rpms/glusterfs
5
0
Fork 0
Code Issues Pull Requests Releases Activity

autobuild v3.12.2-26

Browse Source 
Resolves: bz#1479446 bz#1520882 bz#1579758 bz#1598407 bz#1599808
Resolves: bz#1603118 bz#1619357 bz#1622001 bz#1622308 bz#1631166
Resolves: bz#1631418 bz#1632563 bz#1634649 bz#1635071 bz#1635100
Resolves: bz#1635136 bz#1636291 bz#1638069 bz#1640347 bz#1642854
Resolves: bz#1643035 bz#1644120 bz#1644279 bz#1645916 bz#1647675
Signed-off-by: Milind Changire <mchangir@redhat.com>
This commit is contained in:
Milind Changire 2018-11-08 22:30:35 -05:00
parent 6d7b39fa6f
commit b7dd6f45c1
38 changed files with 9498 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

27
0408-Update-rfc.sh-to-rhgs-3.4.2.patch Normal file
View File

@ -0,0 +1,27 @@
From 1c73105778ac1a7193af747fe9a741690fea010c Mon Sep 17 00:00:00 2001
From: Milind Changire <mchangir@redhat.com>
Date: Fri, 2 Nov 2018 20:56:59 +0530
Subject: [PATCH 408/444] Update rfc.sh to rhgs-3.4.2
Change-Id: Ibe0a72f24c36248f0f637417247e9e359fa23075
Signed-off-by: Milind Changire <mchangir@redhat.com>
---
rfc.sh | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/rfc.sh b/rfc.sh
index 8c4b5ac..5b09aad 100755
--- a/rfc.sh
+++ b/rfc.sh
@@ -17,7 +17,7 @@ done
shift $((OPTIND-1))
-branch="rhgs-3.4.1";
+branch="rhgs-3.4.2";
set_hooks_commit_msg()
{
--
1.8.3.1

50
0409-Update-database-profile-group.patch Normal file
View File

@ -0,0 +1,50 @@
From f56ad2fc0ba3f3b78dc854b0c09c5c8f9bb9db77 Mon Sep 17 00:00:00 2001
From: Xavi Hernandez <xhernandez@redhat.com>
Date: Fri, 21 Sep 2018 21:51:46 +0200
Subject: [PATCH 409/444] Update database profile group
Some performance testing has revealed that pgbench performs 3x better
when these options are set:
performance.client-io-threads=on
performance.open-behind=on
performance.read-after-open=yes
server.event-threads=4
client.event-threads=4
> Upstream patch: https://review.gluster.org/c/glusterfs/+/21247
> Change-Id: I36ce389f893a8af13aac5f8285104d749b73d098
> fixes: bz#1631886
> Signed-off-by: Xavi Hernandez <xhernandez@redhat.com>
Change-Id: I36ce389f893a8af13aac5f8285104d749b73d098
BUG: 1644120
Signed-off-by: Xavi Hernandez <xhernandez@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/154881
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Sunil Kumar Heggodu Gopala Acharya <sheggodu@redhat.com>
---
extras/group-db-workload | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/extras/group-db-workload b/extras/group-db-workload
index c9caf21..9334d6f 100644
--- a/extras/group-db-workload
+++ b/extras/group-db-workload
@@ -1,4 +1,4 @@
-performance.open-behind=off
+performance.open-behind=on
performance.write-behind=off
performance.stat-prefetch=off
performance.quick-read=off
@@ -6,3 +6,7 @@ performance.strict-o-direct=on
performance.read-ahead=off
performance.io-cache=off
performance.readdir-ahead=off
+performance.client-io-threads=on
+server.event-threads=4
+client.event-threads=4
+performance.read-after-open=yes
--
1.8.3.1

42
0410-cli-fix-glusterd-memory-leak-cause-by-gluster-v-stat.patch Normal file
View File

@ -0,0 +1,42 @@
From 4bfbc59a0cbfb28325c16e81480decab003fe6d1 Mon Sep 17 00:00:00 2001
From: shujun10086 <shujun.huang@nokia-sbell.com>
Date: Tue, 2 Oct 2018 08:37:17 +0000
Subject: [PATCH 410/444] cli: fix glusterd memory leak cause by "gluster v
status volume_name"
If use this command every some seconds for example 15s to check gluster brick
status, the glusterd will use about 1G memory in a year. free the value of rsp
in gf_cli_status_cbk. glusterd allocate the value of rsp and send it to cli, but
cli do not free the value, that cause glusterd memory leak.
> fixes: bz#1635480
> Change-Id: I3f19cd0d4b791ae1b35f9664b3a668b1579f1178
> Signed-off-by: shujun10086 <shujun.huang@nokia-sbell.com>
upstream patch: https://review.gluster.org/#/c/21316/
Change-Id: I3f19cd0d4b791ae1b35f9664b3a668b1579f1178
BUG: 1635100
Signed-off-by: Sanju Rakonde <srakonde@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/154882
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Sunil Kumar Heggodu Gopala Acharya <sheggodu@redhat.com>
---
cli/src/cli-rpc-ops.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/cli/src/cli-rpc-ops.c b/cli/src/cli-rpc-ops.c
index 54b61ee65..10f772c 100644
--- a/cli/src/cli-rpc-ops.c
+++ b/cli/src/cli-rpc-ops.c
@@ -8515,6 +8515,7 @@ cont:
ret = rsp.op_ret;
out:
+ FREE(rsp.dict.dict_val);
if (dict)
dict_unref (dict);
GF_FREE (status.brick);
--
1.8.3.1

86
0411-glusterd-ensure-volinfo-caps-is-set-to-correct-value.patch Normal file
View File

@ -0,0 +1,86 @@
From da38c139d41c839244cd5acc0464ddf06fa51c78 Mon Sep 17 00:00:00 2001
From: Sanju Rakonde <srakonde@redhat.com>
Date: Wed, 3 Oct 2018 23:58:37 +0530
Subject: [PATCH 411/444] glusterd: ensure volinfo->caps is set to correct
value
With the commit febf5ed4848, during the volume create op,
we are setting volinfo->caps to 0, only if any of the bricks
belong to the same node and brickinfo->vg[0] is null.
Previously, we used to set volinfo->caps to 0, when
either brick doesn't belong to the same node or brickinfo->vg[0]
is null.
With this patch, we set volinfo->caps to 0, when either brick
doesn't belong to the same node or brickinfo->vg[0] is null.
(as we do earlier without commit febf5ed4848).
> fixes: bz#1635820
> Change-Id: I00a97415786b775fb088ac45566ad52b402f1a49
> Signed-off-by: Sanju Rakonde <srakonde@redhat.com>
upstream patch: https://review.gluster.org/#/c/glusterfs/+/21336/
Change-Id: I00a97415786b775fb088ac45566ad52b402f1a49
BUG: 1635136
Signed-off-by: Sanju Rakonde <srakonde@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/154909
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Sunil Kumar Heggodu Gopala Acharya <sheggodu@redhat.com>
---
.../bug-1636957-peer-reject-on-glusterd-reboot.t | 29 ++++++++++++++++++++++
xlators/mgmt/glusterd/src/glusterd-volume-ops.c | 2 ++
2 files changed, 31 insertions(+)
create mode 100644 tests/bugs/glusterd/bug-1636957-peer-reject-on-glusterd-reboot.t
diff --git a/tests/bugs/glusterd/bug-1636957-peer-reject-on-glusterd-reboot.t b/tests/bugs/glusterd/bug-1636957-peer-reject-on-glusterd-reboot.t
new file mode 100644
index 0000000..b462b38
--- /dev/null
+++ b/tests/bugs/glusterd/bug-1636957-peer-reject-on-glusterd-reboot.t
@@ -0,0 +1,29 @@
+#!/bin/bash
+
+. $(dirname $0)/../../include.rc
+. $(dirname $0)/../../cluster.rc
+. $(dirname $0)/../../volume.rc
+
+function peer_count {
+eval \$CLI_$1 peer status | grep 'Peer in Cluster (Connected)' | wc -l
+}
+
+cleanup
+
+TEST launch_cluster 2
+
+TEST $CLI_1 peer probe $H2;
+EXPECT_WITHIN $PROBE_TIMEOUT 1 peer_count 1
+EXPECT_WITHIN $PROBE_TIMEOUT 1 peer_count 2
+
+TEST $CLI_1 volume create $V0 $H1:$B1/$V0 $H2:$B2/$V0
+
+# rebooting a node which doesn't host bricks for any one volume
+# peer should not go into rejected state
+TEST kill_glusterd 2
+TEST start_glusterd 2
+
+EXPECT_WITHIN $PROBE_TIMEOUT 1 peer_count 1
+EXPECT_WITHIN $PROBE_TIMEOUT 1 peer_count 2
+
+cleanup
diff --git a/xlators/mgmt/glusterd/src/glusterd-volume-ops.c b/xlators/mgmt/glusterd/src/glusterd-volume-ops.c
index 36d9bff..87b7acc 100644
--- a/xlators/mgmt/glusterd/src/glusterd-volume-ops.c
+++ b/xlators/mgmt/glusterd/src/glusterd-volume-ops.c
@@ -2485,6 +2485,8 @@ glusterd_op_create_volume (dict_t *dict, char **op_errstr)
caps = 0;
}
#endif
+ } else {
+ caps = 0;
}
cds_list_add_tail (&brickinfo->brick_list, &volinfo->bricks);
--
1.8.3.1

126
0412-glusterd-set-fsid-while-performing-replace-brick.patch Normal file
View File

@ -0,0 +1,126 @@
From 1e1495a8d5356e6a4f724c211cdd17c5e3f399b5 Mon Sep 17 00:00:00 2001
From: Sanju Rakonde <srakonde@redhat.com>
Date: Tue, 30 Oct 2018 16:36:50 +0530
Subject: [PATCH 412/444] glusterd: set fsid while performing replace brick
While performing the replace-brick operation, we should set
fsid value to the new brick.
> fixes: bz#1637196
> Change-Id: I9e9a4962fc0c2f5dff43e4ac11767814a0c0beaf
> Signed-off-by: Sanju Rakonde <srakonde@redhat.com>
upstream patch: https://review.gluster.org/#/c/glusterfs/+/21513/
Change-Id: I9e9a4962fc0c2f5dff43e4ac11767814a0c0beaf
BUG: 1644279
Signed-off-by: Sanju Rakonde <srakonde@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/154907
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Sunil Kumar Heggodu Gopala Acharya <sheggodu@redhat.com>
---
.../df-results-post-replace-brick-operations.t | 58 ++++++++++++++++++++++
xlators/mgmt/glusterd/src/glusterd-replace-brick.c | 15 ++++++
2 files changed, 73 insertions(+)
create mode 100644 tests/bugs/glusterd/df-results-post-replace-brick-operations.t
diff --git a/tests/bugs/glusterd/df-results-post-replace-brick-operations.t b/tests/bugs/glusterd/df-results-post-replace-brick-operations.t
new file mode 100644
index 0000000..443911c
--- /dev/null
+++ b/tests/bugs/glusterd/df-results-post-replace-brick-operations.t
@@ -0,0 +1,58 @@
+#!/bin/bash
+
+. $(dirname $0)/../../include.rc
+. $(dirname $0)/../../volume.rc
+
+cleanup
+TEST glusterd
+
+#Create brick partitions
+TEST truncate -s 100M $B0/brick1
+TEST truncate -s 100M $B0/brick2
+TEST truncate -s 100M $B0/brick3
+TEST truncate -s 100M $B0/brick4
+TEST truncate -s 100M $B0/brick5
+
+LO1=`SETUP_LOOP $B0/brick1`
+TEST [ $? -eq 0 ]
+TEST MKFS_LOOP $LO1
+
+LO2=`SETUP_LOOP $B0/brick2`
+TEST [ $? -eq 0 ]
+TEST MKFS_LOOP $LO2
+
+LO3=`SETUP_LOOP $B0/brick3`
+TEST [ $? -eq 0 ]
+TEST MKFS_LOOP $LO3
+
+LO4=`SETUP_LOOP $B0/brick4`
+TEST [ $? -eq 0 ]
+TEST MKFS_LOOP $LO4
+
+LO5=`SETUP_LOOP $B0/brick5`
+TEST [ $? -eq 0 ]
+TEST MKFS_LOOP $LO5
+
+TEST mkdir -p $B0/${V0}1 $B0/${V0}2 $B0/${V0}3 $B0/${V0}4 $B0/${V0}5
+TEST MOUNT_LOOP $LO1 $B0/${V0}1
+TEST MOUNT_LOOP $LO2 $B0/${V0}2
+TEST MOUNT_LOOP $LO3 $B0/${V0}3
+TEST MOUNT_LOOP $LO4 $B0/${V0}4
+TEST MOUNT_LOOP $LO5 $B0/${V0}5
+
+# create a subdirectory in mount point and use it for volume creation
+TEST $CLI volume create $V0 replica 3 $H0:$B0/${V0}1/brick1 $H0:$B0/${V0}2/brick1 $H0:$B0/${V0}3/brick1
+TEST $CLI volume start $V0
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "3" online_brick_count
+
+# mount the volume and check the size at mount point
+TEST $GFS --volfile-server=$H0 --volfile-id=$V0 $M0
+total_space=$(df -P $M0 | tail -1 | awk '{ print $2}')
+
+# perform replace brick operations
+TEST $CLI volume replace-brick $V0 $H0:$B0/${V0}1/brick1 $H0:$B0/${V0}4/brick1 commit force
+TEST $CLI volume replace-brick $V0 $H0:$B0/${V0}2/brick1 $H0:$B0/${V0}5/brick1 commit force
+
+# check for the size at mount point, it should be same as previous
+total_space_new=$(df -P $M0 | tail -1 | awk '{ print $2}')
+TEST [ $total_space -eq $total_space_new ]
diff --git a/xlators/mgmt/glusterd/src/glusterd-replace-brick.c b/xlators/mgmt/glusterd/src/glusterd-replace-brick.c
index a037323..5fc3669 100644
--- a/xlators/mgmt/glusterd/src/glusterd-replace-brick.c
+++ b/xlators/mgmt/glusterd/src/glusterd-replace-brick.c
@@ -362,6 +362,7 @@ glusterd_op_perform_replace_brick (glusterd_volinfo_t *volinfo,
int32_t ret = -1;
xlator_t *this = NULL;
glusterd_conf_t *conf = NULL;
+ struct statvfs brickstat = {0,};
this = THIS;
GF_ASSERT (this);
@@ -379,6 +380,20 @@ glusterd_op_perform_replace_brick (glusterd_volinfo_t *volinfo,
ret = glusterd_resolve_brick (new_brickinfo);
if (ret)
goto out;
+ if (!gf_uuid_compare(new_brickinfo->uuid, MY_UUID)) {
+ ret = sys_statvfs(new_brickinfo->path, &brickstat);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, errno, GD_MSG_STATVFS_FAILED,
+ "Failed to fetch disk utilization "
+ "from the brick (%s:%s). Please check the health of "
+ "the brick. Error code was %s",
+ new_brickinfo->hostname, new_brickinfo->path,
+ strerror(errno));
+
+ goto out;
+ }
+ new_brickinfo->statfs_fsid = brickstat.f_fsid;
+ }
ret = glusterd_volume_brickinfo_get_by_brick (old_brick,
volinfo, &old_brickinfo,
--
1.8.3.1

66
0413-glusterfind-add-logs-to-identify-parsing-phases.patch Normal file
View File

@ -0,0 +1,66 @@
From 23dda42f6fa9fe0e6def1b6b0cef8dfdd9a5dcb3 Mon Sep 17 00:00:00 2001
From: Milind Changire <mchangir@redhat.com>
Date: Mon, 5 Nov 2018 15:02:36 +0530
Subject: [PATCH 413/444] glusterfind: add logs to identify parsing phases
Add logs to idenitfy start and finish of changelog parsing phases.
mainline:
> fixes: bz#1632236
> Reviewed-on: https://review.gluster.org/c/glusterfs/+/21262
> Change-Id: Id250231f2af7829f887401d30ac98875ae1ae793
> Signed-off-by: Milind Changire <mchangir@redhat.com>
Change-Id: Id250231f2af7829f887401d30ac98875ae1ae793
BUG: 1631166
Signed-off-by: Milind Changire <mchangir@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/154905
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Sunil Kumar Heggodu Gopala Acharya <sheggodu@redhat.com>
---
tools/glusterfind/src/changelog.py | 10 ++++++++++
1 file changed, 10 insertions(+)
diff --git a/tools/glusterfind/src/changelog.py b/tools/glusterfind/src/changelog.py
index 2376af2..8354cc9 100644
--- a/tools/glusterfind/src/changelog.py
+++ b/tools/glusterfind/src/changelog.py
@@ -273,6 +273,7 @@ def get_changes(brick, hash_dir, log_file, start, end, args):
fail("%s: %s Historical Changelogs not available: %s" %
(args.node, brick, e), logger=logger)
+ logger.info("[1/4] Starting changelog parsing ...")
try:
# scan followed by getchanges till scan returns zero.
# history_scan() is blocking call, till it gets the number
@@ -301,18 +302,27 @@ def get_changes(brick, hash_dir, log_file, start, end, args):
fail("%s Error during Changelog Crawl: %s" % (brick, e),
logger=logger)
+ logger.info("[1/4] Finished changelog parsing.")
+
# Convert all pgfid available from Changelogs
+ logger.info("[2/4] Starting 'pgfid to path' conversions ...")
pgfid_to_path(brick, changelog_data)
changelog_data.commit()
+ logger.info("[2/4] Finished 'pgfid to path' conversions.")
# Convert all GFIDs for which no other additional details available
+ logger.info("[3/4] Starting 'gfid to path using pgfid' conversions ...")
gfid_to_path_using_pgfid(brick, changelog_data, args)
changelog_data.commit()
+ logger.info("[3/4] Finished 'gfid to path using pgfid' conversions.")
# If some GFIDs fail to get converted from previous step,
# convert using find
+ logger.info("[4/4] Starting 'gfid to path using batchfind' "
+ "conversions ...")
gfid_to_path_using_batchfind(brick, changelog_data)
changelog_data.commit()
+ logger.info("[4/4] Finished 'gfid to path using batchfind' conversions.")
return actual_end
--
1.8.3.1

121
0414-logrotate-utilize-the-new-maxsize-option.patch Normal file
View File

@ -0,0 +1,121 @@
From 405a367205c72318fc48d014a201eab3b7031010 Mon Sep 17 00:00:00 2001
From: Amar Tumballi <amarts@redhat.com>
Date: Mon, 5 Nov 2018 10:27:10 +0530
Subject: [PATCH 414/444] logrotate: utilize the new 'maxsize' option
Since logrotate 3.8.x version, a new option 'maxsize' is supported,
which helps in rotating the logs before the specified time if the
size exceeds maxsize limit. This should help in reducing the
overflow of gluster logs.
Upstream:
> URL: https://review.gluster.org/21187
BUG: 1599808
Change-Id: Ic662ada8b73798146736ff81963053d8981745b8
Signed-off-by: Amar Tumballi <amarts@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/154846
Reviewed-by: Sunil Kumar Heggodu Gopala Acharya <sheggodu@redhat.com>
Tested-by: RHGS Build Bot <nigelb@redhat.com>
---
.testignore | 2 ++
extras/glusterfs-georep-logrotate | 24 +++++++++++++++++++++---
extras/glusterfs-logrotate | 14 ++++++++++++--
3 files changed, 35 insertions(+), 5 deletions(-)
diff --git a/.testignore b/.testignore
index 4a72bc4..6e5df3a 100644
--- a/.testignore
+++ b/.testignore
@@ -32,6 +32,8 @@ extras/cliutils/README.md
extras/command-completion/README
extras/create_new_xlator/README.md
extras/glusterfs.vim
+extras/glusterfs-logrotate
+extras/glusterfs-georep-logrotate
extras/group-gluster-block
extras/group-db-workload
extras/group-metadata-cache
diff --git a/extras/glusterfs-georep-logrotate b/extras/glusterfs-georep-logrotate
index 6fdb8c6..3e7ecf3 100644
--- a/extras/glusterfs-georep-logrotate
+++ b/extras/glusterfs-georep-logrotate
@@ -1,6 +1,12 @@
/var/log/glusterfs/geo-replication/*/*.log {
sharedscripts
- rotate 52
+ weekly
+ maxsize 10M
+ minsize 100k
+
+ # 6 months of logs are good enough
+ rotate 26
+
missingok
compress
delaycompress
@@ -15,7 +21,13 @@
/var/log/glusterfs/geo-replication-slaves/*.log {
sharedscripts
- rotate 52
+ weekly
+ maxsize 10M
+ minsize 100k
+
+ # 6 months of logs are good enough
+ rotate 26
+
missingok
compress
delaycompress
@@ -30,7 +42,13 @@
/var/log/glusterfs/geo-replication-slaves/*/*.log {
sharedscripts
- rotate 52
+ weekly
+ maxsize 10M
+ minsize 100k
+
+ # 6 months of logs are good enough
+ rotate 26
+
missingok
compress
delaycompress
diff --git a/extras/glusterfs-logrotate b/extras/glusterfs-logrotate
index 575c0ee..75f700e 100644
--- a/extras/glusterfs-logrotate
+++ b/extras/glusterfs-logrotate
@@ -2,7 +2,12 @@
/var/log/glusterfs/*.log {
sharedscripts
weekly
- rotate 52
+ maxsize 10M
+ minsize 100k
+
+# 6 months of logs are good enough
+ rotate 26
+
missingok
compress
delaycompress
@@ -17,7 +22,12 @@
/var/log/glusterfs/bricks/*.log {
sharedscripts
weekly
- rotate 52
+ maxsize 10M
+ minsize 100k
+
+# 6 months of logs are good enough
+ rotate 26
+
missingok
compress
delaycompress
--
1.8.3.1

39
0415-statedump-fix-clang-null-dereference-error.patch Normal file
View File

@ -0,0 +1,39 @@
From a469cad3a6b7f340c6ac6fad7c2186299d675d70 Mon Sep 17 00:00:00 2001
From: Amar Tumballi <amarts@redhat.com>
Date: Mon, 5 Nov 2018 10:22:44 +0530
Subject: [PATCH 415/444] statedump: fix clang null dereference error
ctx->active can be null, and is checked elsewhere in the
same function. In another case, where 'ctx->active' gets
dereferenced, it needs to be validated before the loop
is hit.
Upstream:
> URL: https://review.gluster.org/21493
BUG: 1643035
Change-Id: I799d92c8089ddbfd9171da4e7e1d77ac91133aba
Signed-off-by: Amar Tumballi <amarts@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/154845
Reviewed-by: Sunil Kumar Heggodu Gopala Acharya <sheggodu@redhat.com>
Tested-by: RHGS Build Bot <nigelb@redhat.com>
---
libglusterfs/src/statedump.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/libglusterfs/src/statedump.c b/libglusterfs/src/statedump.c
index a123adb..a4635f3 100644
--- a/libglusterfs/src/statedump.c
+++ b/libglusterfs/src/statedump.c
@@ -812,7 +812,7 @@ gf_proc_dump_info (int signum, glusterfs_ctx_t *ctx)
if (!ctx)
goto out;
- if (ctx) {
+ if (ctx && ctx->active) {
top = ctx->active->first;
for (trav_p = &top->children; *trav_p;
trav_p = &(*trav_p)->next) {
--
1.8.3.1

63
0416-glusterd-ignore-RPC-events-when-glusterd-is-shutting.patch Normal file
View File

@ -0,0 +1,63 @@
From 04e697b79edd55680a319e6fdb5983a1e5686db9 Mon Sep 17 00:00:00 2001
From: Atin Mukherjee <amukherj@redhat.com>
Date: Wed, 3 Oct 2018 16:34:54 +0530
Subject: [PATCH 416/444] glusterd: ignore RPC events when glusterd is shutting
down
When glusterd receives a SIGTERM while it receives RPC
connect/disconnect/destroy events, the thread might lead to a crash
while accessing rcu_read_lock () as the clean up thread might have
already freed up the resources. This is more observable when glusterd
comes up with upgrade mode = on during upgrade process.
The solution is to ignore these events if glusterd is already in the
middle of cleanup_and_exit ().
> upstream patch : https://review.gluster.org/#/c/glusterfs/+/21330/
>Fixes: bz#1635593
>Change-Id: I12831d31c2f689d4deb038b83b9421bd5cce26d9
>Signed-off-by: Atin Mukherjee <amukherj@redhat.com>
Change-Id: I12831d31c2f689d4deb038b83b9421bd5cce26d9
BUG: 1635071
Signed-off-by: Atin Mukherjee <amukherj@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/154848
Reviewed-by: Sanju Rakonde <srakonde@redhat.com>
Tested-by: Sanju Rakonde <srakonde@redhat.com>
Tested-by: RHGS Build Bot <nigelb@redhat.com>
---
xlators/mgmt/glusterd/src/glusterd-handler.c | 10 +++++++++-
1 file changed, 9 insertions(+), 1 deletion(-)
diff --git a/xlators/mgmt/glusterd/src/glusterd-handler.c b/xlators/mgmt/glusterd/src/glusterd-handler.c
index 861ff17..bf37e70 100644
--- a/xlators/mgmt/glusterd/src/glusterd-handler.c
+++ b/xlators/mgmt/glusterd/src/glusterd-handler.c
@@ -6340,6 +6340,7 @@ __glusterd_peer_rpc_notify (struct rpc_clnt *rpc, void *mydata,
glusterd_peerctx_t *peerctx = NULL;
gf_boolean_t quorum_action = _gf_false;
glusterd_volinfo_t *volinfo = NULL;
+ glusterfs_ctx_t *ctx = NULL;
uuid_t uuid;
peerctx = mydata;
@@ -6355,7 +6356,14 @@ __glusterd_peer_rpc_notify (struct rpc_clnt *rpc, void *mydata,
GF_FREE (peerctx);
return 0;
}
-
+ ctx = this->ctx;
+ GF_VALIDATE_OR_GOTO (this->name, ctx, out);
+ if (ctx->cleanup_started) {
+ gf_log (this->name, GF_LOG_INFO, "glusterd already received a "
+ "SIGTERM, dropping the event %d for peer %s", event,
+ peerctx->peername);
+ return 0;
+ }
rcu_read_lock ();
peerinfo = glusterd_peerinfo_find_by_generation (peerctx->peerinfo_gen);
--
1.8.3.1

190
0417-cli-Add-warning-message-while-converting-to-replica-.patch Normal file
View File

@ -0,0 +1,190 @@
From 1a24a7942fe9ecccaf29ae9bc125cd9b08fc8906 Mon Sep 17 00:00:00 2001
From: karthik-us <ksubrahm@redhat.com>
Date: Mon, 5 Nov 2018 17:57:55 +0530
Subject: [PATCH 417/444] cli: Add warning message while converting to replica
2 configuration
Backport of: https://review.gluster.org/#/c/glusterfs/+/21136/
Currently while creating replica 2 volume we display a warning message
of ending up in split-brain. But while converting an existing volume
from other configuration to replica 2 by add-brick or remove-brick
operations we do not show any such messages.
With this fix in add-brick and remove-brick cases also we will display
the same warning message and prompt for confirmation if the configuration
changes to replica 2.
Change-Id: Id7b1a40e80fca3e1043b802fa5f7c3b656ef2228
BUG: 1579758
Signed-off-by: karthik-us <ksubrahm@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/154947
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Ravishankar Narayanankutty <ravishankar@redhat.com>
Reviewed-by: Atin Mukherjee <amukherj@redhat.com>
---
cli/src/cli-cmd-parser.c | 48 +++++++++++++++++++++++++++++++++++++++++++-----
cli/src/cli-cmd-volume.c | 11 +++++++----
cli/src/cli.h | 10 +++++-----
3 files changed, 55 insertions(+), 14 deletions(-)
diff --git a/cli/src/cli-cmd-parser.c b/cli/src/cli-cmd-parser.c
index 7917d66..3745fb4 100644
--- a/cli/src/cli-cmd-parser.c
+++ b/cli/src/cli-cmd-parser.c
@@ -1774,8 +1774,8 @@ out:
}
int32_t
-cli_cmd_volume_add_brick_parse (const char **words, int wordcount,
- dict_t **options, int *ret_type)
+cli_cmd_volume_add_brick_parse (struct cli_state *state, const char **words,
+ int wordcount, dict_t **options, int *ret_type)
{
dict_t *dict = NULL;
char *volname = NULL;
@@ -1790,6 +1790,8 @@ cli_cmd_volume_add_brick_parse (const char **words, int wordcount,
int index;
gf_boolean_t is_force = _gf_false;
int wc = wordcount;
+ gf_answer_t answer = GF_ANSWER_NO;
+ const char *question = NULL;
GF_ASSERT (words);
GF_ASSERT (options);
@@ -1854,6 +1856,23 @@ cli_cmd_volume_add_brick_parse (const char **words, int wordcount,
goto out;
index = 7;
}
+
+ if (count == 2) {
+ if (strcmp (words[wordcount - 1], "force")) {
+ question = "Replica 2 volumes are prone to "
+ "split-brain. Use Arbiter or "
+ "Replica 3 to avoid this.\n"
+ "Do you still want to continue?\n";
+ answer = cli_cmd_get_confirmation (state,
+ question);
+ if (GF_ANSWER_NO == answer) {
+ gf_log ("cli", GF_LOG_ERROR, "Add brick"
+ " cancelled, exiting");
+ ret = -1;
+ goto out;
+ }
+ }
+ }
} else if ((strcmp (w, "stripe")) == 0) {
type = GF_CLUSTER_TYPE_STRIPE;
count = strtol (words[4], NULL, 0);
@@ -2061,9 +2080,9 @@ out:
}
int32_t
-cli_cmd_volume_remove_brick_parse (const char **words, int wordcount,
- dict_t **options, int *question,
- int *brick_count)
+cli_cmd_volume_remove_brick_parse (struct cli_state *state, const char **words,
+ int wordcount, dict_t **options,
+ int *question, int *brick_count)
{
dict_t *dict = NULL;
char *volname = NULL;
@@ -2081,6 +2100,8 @@ cli_cmd_volume_remove_brick_parse (const char **words, int wordcount,
char *w = NULL;
int32_t command = GF_OP_CMD_NONE;
long count = 0;
+ gf_answer_t answer = GF_ANSWER_NO;
+ const char *ques = NULL;
GF_ASSERT (words);
GF_ASSERT (options);
@@ -2115,6 +2136,23 @@ cli_cmd_volume_remove_brick_parse (const char **words, int wordcount,
goto out;
}
+ if (count == 2) {
+ if (strcmp (words[wordcount - 1], "force")) {
+ ques = "Replica 2 volumes are prone to "
+ "split-brain. Use Arbiter or Replica 3 "
+ "to avoid this.\n"
+ "Do you still want to continue?\n";
+ answer = cli_cmd_get_confirmation (state,
+ ques);
+ if (GF_ANSWER_NO == answer) {
+ gf_log ("cli", GF_LOG_ERROR, "Remove "
+ "brick cancelled, exiting");
+ ret = -1;
+ goto out;
+ }
+ }
+ }
+
ret = dict_set_int32 (dict, "replica-count", count);
if (ret)
goto out;
diff --git a/cli/src/cli-cmd-volume.c b/cli/src/cli-cmd-volume.c
index a1f0840..32efa73 100644
--- a/cli/src/cli-cmd-volume.c
+++ b/cli/src/cli-cmd-volume.c
@@ -1021,7 +1021,8 @@ cli_cmd_volume_add_brick_cbk (struct cli_state *state,
if (!frame)
goto out;
- ret = cli_cmd_volume_add_brick_parse (words, wordcount, &options, 0);
+ ret = cli_cmd_volume_add_brick_parse (state, words, wordcount, &options,
+ 0);
if (ret) {
cli_usage_out (word->pattern);
parse_error = 1;
@@ -1151,7 +1152,8 @@ do_cli_cmd_volume_attach_tier (struct cli_state *state,
if (!frame)
goto out;
- ret = cli_cmd_volume_add_brick_parse (words, wordcount, &options, &type);
+ ret = cli_cmd_volume_add_brick_parse (state, words, wordcount, &options,
+ &type);
if (ret) {
cli_usage_out (word->pattern);
parse_error = 1;
@@ -2032,8 +2034,9 @@ cli_cmd_volume_remove_brick_cbk (struct cli_state *state,
if (!frame)
goto out;
- ret = cli_cmd_volume_remove_brick_parse (words, wordcount, &options,
- &need_question, &brick_count);
+ ret = cli_cmd_volume_remove_brick_parse (state, words, wordcount,
+ &options, &need_question,
+ &brick_count);
if (ret) {
cli_usage_out (word->pattern);
parse_error = 1;
diff --git a/cli/src/cli.h b/cli/src/cli.h
index c9bf93d..109dcd4 100644
--- a/cli/src/cli.h
+++ b/cli/src/cli.h
@@ -264,8 +264,8 @@ cli_cmd_get_state_parse (struct cli_state *state, const char **words,
int wordcount, dict_t **options, char **op_errstr);
int32_t
-cli_cmd_volume_add_brick_parse (const char **words, int wordcount,
- dict_t **options, int *type);
+cli_cmd_volume_add_brick_parse (struct cli_state *state, const char **words,
+ int wordcount, dict_t **options, int *type);
int32_t
cli_cmd_volume_detach_tier_parse (const char **words, int wordcount,
@@ -280,9 +280,9 @@ cli_cmd_volume_old_tier_parse (const char **words, int wordcount,
dict_t **options);
int32_t
-cli_cmd_volume_remove_brick_parse (const char **words, int wordcount,
- dict_t **options, int *question,
- int *brick_count);
+cli_cmd_volume_remove_brick_parse (struct cli_state *state, const char **words,
+ int wordcount, dict_t **options,
+ int *question, int *brick_count);
int32_t
cli_cmd_volume_replace_brick_parse (const char **words, int wordcount,
--
1.8.3.1

58
0418-cli-correct-rebalance-status-elapsed-check.patch Normal file
View File

@ -0,0 +1,58 @@
From 414d6d378b7d63b172859f619bd3ffb72bd3f434 Mon Sep 17 00:00:00 2001
From: N Balachandran <nbalacha@redhat.com>
Date: Tue, 8 Aug 2017 23:11:10 +0530
Subject: [PATCH 418/444] cli: correct rebalance status elapsed check
Check that elapsed time has crossed 10 mins for at least
one rebalance process before displaying the estimates.
Upstream patch: https://review.gluster.org/#/c/glusterfs/+/18000/
> Change-Id: Ib357a6f0d0125a178e94ede1e31514fdc6ce3593
> BUG: 1479528
> Signed-off-by: N Balachandran <nbalacha@redhat.com>
Change-Id: Ic4606acad991b9369c6b674691e0ec15621c6932
BUG: 1479446
Signed-off-by: N Balachandran <nbalacha@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/154929
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Atin Mukherjee <amukherj@redhat.com>
---
cli/src/cli-rpc-ops.c | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/cli/src/cli-rpc-ops.c b/cli/src/cli-rpc-ops.c
index 10f772c..5623950 100644
--- a/cli/src/cli-rpc-ops.c
+++ b/cli/src/cli-rpc-ops.c
@@ -1616,6 +1616,7 @@ gf_cli_print_rebalance_status (dict_t *dict, enum gf_task_types task_type,
gf_boolean_t down = _gf_false;
gf_boolean_t fix_layout = _gf_false;
uint64_t max_time = 0;
+ uint64_t max_elapsed = 0;
uint64_t time_left = 0;
gf_boolean_t show_estimates = _gf_false;
@@ -1758,6 +1759,9 @@ gf_cli_print_rebalance_status (dict_t *dict, enum gf_task_types task_type,
gf_log ("cli", GF_LOG_TRACE,
"failed to get time left");
+ if (elapsed > max_elapsed)
+ max_elapsed = elapsed;
+
if (time_left > max_time)
max_time = time_left;
@@ -1818,7 +1822,7 @@ gf_cli_print_rebalance_status (dict_t *dict, enum gf_task_types task_type,
if (!show_estimates) {
goto out;
}
- if (elapsed <= REBAL_ESTIMATE_START_TIME) {
+ if (max_elapsed <= REBAL_ESTIMATE_START_TIME) {
cli_out ("The estimated time for rebalance to complete "
"will be unavailable for the first 10 "
"minutes.");
--
1.8.3.1

84
0419-glusterfs-During-reconfigure-set-log-level-per-xlato.patch Normal file
View File

@ -0,0 +1,84 @@
From 56fb13d05cb4465c14cc231bab1296a48c33c57d Mon Sep 17 00:00:00 2001
From: Mohit Agrawal <moagrawa@redhat.com>
Date: Tue, 6 Nov 2018 09:06:34 +0530
Subject: [PATCH 419/444] glusterfs: During reconfigure set log-level per
xlator level
Problem: In brick_mux environment, while a user has enabled brick-log-level
for anyone volume, it automatically enables for other volumes
also those are attached with same brick.
Solution: A log-level option is automatically enabled for other volumes
because log-level saved in glusterfsd_ctx and ctx is common for
volumes those are attached with same brick. To resolve it
set log level for all children xlator's at the time of the graph
reconfigure at io-stat xlator.
> Change-Id: Id9a6efa05d286e0bea2d47f49292d084e7bb2fcf
> fixes: bz#1640495
> (Reviwed on upstream link https://review.gluster.org/#/c/glusterfs/+/20488/)
> (Cherry pick from commit c34e4161f3cb6539ec83a9020f3d27eb4759a975)
Change-Id: I1dd57c52997f16e8a05f982c6c05bb4f758e8bd3
BUG: 1598407
Signed-off-by: Mohit Agrawal <moagrawa@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/155021
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Sunil Kumar Heggodu Gopala Acharya <sheggodu@redhat.com>
---
xlators/debug/io-stats/src/io-stats.c | 32 +++++++++++++++++++++++++++++++-
1 file changed, 31 insertions(+), 1 deletion(-)
diff --git a/xlators/debug/io-stats/src/io-stats.c b/xlators/debug/io-stats/src/io-stats.c
index 0f71334..aade097 100644
--- a/xlators/debug/io-stats/src/io-stats.c
+++ b/xlators/debug/io-stats/src/io-stats.c
@@ -3812,6 +3812,35 @@ ios_set_log_format_code (struct ios_conf *conf)
conf->dump_format = IOS_DUMP_TYPE_SAMPLES;
}
+void
+xlator_set_loglevel(xlator_t *this, int log_level)
+{
+ glusterfs_ctx_t *ctx = NULL;
+ glusterfs_graph_t *active = NULL;
+ xlator_t *top = NULL;
+ xlator_t *trav = this;
+
+ ctx = this->ctx;
+ GF_ASSERT(ctx);
+ active = ctx->active;
+ top = active->first;
+
+ if (strcmp(top->type, "protocol/server") || (log_level == -1))
+ return;
+
+ /* Set log-level for server xlator */
+ top->loglevel = log_level;
+
+ /* Set log-level for parent xlator */
+ if (this->parents)
+ this->parents->xlator->loglevel = log_level;
+
+ while (trav) {
+ trav->loglevel = log_level;
+ trav = trav->next;
+ }
+}
+
int
reconfigure (xlator_t *this, dict_t *options)
{
@@ -3867,7 +3896,8 @@ reconfigure (xlator_t *this, dict_t *options)
GF_OPTION_RECONF ("log-level", log_str, options, str, out);
if (log_str) {
log_level = glusterd_check_log_level (log_str);
- gf_log_set_loglevel (log_level);
+ /* Set loglevel for all children and server xlators */
+ xlator_set_loglevel(this, log_level);
}
GF_OPTION_RECONF ("logger", logger_str, options, str, out);
--
1.8.3.1

40
0420-Modify-log-message-DH-ciphers-are-disabled-from-ERRO.patch Normal file
View File

@ -0,0 +1,40 @@
From 9a6ad46e3d7ae9ac683ef790c12937fee8f1143c Mon Sep 17 00:00:00 2001
From: Mohit Agrawal <moagrawa@redhat.com>
Date: Tue, 6 Nov 2018 09:31:50 +0530
Subject: [PATCH 420/444] Modify log message 'DH ciphers are disabled' from
ERROR to INFO
Per the latest comment in bz#1398237 this message is confusing for users
because it suggests an error where none exists.
> Fixes: bz#1626319
> Change-Id: I2f05999da157b11e225bf3d95edb597e964f9923
> Signed-off-by: Omar Kohl <omarkohl@gmail.com>
> (Reviewed on upstream link https://review.gluster.org/#/c/glusterfs/+/21108/)
Change-Id: I154cdd6e33e17d426bcba10fe17fceceba047b16
BUG: 1632563
Signed-off-by: Mohit Agrawal <moagrawa@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/155023
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Sunil Kumar Heggodu Gopala Acharya <sheggodu@redhat.com>
---
rpc/rpc-transport/socket/src/socket.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/rpc/rpc-transport/socket/src/socket.c b/rpc/rpc-transport/socket/src/socket.c
index 243d49c..8a08177 100644
--- a/rpc/rpc-transport/socket/src/socket.c
+++ b/rpc/rpc-transport/socket/src/socket.c
@@ -4378,7 +4378,7 @@ socket_init (rpc_transport_t *this)
#endif
if ((bio = BIO_new_file(dh_param, "r")) == NULL) {
- gf_log(this->name,GF_LOG_ERROR,
+ gf_log(this->name, GF_LOG_INFO,
"failed to open %s, "
"DH ciphers are disabled", dh_param);
}
--
1.8.3.1

66
0421-rpc-handle-EAGAIN-when-SSL_ERROR_SYSCALL-is-returned.patch Normal file
View File

@ -0,0 +1,66 @@
From ce2c9ea016ffa20bf291264a012cc14102040900 Mon Sep 17 00:00:00 2001
From: Milind Changire <mchangir@redhat.com>
Date: Mon, 10 Sep 2018 13:48:18 +0530
Subject: [PATCH 421/444] rpc: handle EAGAIN when SSL_ERROR_SYSCALL is returned
Problem:
A return value of ENODATA was forcibly returned in the case where
SSL_get_error(r) returned SSL_ERROR_SYSCALL. Sometimes SSL_ERROR_SYSCALL
is a transient error which is identified by setting errno to EAGAIN.
EAGAIN is not a fatal error and indicates that the syscall needs to be
retried.
Solution:
Bubble up the errno in case SSL_get_error(r) returns SSL_ERROR_SYSCALL
and let the upper layers handle it appropriately.
mainline:
> Reviewed-on: https://review.gluster.org/c/glusterfs/+/20993
> fixes: bz#1622405
> Change-Id: I76eff278378930ee79abbf9fa267a7e77356eed6
> BUG: 1622405
Change-Id: I76eff278378930ee79abbf9fa267a7e77356eed6
BUG: 1622308
Signed-off-by: Milind Changire <mchangir@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/154868
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Sunil Kumar Heggodu Gopala Acharya <sheggodu@redhat.com>
---
rpc/rpc-transport/socket/src/socket.c | 11 +++++++++--
1 file changed, 9 insertions(+), 2 deletions(-)
diff --git a/rpc/rpc-transport/socket/src/socket.c b/rpc/rpc-transport/socket/src/socket.c
index 8a08177..34a937f 100644
--- a/rpc/rpc-transport/socket/src/socket.c
+++ b/rpc/rpc-transport/socket/src/socket.c
@@ -209,6 +209,7 @@ ssl_do (rpc_transport_t *this, void *buf, size_t len, SSL_trinary_func *func)
int r = (-1);
struct pollfd pfd = {-1,};
socket_private_t *priv = NULL;
+ int myerrno = -1;
GF_VALIDATE_OR_GOTO(this->name,this->private,out);
priv = this->private;
@@ -276,10 +277,16 @@ ssl_do (rpc_transport_t *this, void *buf, size_t len, SSL_trinary_func *func)
}
break;
case SSL_ERROR_SYSCALL:
+ myerrno = errno;
/* This is what we get when remote disconnects. */
gf_log(this->name,GF_LOG_DEBUG,
- "syscall error (probably remote disconnect)");
- errno = ENODATA;
+ "syscall error (probably remote disconnect)"
+ " errno:%d(%s)", errno, strerror(errno));
+ /* sometimes, errno is set to EAGAIN in this case
+ * so let the upper layers do what they need to do
+ * with it
+ */
+ errno = myerrno;
goto out;
default:
errno = EIO;
--
1.8.3.1

46
0422-glusterd-raise-default-transport.listen-backlog.patch Normal file
View File

@ -0,0 +1,46 @@
From ccac7336bb6fa667b4f9b51426440d898ff3d184 Mon Sep 17 00:00:00 2001
From: Milind Changire <mchangir@redhat.com>
Date: Mon, 5 Nov 2018 19:38:08 +0530
Subject: [PATCH 422/444] glusterd: raise default transport.listen-backlog
Problem:
data center setups with large number of bricks with replication
causes a flood of connections from bricks and self-heal daemons
to glusterd causing connections to be dropped due to insufficient
listener socket backlog queue length
Solution:
raise default value of transport.listen-backlog to 1024
mainline:
> Reviewed-on: https://review.gluster.org/c/glusterfs/+/21482
> Change-Id: I879e4161a88f1e30875046dff232499a8e2e6c51
> fixes: bz#1642850
> Signed-off-by: Milind Changire <mchangir@redhat.com>
Change-Id: I879e4161a88f1e30875046dff232499a8e2e6c51
BUG: 1642854
Signed-off-by: Milind Changire <mchangir@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/154959
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Sunil Kumar Heggodu Gopala Acharya <sheggodu@redhat.com>
---
libglusterfs/src/glusterfs.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/libglusterfs/src/glusterfs.h b/libglusterfs/src/glusterfs.h
index d71a9c1..5e641fd 100644
--- a/libglusterfs/src/glusterfs.h
+++ b/libglusterfs/src/glusterfs.h
@@ -73,7 +73,7 @@
#endif
#define GLUSTERD_MAX_SNAP_NAME 255
-#define GLUSTERFS_SOCKET_LISTEN_BACKLOG 10
+#define GLUSTERFS_SOCKET_LISTEN_BACKLOG 1024
#define ZR_MOUNTPOINT_OPT "mountpoint"
#define ZR_ATTR_TIMEOUT_OPT "attribute-timeout"
#define ZR_ENTRY_TIMEOUT_OPT "entry-timeout"
--
1.8.3.1

150
0423-glusterd-acquire-lock-to-update-volinfo-structure.patch Normal file
View File

@ -0,0 +1,150 @@
From 216ac7a1bd22db08cc02d7b8688a3338e78c71cd Mon Sep 17 00:00:00 2001
From: Sanju Rakonde <srakonde@redhat.com>
Date: Tue, 11 Sep 2018 14:19:42 +0530
Subject: [PATCH 423/444] glusterd: acquire lock to update volinfo structure
Problem: With commit cb0339f92, we are using a separate syntask
for restart_bricks. There can be a situation where two threads
are accessing the same volinfo structure at the same time and
updating volinfo structure. This can lead volinfo to have
inconsistent values and assertion failures because of unexpected
values.
Solution: While updating the volinfo structure, acquire a
store_volinfo_lock, and release the lock only when the thread
completed its critical section part.
> Fixes: bz#1627610
> Signed-off-by: Sanju Rakonde <srakonde@redhat.com>
> Change-Id: I545e4e2368e3285d8f7aa28081ff4448abb72f5d
upstream patch: https://review.gluster.org/#/c/glusterfs/+/21150/
Change-Id: I545e4e2368e3285d8f7aa28081ff4448abb72f5d
BUG: 1631418
Signed-off-by: Sanju Rakonde <srakonde@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/154885
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Atin Mukherjee <amukherj@redhat.com>
---
xlators/mgmt/glusterd/src/glusterd-store.c | 67 +++++++++++++------------
xlators/mgmt/glusterd/src/glusterd-volume-ops.c | 2 +
xlators/mgmt/glusterd/src/glusterd.h | 3 ++
3 files changed, 40 insertions(+), 32 deletions(-)
diff --git a/xlators/mgmt/glusterd/src/glusterd-store.c b/xlators/mgmt/glusterd/src/glusterd-store.c
index 015f6c2..37542e7 100644
--- a/xlators/mgmt/glusterd/src/glusterd-store.c
+++ b/xlators/mgmt/glusterd/src/glusterd-store.c
@@ -1796,46 +1796,49 @@ glusterd_store_volinfo (glusterd_volinfo_t *volinfo, glusterd_volinfo_ver_ac_t a
GF_ASSERT (volinfo);
- glusterd_perform_volinfo_version_action (volinfo, ac);
- ret = glusterd_store_create_volume_dir (volinfo);
- if (ret)
- goto out;
-
- ret = glusterd_store_create_volume_run_dir (volinfo);
- if (ret)
- goto out;
+ pthread_mutex_lock(&volinfo->store_volinfo_lock);
+ {
+ glusterd_perform_volinfo_version_action(volinfo, ac);
+ ret = glusterd_store_create_volume_dir(volinfo);
+ if (ret)
+ goto unlock;
- ret = glusterd_store_create_vol_shandle_on_absence (volinfo);
- if (ret)
- goto out;
+ ret = glusterd_store_create_volume_run_dir(volinfo);
+ if (ret)
+ goto unlock;
- ret = glusterd_store_create_nodestate_sh_on_absence (volinfo);
- if (ret)
- goto out;
+ ret = glusterd_store_create_vol_shandle_on_absence(volinfo);
+ if (ret)
+ goto unlock;
- ret = glusterd_store_perform_volume_store (volinfo);
- if (ret)
- goto out;
+ ret = glusterd_store_create_nodestate_sh_on_absence(volinfo);
+ if (ret)
+ goto unlock;
- ret = glusterd_store_volume_atomic_update (volinfo);
- if (ret) {
- glusterd_perform_volinfo_version_action (volinfo,
- GLUSTERD_VOLINFO_VER_AC_DECREMENT);
- goto out;
- }
+ ret = glusterd_store_perform_volume_store(volinfo);
+ if (ret)
+ goto unlock;
- ret = glusterd_store_perform_node_state_store (volinfo);
- if (ret)
- goto out;
+ ret = glusterd_store_volume_atomic_update(volinfo);
+ if (ret) {
+ glusterd_perform_volinfo_version_action(volinfo,
+ GLUSTERD_VOLINFO_VER_AC_DECREMENT);
+ goto unlock;
+ }
- /* checksum should be computed at the end */
- ret = glusterd_compute_cksum (volinfo, _gf_false);
- if (ret)
- goto out;
+ ret = glusterd_store_perform_node_state_store(volinfo);
+ if (ret)
+ goto unlock;
-out:
+ /* checksum should be computed at the end */
+ ret = glusterd_compute_cksum(volinfo, _gf_false);
+ if (ret)
+ goto unlock;
+ }
+unlock:
+ pthread_mutex_unlock(&volinfo->store_volinfo_lock);
if (ret)
- glusterd_store_volume_cleanup_tmp (volinfo);
+ glusterd_store_volume_cleanup_tmp(volinfo);
gf_msg_debug (THIS->name, 0, "Returning %d", ret);
diff --git a/xlators/mgmt/glusterd/src/glusterd-volume-ops.c b/xlators/mgmt/glusterd/src/glusterd-volume-ops.c
index 87b7acc..b91a516 100644
--- a/xlators/mgmt/glusterd/src/glusterd-volume-ops.c
+++ b/xlators/mgmt/glusterd/src/glusterd-volume-ops.c
@@ -2198,6 +2198,8 @@ glusterd_op_create_volume (dict_t *dict, char **op_errstr)
goto out;
}
+ pthread_mutex_init(&volinfo->store_volinfo_lock, NULL);
+
ret = dict_get_str (dict, "volname", &volname);
if (ret) {
diff --git a/xlators/mgmt/glusterd/src/glusterd.h b/xlators/mgmt/glusterd/src/glusterd.h
index 8c70d48..edd41aa 100644
--- a/xlators/mgmt/glusterd/src/glusterd.h
+++ b/xlators/mgmt/glusterd/src/glusterd.h
@@ -478,6 +478,9 @@ struct glusterd_volinfo_ {
gf_boolean_t stage_deleted; /* volume has passed staging
* for delete operation
*/
+ pthread_mutex_t store_volinfo_lock; /* acquire lock for
+ * updating the volinfo
+ */
};
typedef enum gd_snap_status_ {
--
1.8.3.1

272
0424-cluster-afr-Delegate-metadata-heal-with-pending-xatt.patch Normal file
View File

@ -0,0 +1,272 @@
From 68b0db385ce968547349b187222b9a9401faee12 Mon Sep 17 00:00:00 2001
From: Pranith Kumar K <pkarampu@redhat.com>
Date: Mon, 27 Aug 2018 11:46:33 +0530
Subject: [PATCH 424/444] cluster/afr: Delegate metadata heal with pending
xattrs to SHD
Problem:
When metadata-self-heal is triggered on the mount, it blocks
lookup until metadata-self-heal completes. But that can lead
to hangs when lot of clients are accessing a directory which
needs metadata heal and all of them trigger heals waiting
for other clients to complete heal.
Fix:
Only when the heal is needed but the pending xattrs are not set,
trigger metadata heal that could block lookup. This is the only
case where different clients may give different metadata to the
clients without heals, which should be avoided.
BUG: 1619357
Upstream Patch: https://review.gluster.org/c/glusterfs/+/21086
Change-Id: I6089e9fda0770a83fb287941b229c882711f4e66
Signed-off-by: Pranith Kumar K <pkarampu@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/155028
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Sunil Kumar Heggodu Gopala Acharya <sheggodu@redhat.com>
---
tests/basic/afr/client-side-heal.t | 28 ++++++++++------
tests/bugs/glusterfs/bug-906646.t | 10 ++++--
xlators/cluster/afr/src/afr-common.c | 44 ++++++++++++++++++++++++++
xlators/cluster/afr/src/afr-self-heal-common.c | 38 ----------------------
xlators/cluster/afr/src/afr.h | 3 ++
5 files changed, 72 insertions(+), 51 deletions(-)
diff --git a/tests/basic/afr/client-side-heal.t b/tests/basic/afr/client-side-heal.t
index eba7dc2..1e93361 100755
--- a/tests/basic/afr/client-side-heal.t
+++ b/tests/basic/afr/client-side-heal.t
@@ -17,6 +17,7 @@ TEST $GFS --volfile-id=$V0 --volfile-server=$H0 $M0;
echo "some data" > $M0/datafile
EXPECT 0 echo $?
TEST touch $M0/mdatafile
+TEST touch $M0/mdatafile-backend-direct-modify
TEST mkdir $M0/dir
#Kill a brick and perform I/O to have pending heals.
@@ -29,6 +30,7 @@ EXPECT 0 echo $?
#pending metadata heal
TEST chmod +x $M0/mdatafile
+TEST chmod +x $B0/${V0}0/mdatafile-backend-direct-modify
#pending entry heal. Also causes pending metadata/data heals on file{1..5}
TEST touch $M0/dir/file{1..5}
@@ -40,9 +42,12 @@ TEST $CLI volume start $V0 force
EXPECT_WITHIN $PROCESS_UP_TIMEOUT "1" afr_child_up_status $V0 0
#Medatada heal via explicit lookup must not happen
-TEST ls $M0/mdatafile
+TEST getfattr -d -m. -e hex $M0/mdatafile
+TEST ls $M0/mdatafile-backend-direct-modify
-#Inode refresh must not trigger data and entry heals.
+TEST [[ "$(stat -c %A $B0/${V0}0/mdatafile-backend-direct-modify)" != "$(stat -c %A $B0/${V0}1/mdatafile-backend-direct-modify)" ]]
+
+#Inode refresh must not trigger data metadata and entry heals.
#To trigger inode refresh for sure, the volume is unmounted and mounted each time.
#Check that data heal does not happen.
EXPECT_WITHIN $UMOUNT_TIMEOUT "Y" force_umount $M0
@@ -52,7 +57,6 @@ TEST cat $M0/datafile
EXPECT_WITHIN $UMOUNT_TIMEOUT "Y" force_umount $M0
TEST $GFS --volfile-id=$V0 --volfile-server=$H0 $M0;
TEST ls $M0/dir
-
#No heal must have happened
EXPECT 8 get_pending_heal_count $V0
@@ -61,21 +65,25 @@ TEST $CLI volume set $V0 cluster.data-self-heal on
TEST $CLI volume set $V0 cluster.metadata-self-heal on
TEST $CLI volume set $V0 cluster.entry-self-heal on
-#Metadata heal is triggered by lookup without need for inode refresh.
-TEST ls $M0/mdatafile
-EXPECT 7 get_pending_heal_count $V0
-
-#Inode refresh must trigger data and entry heals.
+#Inode refresh must trigger data metadata and entry heals.
#To trigger inode refresh for sure, the volume is unmounted and mounted each time.
EXPECT_WITHIN $UMOUNT_TIMEOUT "Y" force_umount $M0
TEST $GFS --volfile-id=$V0 --volfile-server=$H0 $M0;
+TEST ls $M0/mdatafile-backend-direct-modify
+
+TEST [[ "$(stat -c %A $B0/${V0}0/mdatafile-backend-direct-modify)" == "$(stat -c %A $B0/${V0}1/mdatafile-backend-direct-modify)" ]]
+
+
+TEST getfattr -d -m. -e hex $M0/mdatafile
+EXPECT_WITHIN $HEAL_TIMEOUT 7 get_pending_heal_count $V0
+
TEST cat $M0/datafile
EXPECT_WITHIN $HEAL_TIMEOUT 6 get_pending_heal_count $V0
EXPECT_WITHIN $UMOUNT_TIMEOUT "Y" force_umount $M0
TEST $GFS --volfile-id=$V0 --volfile-server=$H0 $M0;
TEST ls $M0/dir
-EXPECT 5 get_pending_heal_count $V0
+EXPECT_WITHIN $HEAL_TIMEOUT 5 get_pending_heal_count $V0
TEST cat $M0/dir/file1
TEST cat $M0/dir/file2
@@ -83,5 +91,5 @@ TEST cat $M0/dir/file3
TEST cat $M0/dir/file4
TEST cat $M0/dir/file5
-EXPECT 0 get_pending_heal_count $V0
+EXPECT_WITHIN $HEAL_TIMEOUT 0 get_pending_heal_count $V0
cleanup;
diff --git a/tests/bugs/glusterfs/bug-906646.t b/tests/bugs/glusterfs/bug-906646.t
index 45c85d9..37b8fe5 100644
--- a/tests/bugs/glusterfs/bug-906646.t
+++ b/tests/bugs/glusterfs/bug-906646.t
@@ -13,7 +13,6 @@ TEST pidof glusterd
TEST $CLI volume create $V0 replica $REPLICA $H0:$B0/${V0}-00 $H0:$B0/${V0}-01 $H0:$B0/${V0}-10 $H0:$B0/${V0}-11
TEST $CLI volume start $V0
-TEST $CLI volume set $V0 cluster.self-heal-daemon off
TEST $CLI volume set $V0 cluster.background-self-heal-count 0
## Mount FUSE with caching disabled
@@ -82,10 +81,15 @@ EXPECT 1 xattr_query_check ${backend_paths_array[1]} "trusted.name"
# restart the brick process
TEST $CLI volume start $V0 force
-EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status $V0 `expr $brick_id - 1`
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "Y" glustershd_up_status
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status_in_shd $V0 0
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status_in_shd $V0 1
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status_in_shd $V0 2
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status_in_shd $V0 3
-cat $pth >/dev/null
+TEST $CLI volume heal $V0
+EXPECT_WITHIN $HEAL_TIMEOUT "0" get_pending_heal_count $V0
# check backends - xattr should not be present anywhere
EXPECT 1 xattr_query_check ${backend_paths_array[0]} "trusted.name"
EXPECT 1 xattr_query_check ${backend_paths_array[1]} "trusted.name"
diff --git a/xlators/cluster/afr/src/afr-common.c b/xlators/cluster/afr/src/afr-common.c
index e8107c9..e74fdec 100644
--- a/xlators/cluster/afr/src/afr-common.c
+++ b/xlators/cluster/afr/src/afr-common.c
@@ -2571,6 +2571,42 @@ out:
return 0;
}
+gf_boolean_t
+afr_is_pending_set (xlator_t *this, dict_t *xdata, int type)
+{
+ int idx = -1;
+ afr_private_t *priv = NULL;
+ void *pending_raw = NULL;
+ int *pending_int = NULL;
+ int i = 0;
+
+ priv = this->private;
+ idx = afr_index_for_transaction_type (type);
+
+ if (dict_get_ptr (xdata, AFR_DIRTY, &pending_raw) == 0) {
+ if (pending_raw) {
+ pending_int = pending_raw;
+
+ if (ntoh32 (pending_int[idx]))
+ return _gf_true;
+ }
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (dict_get_ptr (xdata, priv->pending_key[i],
+ &pending_raw))
+ continue;
+ if (!pending_raw)
+ continue;
+ pending_int = pending_raw;
+
+ if (ntoh32 (pending_int[idx]))
+ return _gf_true;
+ }
+
+ return _gf_false;
+}
+
static gf_boolean_t
afr_can_start_metadata_self_heal(call_frame_t *frame, xlator_t *this)
{
@@ -2597,6 +2633,14 @@ afr_can_start_metadata_self_heal(call_frame_t *frame, xlator_t *this)
continue;
}
+ if (afr_is_pending_set (this, replies[i].xdata,
+ AFR_METADATA_TRANSACTION)) {
+ /* Let shd do the heal so that lookup is not blocked
+ * on getting metadata lock/doing the heal */
+ start = _gf_false;
+ break;
+ }
+
if (gf_uuid_compare (stbuf.ia_gfid, replies[i].poststat.ia_gfid)) {
start = _gf_false;
break;
diff --git a/xlators/cluster/afr/src/afr-self-heal-common.c b/xlators/cluster/afr/src/afr-self-heal-common.c
index d04f11d..c6ee75b 100644
--- a/xlators/cluster/afr/src/afr-self-heal-common.c
+++ b/xlators/cluster/afr/src/afr-self-heal-common.c
@@ -2182,44 +2182,6 @@ afr_selfheal_unentrylk (call_frame_t *frame, xlator_t *this, inode_t *inode,
return 0;
}
-
-gf_boolean_t
-afr_is_pending_set (xlator_t *this, dict_t *xdata, int type)
-{
- int idx = -1;
- afr_private_t *priv = NULL;
- void *pending_raw = NULL;
- int *pending_int = NULL;
- int i = 0;
-
- priv = this->private;
- idx = afr_index_for_transaction_type (type);
-
- if (dict_get_ptr (xdata, AFR_DIRTY, &pending_raw) == 0) {
- if (pending_raw) {
- pending_int = pending_raw;
-
- if (ntoh32 (pending_int[idx]))
- return _gf_true;
- }
- }
-
- for (i = 0; i < priv->child_count; i++) {
- if (dict_get_ptr (xdata, priv->pending_key[i],
- &pending_raw))
- continue;
- if (!pending_raw)
- continue;
- pending_int = pending_raw;
-
- if (ntoh32 (pending_int[idx]))
- return _gf_true;
- }
-
- return _gf_false;
-}
-
-
gf_boolean_t
afr_is_data_set (xlator_t *this, dict_t *xdata)
{
diff --git a/xlators/cluster/afr/src/afr.h b/xlators/cluster/afr/src/afr.h
index 2e6d995..af9dbc8 100644
--- a/xlators/cluster/afr/src/afr.h
+++ b/xlators/cluster/afr/src/afr.h
@@ -1225,4 +1225,7 @@ afr_set_inode_local (xlator_t *this, afr_local_t *local, inode_t *inode);
gf_boolean_t
afr_is_symmetric_error (call_frame_t *frame, xlator_t *this);
+
+gf_boolean_t
+afr_is_pending_set (xlator_t *this, dict_t *xdata, int type);
#endif /* __AFR_H__ */
--
1.8.3.1

352
0425-cluster-afr-Delegate-name-heal-when-possible.patch Normal file
View File

@ -0,0 +1,352 @@
From 8a3c0fb64c8798ecf5a3635fe0922e3cfd476817 Mon Sep 17 00:00:00 2001
From: Pranith Kumar K <pkarampu@redhat.com>
Date: Mon, 27 Aug 2018 12:40:16 +0530
Subject: [PATCH 425/444] cluster/afr: Delegate name-heal when possible
Problem:
When name-self-heal is triggered on the mount, it blocks
lookup until name-self-heal completes. But that can lead
to hangs when lot of clients are accessing a directory which
needs name heal and all of them trigger heals waiting
for other clients to complete heal.
Fix:
When a name-heal is needed but quorum number of names have the
file and pending xattrs exist on the parent, then better to
delegate the heal to SHD which will be completed as part of
entry-heal of the parent directory. We could also do the same
for quorum-number of names not present but we don't have
any known use-case where this is a frequent occurrence so
not changing that part at the moment. When there is a gfid
mismatch or missing gfid it is important to complete the heal
so that next rename doesn't assume everything is fine and
perform a rename etc
BUG: 1619357
Upstream Patch: https://review.gluster.org/c/glusterfs/+/21087
Change-Id: I8b002c85dffc6eb6f2833e742684a233daefeb2c
Signed-off-by: Pranith Kumar K <pkarampu@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/155029
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Sunil Kumar Heggodu Gopala Acharya <sheggodu@redhat.com>
---
tests/afr.rc | 8 ++
tests/basic/afr/name-self-heal.t | 112 +++++++++++++++++++++++++++
xlators/cluster/afr/src/afr-common.c | 100 ++++++++++++++++++------
xlators/cluster/afr/src/afr-self-heal-name.c | 12 ++-
4 files changed, 205 insertions(+), 27 deletions(-)
create mode 100644 tests/basic/afr/name-self-heal.t
diff --git a/tests/afr.rc b/tests/afr.rc
index 1fd0310..a1e8a44 100644
--- a/tests/afr.rc
+++ b/tests/afr.rc
@@ -89,3 +89,11 @@ function count_index_entries()
{
ls $1/.glusterfs/indices/xattrop | wc -l
}
+
+function get_quorum_type()
+{
+ local m="$1"
+ local v="$2"
+ local repl_id="$3"
+ cat $m/.meta/graphs/active/$v-replicate-$repl_id/private|grep quorum-type|awk '{print $3}'
+}
diff --git a/tests/basic/afr/name-self-heal.t b/tests/basic/afr/name-self-heal.t
new file mode 100644
index 0000000..50fc2ec
--- /dev/null
+++ b/tests/basic/afr/name-self-heal.t
@@ -0,0 +1,112 @@
+#!/bin/bash
+#Self-heal tests
+
+. $(dirname $0)/../../include.rc
+. $(dirname $0)/../../volume.rc
+. $(dirname $0)/../../afr.rc
+cleanup;
+
+#Check that when quorum is not enabled name-heal happens correctly
+TEST glusterd
+TEST pidof glusterd
+TEST $CLI volume create $V0 replica 2 $H0:$B0/brick{0,1}
+TEST $CLI volume set $V0 performance.stat-prefetch off
+TEST $CLI volume start $V0
+TEST $CLI volume heal $V0 disable
+TEST $CLI volume set $V0 cluster.entry-self-heal off
+TEST $GFS --volfile-id=$V0 --volfile-server=$H0 $M0;
+
+TEST touch $M0/a
+TEST touch $M0/c
+TEST kill_brick $V0 $H0 $B0/brick0
+TEST touch $M0/b
+TEST rm -f $M0/a
+TEST rm -f $M0/c
+TEST touch $M0/c #gfid mismatch case
+c_gfid=$(gf_get_gfid_xattr $B0/brick1/c)
+TEST $CLI volume start $V0 force
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status $V0 0
+TEST ! stat $M0/a
+TEST ! stat $B0/brick0/a
+TEST ! stat $B0/brick1/a
+
+TEST stat $M0/b
+TEST stat $B0/brick0/b
+TEST stat $B0/brick1/b
+TEST [[ "$(gf_get_gfid_xattr $B0/brick0/b)" == "$(gf_get_gfid_xattr $B0/brick1/b)" ]]
+
+TEST stat $M0/c
+TEST stat $B0/brick0/c
+TEST stat $B0/brick1/c
+TEST [[ "$(gf_get_gfid_xattr $B0/brick0/c)" == "$c_gfid" ]]
+
+cleanup;
+
+#Check that when quorum is enabled name-heal happens as expected
+TEST glusterd
+TEST pidof glusterd
+TEST $CLI volume create $V0 replica 3 $H0:$B0/brick{0,1,2}
+TEST $CLI volume set $V0 performance.stat-prefetch off
+TEST $CLI volume start $V0
+TEST $CLI volume heal $V0 disable
+TEST $CLI volume set $V0 cluster.entry-self-heal off
+TEST $GFS --volfile-id=$V0 --volfile-server=$H0 $M0;
+
+TEST touch $M0/a
+TEST touch $M0/c
+TEST kill_brick $V0 $H0 $B0/brick0
+TEST touch $M0/b
+TEST rm -f $M0/a
+TEST rm -f $M0/c
+TEST touch $M0/c #gfid mismatch case
+c_gfid=$(gf_get_gfid_xattr $B0/brick1/c)
+TEST $CLI volume start $V0 force
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status $V0 0
+TEST ! stat $M0/a
+TEST ! stat $B0/brick0/a
+TEST ! stat $B0/brick1/a
+TEST ! stat $B0/brick2/a
+
+TEST stat $M0/b
+TEST ! stat $B0/brick0/b #Name heal shouldn't be triggered
+TEST stat $B0/brick1/b
+TEST stat $B0/brick2/b
+
+TEST stat $M0/c
+TEST stat $B0/brick0/c
+TEST stat $B0/brick1/c
+TEST stat $B0/brick2/c
+TEST [[ "$(gf_get_gfid_xattr $B0/brick0/c)" == "$c_gfid" ]]
+
+TEST $CLI volume set $V0 cluster.quorum-type none
+EXPECT_WITHIN $CONFIG_UPDATE_TIMEOUT "none" get_quorum_type $M0 $V0 0
+TEST stat $M0/b
+TEST stat $B0/brick0/b #Name heal should be triggered
+TEST stat $B0/brick1/b
+TEST stat $B0/brick2/b
+TEST [[ "$(gf_get_gfid_xattr $B0/brick0/b)" == "$(gf_get_gfid_xattr $B0/brick1/b)" ]]
+TEST $CLI volume set $V0 cluster.quorum-type auto
+EXPECT_WITHIN $CONFIG_UPDATE_TIMEOUT "auto" get_quorum_type $M0 $V0 0
+
+EXPECT_WITHIN $UMOUNT_TIMEOUT "Y" force_umount $M0
+
+#Missing parent xattrs cases
+TEST $CLI volume heal $V0 enable
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "Y" glustershd_up_status
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status_in_shd $V0 0
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status_in_shd $V0 1
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status_in_shd $V0 2
+TEST $CLI volume heal $V0
+EXPECT_WITHIN $HEAL_TIMEOUT "0" get_pending_heal_count $V0
+
+TEST $GFS --volfile-id=$V0 --volfile-server=$H0 $M0;
+TEST $CLI volume heal $V0 disable
+#In cases where a good parent doesn't have pending xattrs and a file,
+#name-heal will be triggered
+TEST gf_rm_file_and_gfid_link $B0/brick1 c
+TEST stat $M0/c
+TEST stat $B0/brick0/c
+TEST stat $B0/brick1/c
+TEST stat $B0/brick2/c
+TEST [[ "$(gf_get_gfid_xattr $B0/brick0/c)" == "$c_gfid" ]]
+cleanup
diff --git a/xlators/cluster/afr/src/afr-common.c b/xlators/cluster/afr/src/afr-common.c
index e74fdec..ce2b17a 100644
--- a/xlators/cluster/afr/src/afr-common.c
+++ b/xlators/cluster/afr/src/afr-common.c
@@ -2302,8 +2302,6 @@ afr_lookup_done (call_frame_t *frame, xlator_t *this)
*/
for (i = 0; i < priv->child_count; i++) {
if (!replies[i].valid || replies[i].op_ret == -1) {
- if (priv->child_up[i])
- can_interpret = _gf_false;
continue;
}
@@ -2742,21 +2740,52 @@ afr_lookup_entry_heal (call_frame_t *frame, xlator_t *this)
afr_private_t *priv = NULL;
call_frame_t *heal = NULL;
int i = 0, first = -1;
- gf_boolean_t need_heal = _gf_false;
+ gf_boolean_t name_state_mismatch = _gf_false;
struct afr_reply *replies = NULL;
int ret = 0;
+ unsigned char *par_readables = NULL;
+ unsigned char *success = NULL;
+ int32_t op_errno = 0;
+ uuid_t gfid = {0};
local = frame->local;
replies = local->replies;
priv = this->private;
+ par_readables = alloca0(priv->child_count);
+ success = alloca0(priv->child_count);
+
+ ret = afr_inode_read_subvol_get (local->loc.parent, this, par_readables,
+ NULL, NULL);
+ if (ret < 0 || AFR_COUNT (par_readables, priv->child_count) == 0) {
+ /* In this case set par_readables to all 1 so that name_heal
+ * need checks at the end of this function will flag missing
+ * entry when name state mismatches*/
+ memset (par_readables, 1, priv->child_count);
+ }
for (i = 0; i < priv->child_count; i++) {
if (!replies[i].valid)
continue;
+ if (replies[i].op_ret == 0) {
+ if (uuid_is_null (gfid)) {
+ gf_uuid_copy (gfid,
+ replies[i].poststat.ia_gfid);
+ }
+ success[i] = 1;
+ } else {
+ if ((replies[i].op_errno != ENOTCONN) &&
+ (replies[i].op_errno != ENOENT) &&
+ (replies[i].op_errno != ESTALE)) {
+ op_errno = replies[i].op_errno;
+ }
+ }
+
+ /*gfid is missing, needs heal*/
if ((replies[i].op_ret == -1) &&
- (replies[i].op_errno == ENODATA))
- need_heal = _gf_true;
+ (replies[i].op_errno == ENODATA)) {
+ goto name_heal;
+ }
if (first == -1) {
first = i;
@@ -2764,30 +2793,53 @@ afr_lookup_entry_heal (call_frame_t *frame, xlator_t *this)
}
if (replies[i].op_ret != replies[first].op_ret) {
- need_heal = _gf_true;
- break;
+ name_state_mismatch = _gf_true;
}
- if (gf_uuid_compare (replies[i].poststat.ia_gfid,
- replies[first].poststat.ia_gfid)) {
- need_heal = _gf_true;
- break;
- }
+ if (replies[i].op_ret == 0) {
+ /* Rename after this lookup may succeed if we don't do
+ * a name-heal and the destination may not have pending xattrs
+ * to indicate which name is good and which is bad so always do
+ * this heal*/
+ if (gf_uuid_compare (replies[i].poststat.ia_gfid,
+ gfid)) {
+ goto name_heal;
+ }
+ }
}
- if (need_heal) {
- heal = afr_frame_create (this, NULL);
- if (!heal)
- goto metadata_heal;
-
- ret = synctask_new (this->ctx->env, afr_lookup_selfheal_wrap,
- afr_refresh_selfheal_done, heal, frame);
- if (ret) {
- AFR_STACK_DESTROY (heal);
- goto metadata_heal;
+ if (name_state_mismatch) {
+ if (!priv->quorum_count)
+ goto name_heal;
+ if (!afr_has_quorum (success, this))
+ goto name_heal;
+ if (op_errno)
+ goto name_heal;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid)
+ continue;
+ if (par_readables[i] && replies[i].op_ret < 0 &&
+ replies[i].op_errno != ENOTCONN) {
+ goto name_heal;
+ }
}
- return ret;
- }
+ }
+
+ goto metadata_heal;
+
+name_heal:
+ heal = afr_frame_create (this, NULL);
+ if (!heal)
+ goto metadata_heal;
+
+ ret = synctask_new (this->ctx->env, afr_lookup_selfheal_wrap,
+ afr_refresh_selfheal_done, heal, frame);
+ if (ret) {
+ AFR_STACK_DESTROY (heal);
+ goto metadata_heal;
+ }
+ return ret;
+
metadata_heal:
ret = afr_lookup_metadata_heal_check (frame, this);
diff --git a/xlators/cluster/afr/src/afr-self-heal-name.c b/xlators/cluster/afr/src/afr-self-heal-name.c
index bcd0e60..0a5be29 100644
--- a/xlators/cluster/afr/src/afr-self-heal-name.c
+++ b/xlators/cluster/afr/src/afr-self-heal-name.c
@@ -634,20 +634,26 @@ afr_selfheal_name_unlocked_inspect (call_frame_t *frame, xlator_t *this,
continue;
if ((replies[i].op_ret == -1) &&
- (replies[i].op_errno == ENODATA))
+ (replies[i].op_errno == ENODATA)) {
*need_heal = _gf_true;
+ break;
+ }
if (first_idx == -1) {
first_idx = i;
continue;
}
- if (replies[i].op_ret != replies[first_idx].op_ret)
+ if (replies[i].op_ret != replies[first_idx].op_ret) {
*need_heal = _gf_true;
+ break;
+ }
if (gf_uuid_compare (replies[i].poststat.ia_gfid,
- replies[first_idx].poststat.ia_gfid))
+ replies[first_idx].poststat.ia_gfid)) {
*need_heal = _gf_true;
+ break;
+ }
}
if (inode)
--
1.8.3.1

582
0426-features-shard-Make-operations-on-internal-directori.patch Normal file
View File

@ -0,0 +1,582 @@
From 9be984ac2b71423b72ab3b1fa45b4d77a263ce1e Mon Sep 17 00:00:00 2001
From: Krutika Dhananjay <kdhananj@redhat.com>
Date: Wed, 28 Mar 2018 12:09:27 +0530
Subject: [PATCH 426/444] features/shard: Make operations on internal
directories generic
> Upstream: https://review.gluster.org/19892
> BUG: 1568521
> Change-Id: Iea7ad2102220c6d415909f8caef84167ce2d6818
Change-Id: Iea7ad2102220c6d415909f8caef84167ce2d6818
BUG: 1520882
Signed-off-by: Krutika Dhananjay <kdhananj@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/154860
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Xavi Hernandez <xhernandez@redhat.com>
---
xlators/features/shard/src/shard.c | 295 +++++++++++++++++++++++++------------
xlators/features/shard/src/shard.h | 4 +
2 files changed, 206 insertions(+), 93 deletions(-)
diff --git a/xlators/features/shard/src/shard.c b/xlators/features/shard/src/shard.c
index f5fb181..5ff04df 100644
--- a/xlators/features/shard/src/shard.c
+++ b/xlators/features/shard/src/shard.c
@@ -546,30 +546,55 @@ shard_call_count_return (call_frame_t *frame)
return call_count;
}
+static char *
+shard_internal_dir_string (shard_internal_dir_type_t type)
+{
+ char *str = NULL;
+
+ switch (type) {
+ case SHARD_INTERNAL_DIR_DOT_SHARD:
+ str = ".shard";
+ break;
+ default:
+ break;
+ }
+ return str;
+}
+
static int
-shard_init_dot_shard_loc (xlator_t *this, shard_local_t *local)
+shard_init_internal_dir_loc (xlator_t *this, shard_local_t *local,
+ shard_internal_dir_type_t type)
{
- int ret = -1;
- loc_t *dot_shard_loc = NULL;
+ int ret = -1;
+ char *bname = NULL;
+ loc_t *internal_dir_loc = NULL;
if (!local)
return -1;
- dot_shard_loc = &local->dot_shard_loc;
- dot_shard_loc->inode = inode_new (this->itable);
- dot_shard_loc->parent = inode_ref (this->itable->root);
- ret = inode_path (dot_shard_loc->parent, GF_SHARD_DIR,
- (char **)&dot_shard_loc->path);
- if (ret < 0 || !(dot_shard_loc->inode)) {
+ switch (type) {
+ case SHARD_INTERNAL_DIR_DOT_SHARD:
+ internal_dir_loc = &local->dot_shard_loc;
+ bname = GF_SHARD_DIR;
+ break;
+ default:
+ break;
+ }
+
+ internal_dir_loc->inode = inode_new (this->itable);
+ internal_dir_loc->parent = inode_ref (this->itable->root);
+ ret = inode_path (internal_dir_loc->parent, bname,
+ (char **)&internal_dir_loc->path);
+ if (ret < 0 || !(internal_dir_loc->inode)) {
gf_msg (this->name, GF_LOG_ERROR, 0,
SHARD_MSG_INODE_PATH_FAILED,
- "Inode path failed on %s", GF_SHARD_DIR);
+ "Inode path failed on %s", bname);
goto out;
}
- dot_shard_loc->name = strrchr (dot_shard_loc->path, '/');
- if (dot_shard_loc->name)
- dot_shard_loc->name++;
+ internal_dir_loc->name = strrchr (internal_dir_loc->path, '/');
+ if (internal_dir_loc->name)
+ internal_dir_loc->name++;
ret = 0;
out:
@@ -1029,28 +1054,42 @@ out:
}
static inode_t *
-shard_link_dot_shard_inode (shard_local_t *local, inode_t *inode,
- struct iatt *buf)
+shard_link_internal_dir_inode (shard_local_t *local, inode_t *inode,
+ struct iatt *buf, shard_internal_dir_type_t type)
{
inode_t *linked_inode = NULL;
shard_priv_t *priv = NULL;
+ char *bname = NULL;
+ inode_t **priv_inode = NULL;
priv = THIS->private;
- linked_inode = inode_link (inode, inode->table->root, ".shard", buf);
+ switch (type) {
+ case SHARD_INTERNAL_DIR_DOT_SHARD:
+ bname = ".shard";
+ priv_inode = &priv->dot_shard_inode;
+ break;
+ default:
+ break;
+ }
+
+ linked_inode = inode_link (inode, inode->table->root, bname, buf);
inode_lookup (linked_inode);
- priv->dot_shard_inode = linked_inode;
+ *priv_inode = linked_inode;
return linked_inode;
}
int
-shard_refresh_dot_shard_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata,
- struct iatt *postparent)
+shard_refresh_internal_dir_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret,
+ int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xdata,
+ struct iatt *postparent)
{
- shard_local_t *local = NULL;
+ shard_local_t *local = NULL;
+ inode_t *linked_inode = NULL;
+ shard_internal_dir_type_t type = (shard_internal_dir_type_t) cookie;
local = frame->local;
@@ -1061,27 +1100,37 @@ shard_refresh_dot_shard_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
}
/* To-Do: Fix refcount increment per call to
- * shard_link_dot_shard_inode().
+ * shard_link_internal_dir_inode().
*/
- shard_link_dot_shard_inode (local, inode, buf);
- shard_inode_ctx_set_refreshed_flag (inode, this);
+ linked_inode = shard_link_internal_dir_inode (local, inode, buf, type);
+ shard_inode_ctx_set_refreshed_flag (linked_inode, this);
out:
shard_common_resolve_shards (frame, this, local->post_res_handler);
return 0;
}
int
-shard_refresh_dot_shard (call_frame_t *frame, xlator_t *this)
+shard_refresh_internal_dir (call_frame_t *frame, xlator_t *this,
+ shard_internal_dir_type_t type)
{
loc_t loc = {0,};
inode_t *inode = NULL;
shard_priv_t *priv = NULL;
shard_local_t *local = NULL;
+ uuid_t gfid = {0,};
local = frame->local;
priv = this->private;
- inode = inode_find (this->itable, priv->dot_shard_gfid);
+ switch (type) {
+ case SHARD_INTERNAL_DIR_DOT_SHARD:
+ gf_uuid_copy (gfid, priv->dot_shard_gfid);
+ break;
+ default:
+ break;
+ }
+
+ inode = inode_find (this->itable, gfid);
if (!shard_inode_ctx_needs_lookup (inode, this)) {
local->op_ret = 0;
@@ -1092,10 +1141,11 @@ shard_refresh_dot_shard (call_frame_t *frame, xlator_t *this)
* call to inode_find()
*/
loc.inode = inode;
- gf_uuid_copy (loc.gfid, priv->dot_shard_gfid);
+ gf_uuid_copy (loc.gfid, gfid);
- STACK_WIND (frame, shard_refresh_dot_shard_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lookup, &loc, NULL);
+ STACK_WIND_COOKIE (frame, shard_refresh_internal_dir_cbk,
+ (void *)(long) type, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lookup, &loc, NULL);
loc_wipe (&loc);
return 0;
@@ -1106,13 +1156,14 @@ out:
}
int
-shard_lookup_dot_shard_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata,
- struct iatt *postparent)
+shard_lookup_internal_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xdata,
+ struct iatt *postparent)
{
- inode_t *link_inode = NULL;
- shard_local_t *local = NULL;
+ inode_t *link_inode = NULL;
+ shard_local_t *local = NULL;
+ shard_internal_dir_type_t type = (shard_internal_dir_type_t) cookie;
local = frame->local;
@@ -1124,17 +1175,17 @@ shard_lookup_dot_shard_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
if (!IA_ISDIR (buf->ia_type)) {
gf_msg (this->name, GF_LOG_CRITICAL, 0,
- SHARD_MSG_DOT_SHARD_NODIR, "/.shard already exists and "
- "is not a directory. Please remove /.shard from all "
- "bricks and try again");
+ SHARD_MSG_DOT_SHARD_NODIR, "%s already exists and "
+ "is not a directory. Please remove it from all bricks "
+ "and try again", shard_internal_dir_string (type));
local->op_ret = -1;
local->op_errno = EIO;
goto unwind;
}
- link_inode = shard_link_dot_shard_inode (local, inode, buf);
+ link_inode = shard_link_internal_dir_inode (local, inode, buf, type);
if (link_inode != inode) {
- shard_refresh_dot_shard (frame, this);
+ shard_refresh_internal_dir (frame, this, type);
} else {
shard_inode_ctx_set_refreshed_flag (link_inode, this);
shard_common_resolve_shards (frame, this,
@@ -1148,18 +1199,26 @@ unwind:
}
int
-shard_lookup_dot_shard (call_frame_t *frame, xlator_t *this,
- shard_post_resolve_fop_handler_t post_res_handler)
+shard_lookup_internal_dir (call_frame_t *frame, xlator_t *this,
+ shard_post_resolve_fop_handler_t post_res_handler,
+ shard_internal_dir_type_t type)
{
int ret = -1;
dict_t *xattr_req = NULL;
shard_priv_t *priv = NULL;
shard_local_t *local = NULL;
+ uuid_t *gfid = NULL;
+ loc_t *loc = NULL;
+ gf_boolean_t free_gfid = _gf_true;
local = frame->local;
priv = this->private;
local->post_res_handler = post_res_handler;
+ gfid = GF_CALLOC (1, sizeof(uuid_t), gf_common_mt_uuid_t);
+ if (!gfid)
+ goto err;
+
xattr_req = dict_new ();
if (!xattr_req) {
local->op_ret = -1;
@@ -1167,26 +1226,38 @@ shard_lookup_dot_shard (call_frame_t *frame, xlator_t *this,
goto err;
}
- ret = dict_set_static_bin (xattr_req, "gfid-req", priv->dot_shard_gfid,
- 16);
+ switch (type) {
+ case SHARD_INTERNAL_DIR_DOT_SHARD:
+ gf_uuid_copy (*gfid, priv->dot_shard_gfid);
+ loc = &local->dot_shard_loc;
+ break;
+ default:
+ break;
+ }
+
+ ret = dict_set_bin (xattr_req, "gfid-req", *gfid, 16);
if (ret) {
gf_msg (this->name, GF_LOG_ERROR, 0, SHARD_MSG_DICT_SET_FAILED,
- "Failed to set gfid of /.shard into dict");
+ "Failed to set gfid of %s into dict",
+ shard_internal_dir_string (type));
local->op_ret = -1;
local->op_errno = ENOMEM;
goto err;
+ } else {
+ free_gfid = _gf_false;
}
- STACK_WIND (frame, shard_lookup_dot_shard_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lookup, &local->dot_shard_loc,
- xattr_req);
-
+ STACK_WIND_COOKIE (frame, shard_lookup_internal_dir_cbk,
+ (void *) (long) type, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lookup, loc, xattr_req);
dict_unref (xattr_req);
return 0;
err:
if (xattr_req)
dict_unref (xattr_req);
+ if (free_gfid)
+ GF_FREE (gfid);
post_res_handler (frame, this);
return 0;
}
@@ -2203,14 +2274,17 @@ shard_truncate_begin (call_frame_t *frame, xlator_t *this)
local->dot_shard_loc.inode = inode_find (this->itable,
priv->dot_shard_gfid);
if (!local->dot_shard_loc.inode) {
- ret = shard_init_dot_shard_loc (this, local);
+ ret = shard_init_internal_dir_loc (this, local,
+ SHARD_INTERNAL_DIR_DOT_SHARD);
if (ret)
goto err;
- shard_lookup_dot_shard (frame, this,
- shard_post_resolve_truncate_handler);
+ shard_lookup_internal_dir (frame, this,
+ shard_post_resolve_truncate_handler,
+ SHARD_INTERNAL_DIR_DOT_SHARD);
} else {
local->post_res_handler = shard_post_resolve_truncate_handler;
- shard_refresh_dot_shard (frame, this);
+ shard_refresh_internal_dir (frame, this,
+ SHARD_INTERNAL_DIR_DOT_SHARD);
}
return 0;
@@ -2682,14 +2756,17 @@ shard_unlink_base_file_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local->dot_shard_loc.inode = inode_find (this->itable,
priv->dot_shard_gfid);
if (!local->dot_shard_loc.inode) {
- ret = shard_init_dot_shard_loc (this, local);
+ ret = shard_init_internal_dir_loc (this, local,
+ SHARD_INTERNAL_DIR_DOT_SHARD);
if (ret)
goto unwind;
- shard_lookup_dot_shard (frame, this,
- shard_post_resolve_unlink_handler);
+ shard_lookup_internal_dir (frame, this,
+ shard_post_resolve_unlink_handler,
+ SHARD_INTERNAL_DIR_DOT_SHARD);
} else {
local->post_res_handler = shard_post_resolve_unlink_handler;
- shard_refresh_dot_shard (frame, this);
+ shard_refresh_internal_dir (frame, this,
+ SHARD_INTERNAL_DIR_DOT_SHARD);
}
return 0;
@@ -3048,14 +3125,17 @@ shard_rename_unlink_dst_shards_do (call_frame_t *frame, xlator_t *this)
local->dot_shard_loc.inode = inode_find (this->itable,
priv->dot_shard_gfid);
if (!local->dot_shard_loc.inode) {
- ret = shard_init_dot_shard_loc (this, local);
+ ret = shard_init_internal_dir_loc (this, local,
+ SHARD_INTERNAL_DIR_DOT_SHARD);
if (ret)
goto out;
- shard_lookup_dot_shard (frame, this,
- shard_post_resolve_unlink_handler);
+ shard_lookup_internal_dir (frame, this,
+ shard_post_resolve_unlink_handler,
+ SHARD_INTERNAL_DIR_DOT_SHARD);
} else {
local->post_res_handler = shard_post_resolve_unlink_handler;
- shard_refresh_dot_shard (frame, this);
+ shard_refresh_internal_dir (frame, this,
+ SHARD_INTERNAL_DIR_DOT_SHARD);
}
return 0;
@@ -3811,14 +3891,17 @@ shard_post_lookup_readv_handler (call_frame_t *frame, xlator_t *this)
local->dot_shard_loc.inode = inode_find (this->itable,
priv->dot_shard_gfid);
if (!local->dot_shard_loc.inode) {
- ret = shard_init_dot_shard_loc (this, local);
+ ret = shard_init_internal_dir_loc (this, local,
+ SHARD_INTERNAL_DIR_DOT_SHARD);
if (ret)
goto err;
- shard_lookup_dot_shard (frame, this,
- shard_post_resolve_readv_handler);
+ shard_lookup_internal_dir (frame, this,
+ shard_post_resolve_readv_handler,
+ SHARD_INTERNAL_DIR_DOT_SHARD);
} else {
local->post_res_handler = shard_post_resolve_readv_handler;
- shard_refresh_dot_shard (frame, this);
+ shard_refresh_internal_dir (frame, this,
+ SHARD_INTERNAL_DIR_DOT_SHARD);
}
return 0;
@@ -4249,8 +4332,9 @@ shard_common_inode_write_post_mknod_handler (call_frame_t *frame,
}
int
-shard_mkdir_dot_shard (call_frame_t *frame, xlator_t *this,
- shard_post_resolve_fop_handler_t handler);
+shard_mkdir_internal_dir (call_frame_t *frame, xlator_t *this,
+ shard_post_resolve_fop_handler_t handler,
+ shard_internal_dir_type_t type);
int
shard_common_inode_write_post_resolve_handler (call_frame_t *frame,
xlator_t *this)
@@ -4323,26 +4407,28 @@ shard_common_inode_write_post_lookup_handler (call_frame_t *frame,
if (!local->dot_shard_loc.inode) {
/*change handler*/
- shard_mkdir_dot_shard (frame, this,
- shard_common_inode_write_post_resolve_handler);
+ shard_mkdir_internal_dir (frame, this,
+ shard_common_inode_write_post_resolve_handler,
+ SHARD_INTERNAL_DIR_DOT_SHARD);
} else {
/*change handler*/
local->post_res_handler =
shard_common_inode_write_post_resolve_handler;
- shard_refresh_dot_shard (frame, this);
+ shard_refresh_internal_dir (frame, this,
+ SHARD_INTERNAL_DIR_DOT_SHARD);
}
return 0;
}
int
-shard_mkdir_dot_shard_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret,
- int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+shard_mkdir_internal_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- inode_t *link_inode = NULL;
- shard_local_t *local = NULL;
+ inode_t *link_inode = NULL;
+ shard_local_t *local = NULL;
+ shard_internal_dir_type_t type = (shard_internal_dir_type_t) cookie;
local = frame->local;
@@ -4354,17 +4440,19 @@ shard_mkdir_dot_shard_cbk (call_frame_t *frame, void *cookie,
local->op_errno = op_errno;
goto unwind;
} else {
- gf_msg_debug (this->name, 0, "mkdir on /.shard failed "
- "with EEXIST. Attempting lookup now");
- shard_lookup_dot_shard (frame, this,
- local->post_res_handler);
+ gf_msg_debug (this->name, 0, "mkdir on %s failed "
+ "with EEXIST. Attempting lookup now",
+ shard_internal_dir_string (type));
+ shard_lookup_internal_dir (frame, this,
+ local->post_res_handler,
+ type);
return 0;
}
}
- link_inode = shard_link_dot_shard_inode (local, inode, buf);
+ link_inode = shard_link_internal_dir_inode (local, inode, buf, type);
if (link_inode != inode) {
- shard_refresh_dot_shard (frame, this);
+ shard_refresh_internal_dir (frame, this, type);
} else {
shard_inode_ctx_set_refreshed_flag (link_inode, this);
shard_common_resolve_shards (frame, this,
@@ -4377,40 +4465,59 @@ unwind:
}
int
-shard_mkdir_dot_shard (call_frame_t *frame, xlator_t *this,
- shard_post_resolve_fop_handler_t handler)
+shard_mkdir_internal_dir (call_frame_t *frame, xlator_t *this,
+ shard_post_resolve_fop_handler_t handler,
+ shard_internal_dir_type_t type)
{
int ret = -1;
shard_local_t *local = NULL;
shard_priv_t *priv = NULL;
dict_t *xattr_req = NULL;
+ uuid_t *gfid = NULL;
+ loc_t *loc = NULL;
+ gf_boolean_t free_gfid = _gf_true;
local = frame->local;
priv = this->private;
local->post_res_handler = handler;
+ gfid = GF_CALLOC (1, sizeof(uuid_t), gf_common_mt_uuid_t);
+ if (!gfid)
+ goto err;
+
+ switch (type) {
+ case SHARD_INTERNAL_DIR_DOT_SHARD:
+ gf_uuid_copy (*gfid, priv->dot_shard_gfid);
+ loc = &local->dot_shard_loc;
+ break;
+ default:
+ break;
+ }
xattr_req = dict_new ();
if (!xattr_req)
goto err;
- ret = shard_init_dot_shard_loc (this, local);
+ ret = shard_init_internal_dir_loc (this, local, type);
if (ret)
goto err;
- ret = dict_set_static_bin (xattr_req, "gfid-req", priv->dot_shard_gfid,
- 16);
+ ret = dict_set_bin (xattr_req, "gfid-req", *gfid, 16);
if (ret) {
gf_msg (this->name, GF_LOG_ERROR, 0, SHARD_MSG_DICT_SET_FAILED,
- "Failed to set gfid-req for /.shard");
+ "Failed to set gfid-req for %s",
+ shard_internal_dir_string (type));
goto err;
+ } else {
+ free_gfid = _gf_false;
}
SHARD_SET_ROOT_FS_ID (frame, local);
- STACK_WIND (frame, shard_mkdir_dot_shard_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->mkdir,
- &local->dot_shard_loc, 0755, 0, xattr_req);
+ STACK_WIND_COOKIE (frame, shard_mkdir_internal_dir_cbk,
+ (void *)(long) type, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mkdir, loc, 0755, 0,
+ xattr_req);
dict_unref (xattr_req);
return 0;
@@ -4419,6 +4526,8 @@ err:
dict_unref (xattr_req);
local->op_ret = -1;
local->op_errno = ENOMEM;
+ if (free_gfid)
+ GF_FREE (gfid);
handler (frame, this);
return 0;
}
diff --git a/xlators/features/shard/src/shard.h b/xlators/features/shard/src/shard.h
index 75d39a1..a1adb6a 100644
--- a/xlators/features/shard/src/shard.h
+++ b/xlators/features/shard/src/shard.h
@@ -278,4 +278,8 @@ typedef struct shard_inode_ctx {
inode_t *base_inode;
} shard_inode_ctx_t;
+typedef enum {
+ SHARD_INTERNAL_DIR_DOT_SHARD = 1,
+} shard_internal_dir_type_t;
+
#endif /* __SHARD_H__ */
--
1.8.3.1

291
0427-features-shard-Add-option-to-barrier-parallel-lookup.patch Normal file
View File

@ -0,0 +1,291 @@
From 64b238d3a5caf7bdb32bca25946f84e0afe9bc7a Mon Sep 17 00:00:00 2001
From: Krutika Dhananjay <kdhananj@redhat.com>
Date: Tue, 17 Apr 2018 22:14:20 +0530
Subject: [PATCH 427/444] features/shard: Add option to barrier parallel lookup
and unlink of shards
> Upstream: https://review.gluster.org/19915
> BUG: 1568521
> Change-Id: Ib0f90a5f62abdfa89cda7bef9f3ff99f349ec332
Also move the common parallel unlink callback for GF_FOP_TRUNCATE and
GF_FOP_FTRUNCATE into a separate function.
Change-Id: Ib0f90a5f62abdfa89cda7bef9f3ff99f349ec332
BUG: 1520882
Signed-off-by: Krutika Dhananjay <kdhananj@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/154861
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Xavi Hernandez <xhernandez@redhat.com>
Tested-by: Sunil Kumar Heggodu Gopala Acharya <sheggodu@redhat.com>
---
xlators/features/shard/src/shard.c | 113 ++++++++++++++++++++++++++++---------
xlators/features/shard/src/shard.h | 4 ++
2 files changed, 89 insertions(+), 28 deletions(-)
diff --git a/xlators/features/shard/src/shard.c b/xlators/features/shard/src/shard.c
index 5ff04df..268ba20 100644
--- a/xlators/features/shard/src/shard.c
+++ b/xlators/features/shard/src/shard.c
@@ -475,6 +475,7 @@ shard_local_wipe (shard_local_t *local)
count = local->num_blocks;
+ syncbarrier_destroy (&local->barrier);
loc_wipe (&local->loc);
loc_wipe (&local->dot_shard_loc);
loc_wipe (&local->loc2);
@@ -861,6 +862,7 @@ shard_common_resolve_shards (call_frame_t *frame, xlator_t *this,
priv = this->private;
local = frame->local;
+ local->call_count = 0;
shard_idx_iter = local->first_block;
res_inode = local->resolver_base_inode;
@@ -1780,6 +1782,37 @@ shard_unlink_shards_do_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
struct iatt *preparent, struct iatt *postparent,
dict_t *xdata);
+void
+shard_unlink_block_inode (shard_local_t *local, int shard_block_num);
+
+int
+shard_truncate_htol_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
+{
+ int call_count = 0;
+ int shard_block_num = (long) cookie;
+ shard_local_t *local = NULL;
+
+ local = frame->local;
+
+ if (op_ret < 0) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ goto done;
+ }
+
+ shard_unlink_block_inode (local, shard_block_num);
+done:
+ call_count = shard_call_count_return (frame);
+ if (call_count == 0) {
+ SHARD_UNSET_ROOT_FS_ID (frame, local);
+ shard_truncate_last_shard (frame, this, local->inode_list[0]);
+ }
+ return 0;
+}
+
int
shard_truncate_htol (call_frame_t *frame, xlator_t *this, inode_t *inode)
{
@@ -1839,10 +1872,9 @@ shard_truncate_htol (call_frame_t *frame, xlator_t *this, inode_t *inode)
continue;
}
if (wind_failed) {
- shard_unlink_shards_do_cbk (frame,
- (void *)(long) cur_block,
- this, -1, ENOMEM, NULL,
- NULL, NULL);
+ shard_truncate_htol_cbk (frame, (void *)(long) cur_block,
+ this, -1, ENOMEM, NULL, NULL,
+ NULL);
goto next;
}
@@ -1860,10 +1892,9 @@ shard_truncate_htol (call_frame_t *frame, xlator_t *this, inode_t *inode)
local->op_errno = ENOMEM;
loc_wipe (&loc);
wind_failed = _gf_true;
- shard_unlink_shards_do_cbk (frame,
- (void *)(long) cur_block,
- this, -1, ENOMEM, NULL,
- NULL, NULL);
+ shard_truncate_htol_cbk (frame, (void *)(long) cur_block,
+ this, -1, ENOMEM, NULL, NULL,
+ NULL);
goto next;
}
loc.name = strrchr (loc.path, '/');
@@ -1871,7 +1902,7 @@ shard_truncate_htol (call_frame_t *frame, xlator_t *this, inode_t *inode)
loc.name++;
loc.inode = inode_ref (local->inode_list[i]);
- STACK_WIND_COOKIE (frame, shard_unlink_shards_do_cbk,
+ STACK_WIND_COOKIE (frame, shard_truncate_htol_cbk,
(void *) (long) cur_block, FIRST_CHILD(this),
FIRST_CHILD (this)->fops->unlink, &loc,
0, NULL);
@@ -2022,13 +2053,18 @@ shard_common_lookup_shards_cbk (call_frame_t *frame, void *cookie,
done:
call_count = shard_call_count_return (frame);
- if (call_count == 0) {
- if (!local->first_lookup_done)
- local->first_lookup_done = _gf_true;
- if (local->op_ret < 0)
- goto unwind;
- else
- local->pls_fop_handler (frame, this);
+ if (local->lookup_shards_barriered) {
+ syncbarrier_wake (&local->barrier);
+ return 0;
+ } else {
+ if (call_count == 0) {
+ if (!local->first_lookup_done)
+ local->first_lookup_done = _gf_true;
+ if (local->op_ret < 0)
+ goto unwind;
+ else
+ local->pls_fop_handler (frame, this);
+ }
}
return 0;
@@ -2074,6 +2110,7 @@ shard_common_lookup_shards (call_frame_t *frame, xlator_t *this, inode_t *inode,
{
int i = 0;
int ret = 0;
+ int count = 0;
int call_count = 0;
int32_t shard_idx_iter = 0;
int last_block = 0;
@@ -2087,10 +2124,12 @@ shard_common_lookup_shards (call_frame_t *frame, xlator_t *this, inode_t *inode,
priv = this->private;
local = frame->local;
- call_count = local->call_count;
+ count = call_count = local->call_count;
shard_idx_iter = local->first_block;
last_block = local->last_block;
local->pls_fop_handler = handler;
+ if (local->lookup_shards_barriered)
+ local->barrier.waitfor = local->call_count;
while (shard_idx_iter <= last_block) {
if (local->inode_list[i]) {
@@ -2162,7 +2201,8 @@ next:
if (!--call_count)
break;
}
-
+ if (local->lookup_shards_barriered)
+ syncbarrier_wait (&local->barrier, count);
return 0;
}
@@ -2400,6 +2440,9 @@ shard_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
frame->local = local;
+ ret = syncbarrier_init (&local->barrier);
+ if (ret)
+ goto err;
loc_copy (&local->loc, loc);
local->offset = offset;
local->block_size = block_size;
@@ -2450,6 +2493,9 @@ shard_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
goto err;
frame->local = local;
+ ret = syncbarrier_init (&local->barrier);
+ if (ret)
+ goto err;
local->fd = fd_ref (fd);
local->offset = offset;
local->block_size = block_size;
@@ -2881,18 +2927,19 @@ shard_unlink_shards_do_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
done:
call_count = shard_call_count_return (frame);
- if (call_count == 0) {
- SHARD_UNSET_ROOT_FS_ID (frame, local);
+ if (local->unlink_shards_barriered) {
+ syncbarrier_wake (&local->barrier);
+ } else {
- if (local->fop == GF_FOP_UNLINK)
- shard_unlink_cbk (frame, this);
- else if (local->fop == GF_FOP_RENAME)
- shard_rename_cbk (frame, this);
- else
- shard_truncate_last_shard (frame, this,
- local->inode_list[0]);
- }
+ if (call_count == 0) {
+ SHARD_UNSET_ROOT_FS_ID (frame, local);
+ if (local->fop == GF_FOP_UNLINK)
+ shard_unlink_cbk (frame, this);
+ else if (local->fop == GF_FOP_RENAME)
+ shard_rename_cbk (frame, this);
+ }
+ }
return 0;
}
@@ -2952,6 +2999,8 @@ shard_unlink_shards_do (call_frame_t *frame, xlator_t *this, inode_t *inode)
local->call_count = call_count = count;
cur_block = 1;
SHARD_SET_ROOT_FS_ID (frame, local);
+ if (local->unlink_shards_barriered)
+ local->barrier.waitfor = count;
/* Ignore the base file and start iterating from the first block shard.
*/
@@ -3006,6 +3055,8 @@ next:
if (!--call_count)
break;
}
+ if (local->unlink_shards_barriered)
+ syncbarrier_wait (&local->barrier, count);
return 0;
}
@@ -3947,6 +3998,9 @@ shard_readv (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
frame->local = local;
+ ret = syncbarrier_init (&local->barrier);
+ if (ret)
+ goto err;
local->fd = fd_ref (fd);
local->block_size = block_size;
local->offset = offset;
@@ -5414,6 +5468,9 @@ shard_common_inode_write_begin (call_frame_t *frame, xlator_t *this,
frame->local = local;
+ ret = syncbarrier_init (&local->barrier);
+ if (ret)
+ goto out;
local->xattr_req = (xdata) ? dict_ref (xdata) : dict_new ();
if (!local->xattr_req)
goto out;
diff --git a/xlators/features/shard/src/shard.h b/xlators/features/shard/src/shard.h
index a1adb6a..225caa0 100644
--- a/xlators/features/shard/src/shard.h
+++ b/xlators/features/shard/src/shard.h
@@ -15,6 +15,7 @@
#include "xlator.h"
#include "compat-errno.h"
#include "shard-messages.h"
+#include "syncop.h"
#define GF_SHARD_DIR ".shard"
#define SHARD_MIN_BLOCK_SIZE (4 * GF_UNIT_MB)
@@ -257,6 +258,9 @@ typedef struct shard_local {
} lock;
inode_t *resolver_base_inode;
gf_boolean_t first_lookup_done;
+ syncbarrier_t barrier;
+ gf_boolean_t lookup_shards_barriered;
+ gf_boolean_t unlink_shards_barriered;
} shard_local_t;
typedef struct shard_inode_ctx {
--
1.8.3.1

99
0428-libglusterfs-syncop-Handle-barrier_-init-destroy-in-.patch Normal file
View File

@ -0,0 +1,99 @@
From c285acf172d42271d87eb069045ea70bce97b0b1 Mon Sep 17 00:00:00 2001
From: Pranith Kumar K <pkarampu@redhat.com>
Date: Mon, 23 Apr 2018 21:04:58 +0530
Subject: [PATCH 428/444] libglusterfs/syncop: Handle barrier_{init/destroy} in
error cases
> Upstream: https://review.gluster.org/19927
> BUG: 1568521
> Change-Id: I53e60cfcaa7f8edfa5eca47307fa99f10ee64505
Change-Id: I53e60cfcaa7f8edfa5eca47307fa99f10ee64505
BUG: 1520882
Signed-off-by: Pranith Kumar K <pkarampu@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/154862
Tested-by: Krutika Dhananjay <kdhananj@redhat.com>
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Xavi Hernandez <xhernandez@redhat.com>
Tested-by: Sunil Kumar Heggodu Gopala Acharya <sheggodu@redhat.com>
---
libglusterfs/src/syncop.c | 30 ++++++++++++++++++++++++++----
libglusterfs/src/syncop.h | 1 +
2 files changed, 27 insertions(+), 4 deletions(-)
diff --git a/libglusterfs/src/syncop.c b/libglusterfs/src/syncop.c
index ac40a1d..81d73b2 100644
--- a/libglusterfs/src/syncop.c
+++ b/libglusterfs/src/syncop.c
@@ -1087,30 +1087,52 @@ synclock_unlock (synclock_t *lock)
int
syncbarrier_init (struct syncbarrier *barrier)
{
+ int ret = 0;
if (!barrier) {
errno = EINVAL;
return -1;
}
- pthread_cond_init (&barrier->cond, 0);
+ ret = pthread_cond_init (&barrier->cond, 0);
+ if (ret) {
+ errno = ret;
+ return -1;
+ }
barrier->count = 0;
barrier->waitfor = 0;
INIT_LIST_HEAD (&barrier->waitq);
- return pthread_mutex_init (&barrier->guard, 0);
+ ret = pthread_mutex_init (&barrier->guard, 0);
+ if (ret) {
+ (void)pthread_cond_destroy (&barrier->cond);
+ errno = ret;
+ return -1;
+ }
+ barrier->initialized = _gf_true;
+ return 0;
}
int
syncbarrier_destroy (struct syncbarrier *barrier)
{
+ int ret = 0;
+ int ret1 = 0;
if (!barrier) {
errno = EINVAL;
return -1;
}
- pthread_cond_destroy (&barrier->cond);
- return pthread_mutex_destroy (&barrier->guard);
+ if (barrier->initialized) {
+ ret = pthread_cond_destroy (&barrier->cond);
+ ret1 = pthread_mutex_destroy (&barrier->guard);
+ barrier->initialized = _gf_false;
+ }
+ if (ret || ret1) {
+ errno = ret?ret:ret1;
+ return -1;
+ }
+ return 0;
}
diff --git a/libglusterfs/src/syncop.h b/libglusterfs/src/syncop.h
index 5b5ad4e..9ab5ee8 100644
--- a/libglusterfs/src/syncop.h
+++ b/libglusterfs/src/syncop.h
@@ -134,6 +134,7 @@ typedef struct synclock synclock_t;
struct syncbarrier {
+ gf_boolean_t initialized; /*Set on successful initialization*/
pthread_mutex_t guard; /* guard the remaining members, pair @cond */
pthread_cond_t cond; /* waiting non-synctasks */
struct list_head waitq; /* waiting synctasks */
--
1.8.3.1

2749
0429-features-shard-Introducing-.shard-.remove_me-for-ato.patch Normal file
View File

File diff suppressed because it is too large Load Diff

1790
0430-features-shard-Perform-shards-deletion-in-the-backgr.patch Normal file
View File

File diff suppressed because it is too large Load Diff

58
0431-glusterd-Reset-op-version-for-features.shard-deletio.patch Normal file
View File

@ -0,0 +1,58 @@
From c06048a218b4a2e56f72b05b4f9f5842eec611e4 Mon Sep 17 00:00:00 2001
From: Krutika Dhananjay <kdhananj@redhat.com>
Date: Wed, 12 Sep 2018 21:41:35 +0530
Subject: [PATCH 431/444] glusterd: Reset op-version for
"features.shard-deletion-rate"
The op-version for the "features.shard-deletion-rate" option was set to
4.2.0 in the upstream patch and backported at
e75be952569eb69325d5f505f7ab94aace31be52.
This commit reverts the op-version for this option to 3.13.3.
Label: DOWNSTREAM ONLY
Change-Id: Ie3d12f3119ad7a4b40d81bd8bd6ed591658e8371
BUG: 1520882
Signed-off-by: Krutika Dhananjay <kdhananj@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/154865
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Sunil Kumar Heggodu Gopala Acharya <sheggodu@redhat.com>
---
libglusterfs/src/globals.h | 3 ++-
xlators/mgmt/glusterd/src/glusterd-volume-set.c | 2 +-
2 files changed, 3 insertions(+), 2 deletions(-)
diff --git a/libglusterfs/src/globals.h b/libglusterfs/src/globals.h
index 699e73e..97c4fad 100644
--- a/libglusterfs/src/globals.h
+++ b/libglusterfs/src/globals.h
@@ -109,11 +109,12 @@
#define GD_OP_VERSION_3_13_2 31302 /* Op-version for GlusterFS 3.13.2 */
-#define GD_OP_VERSION_4_2_0 40200 /* Op-version for GlusterFs 4.2.0 */
+#define GD_OP_VERSION_3_13_3 31303 /* Op-version for GlusterFS 3.13.3 */
/* Downstream only change */
#define GD_OP_VERSION_3_11_2 31102 /* Op-version for RHGS 3.3.1-async */
#define GD_OP_VERSION_3_13_3 31303 /* Op-version for RHGS-3.4-Batch Update-1*/
+#define GD_OP_VERSION_3_13_4 31304 /* Op-version for RHGS-3.4-Batch Update-2*/
#include "xlator.h"
diff --git a/xlators/mgmt/glusterd/src/glusterd-volume-set.c b/xlators/mgmt/glusterd/src/glusterd-volume-set.c
index 4357562..a825f52 100644
--- a/xlators/mgmt/glusterd/src/glusterd-volume-set.c
+++ b/xlators/mgmt/glusterd/src/glusterd-volume-set.c
@@ -3300,7 +3300,7 @@ struct volopt_map_entry glusterd_volopt_map[] = {
},
{ .key = "features.shard-deletion-rate",
.voltype = "features/shard",
- .op_version = GD_OP_VERSION_4_2_0,
+ .op_version = GD_OP_VERSION_3_13_4,
.flags = OPT_FLAG_CLIENT_OPT
},
{ .key = "features.scrub-throttle",
--
1.8.3.1

250
0432-features-shard-Fix-crash-and-test-case-in-RENAME-fop.patch Normal file
View File

@ -0,0 +1,250 @@
From 212e89f8b257463ace8093dfc72253f515adb234 Mon Sep 17 00:00:00 2001
From: Krutika Dhananjay <kdhananj@redhat.com>
Date: Thu, 2 Aug 2018 21:48:34 +0530
Subject: [PATCH 432/444] features/shard: Fix crash and test case in RENAME fop
> Upstream: https://review.gluster.org/20623
> BUG: 1611692
> Change-Id: Iaf85a5ee3dff8b01a76e11972f10f2bb9dcbd407
Setting the refresh flag in inode ctx in shard_rename_src_cbk()
is applicable only when the dst file exists and is sharded and
has a hard link > 1 at the time of rename.
But this piece of code is exercised even when dst doesn't exist.
In this case, the mount crashes because local->int_inodelk.loc.inode
is NULL.
Change-Id: Iaf85a5ee3dff8b01a76e11972f10f2bb9dcbd407
BUG: 1520882
Signed-off-by: Krutika Dhananjay <kdhananj@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/154866
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Sunil Kumar Heggodu Gopala Acharya <sheggodu@redhat.com>
---
tests/bugs/shard/unlinks-and-renames.t | 96 ++++++++++++++++++++--------------
xlators/features/shard/src/shard.c | 7 ++-
2 files changed, 61 insertions(+), 42 deletions(-)
diff --git a/tests/bugs/shard/unlinks-and-renames.t b/tests/bugs/shard/unlinks-and-renames.t
index 6e5164f..990ca69 100644
--- a/tests/bugs/shard/unlinks-and-renames.t
+++ b/tests/bugs/shard/unlinks-and-renames.t
@@ -31,9 +31,10 @@ TEST mkdir $M0/dir
TEST touch $M0/dir/foo
TEST touch $M0/dir/new
-######################################
-##### Unlink with /.shard absent #####
-######################################
+##########################################
+##### 01. Unlink with /.shard absent #####
+##########################################
+
TEST truncate -s 5M $M0/dir/foo
TEST ! stat $B0/${V0}0/.shard
TEST ! stat $B0/${V0}1/.shard
@@ -45,9 +46,10 @@ TEST stat $B0/${V0}1/.shard/.remove_me
EXPECT_WITHIN $FILE_COUNT_TIME 0 get_file_count $B0/${V0}0/.shard/.remove_me/$gfid_foo
EXPECT_WITHIN $FILE_COUNT_TIME 0 get_file_count $B0/${V0}1/.shard/.remove_me/$gfid_foo
-##################################################
-##### Unlink of a sharded file without holes #####
-##################################################
+######################################################
+##### 02. Unlink of a sharded file without holes #####
+######################################################
+
# Create a 9M sharded file
TEST dd if=/dev/zero of=$M0/dir/new bs=1024 count=9216
gfid_new=$(get_gfid_string $M0/dir/new)
@@ -65,9 +67,10 @@ TEST ! stat $B0/${V0}1/dir/new
EXPECT_WITHIN $FILE_COUNT_TIME 0 get_file_count $B0/${V0}0/.shard/.remove_me/$gfid_new
EXPECT_WITHIN $FILE_COUNT_TIME 0 get_file_count $B0/${V0}1/.shard/.remove_me/$gfid_new
-#######################################
-##### Unlink with /.shard present #####
-#######################################
+###########################################
+##### 03. Unlink with /.shard present #####
+###########################################
+
TEST truncate -s 5M $M0/dir/foo
gfid_foo=$(get_gfid_string $M0/dir/foo)
# Ensure its shards are absent.
@@ -81,9 +84,10 @@ TEST ! stat $M0/dir/foo
EXPECT_WITHIN $FILE_COUNT_TIME 0 get_file_count $B0/${V0}0/.shard/.remove_me/$gfid_foo
EXPECT_WITHIN $FILE_COUNT_TIME 0 get_file_count $B0/${V0}1/.shard/.remove_me/$gfid_foo
-#############################################################
-##### Unlink of a file with only one block (the zeroth) #####
-#############################################################
+#################################################################
+##### 04. Unlink of a file with only one block (the zeroth) #####
+#################################################################
+
TEST touch $M0/dir/foo
gfid_foo=$(get_gfid_string $M0/dir/foo)
TEST dd if=/dev/zero of=$M0/dir/foo bs=1024 count=1024
@@ -95,9 +99,10 @@ TEST ! stat $M0/dir/foo
EXPECT_WITHIN $FILE_COUNT_TIME 0 get_file_count $B0/${V0}0/.shard/.remove_me/$gfid_foo
EXPECT_WITHIN $FILE_COUNT_TIME 0 get_file_count $B0/${V0}1/.shard/.remove_me/$gfid_foo
-####################################################
-##### Unlink of a sharded file with hard-links #####
-####################################################
+########################################################
+##### 05. Unlink of a sharded file with hard-links #####
+########################################################
+
# Create a 9M sharded file
TEST dd if=/dev/zero of=$M0/dir/original bs=1024 count=9216
gfid_original=$(get_gfid_string $M0/dir/original)
@@ -154,9 +159,10 @@ TEST mkdir $M0/dir
TEST touch $M0/dir/src
TEST touch $M0/dir/dst
-######################################
-##### Rename with /.shard absent #####
-######################################
+##########################################
+##### 06. Rename with /.shard absent #####
+##########################################
+
TEST truncate -s 5M $M0/dir/dst
gfid_dst=$(get_gfid_string $M0/dir/dst)
TEST ! stat $B0/${V0}0/.shard
@@ -172,9 +178,10 @@ TEST stat $B0/${V0}1/dir/dst
EXPECT_WITHIN $FILE_COUNT_TIME 0 get_file_count $B0/${V0}0/.shard/.remove_me/$gfid_dst
EXPECT_WITHIN $FILE_COUNT_TIME 0 get_file_count $B0/${V0}1/.shard/.remove_me/$gfid_dst
-##################################################
-##### Rename to a sharded file without holes #####
-##################################################
+######################################################
+##### 07. Rename to a sharded file without holes #####
+######################################################
+
TEST unlink $M0/dir/dst
TEST touch $M0/dir/src
# Create a 9M sharded file
@@ -197,9 +204,10 @@ TEST stat $B0/${V0}1/dir/dst
EXPECT_WITHIN $FILE_COUNT_TIME 0 get_file_count $B0/${V0}0/.shard/.remove_me/$gfid_dst
EXPECT_WITHIN $FILE_COUNT_TIME 0 get_file_count $B0/${V0}1/.shard/.remove_me/$gfid_dst
-###################################################
-##### Rename of dst file with /.shard present #####
-###################################################
+#######################################################
+##### 08. Rename of dst file with /.shard present #####
+#######################################################
+
TEST unlink $M0/dir/dst
TEST touch $M0/dir/src
TEST truncate -s 5M $M0/dir/dst
@@ -215,9 +223,10 @@ TEST stat $B0/${V0}1/dir/dst
EXPECT_WITHIN $FILE_COUNT_TIME 0 get_file_count $B0/${V0}0/.shard/.remove_me/$gfid_dst
EXPECT_WITHIN $FILE_COUNT_TIME 0 get_file_count $B0/${V0}1/.shard/.remove_me/$gfid_dst
-###############################################################
-##### Rename of dst file with only one block (the zeroth) #####
-###############################################################
+###################################################################
+##### 09. Rename of dst file with only one block (the zeroth) #####
+###################################################################
+
TEST unlink $M0/dir/dst
TEST touch $M0/dir/src
TEST dd if=/dev/zero of=$M0/dir/dst bs=1024 count=1024
@@ -233,9 +242,10 @@ TEST stat $B0/${V0}1/dir/dst
EXPECT_WITHIN $FILE_COUNT_TIME 0 get_file_count $B0/${V0}0/.shard/.remove_me/$gfid_dst
EXPECT_WITHIN $FILE_COUNT_TIME 0 get_file_count $B0/${V0}1/.shard/.remove_me/$gfid_dst
-########################################################
-##### Rename to a dst sharded file with hard-links #####
-########################################################
+############################################################
+##### 10. Rename to a dst sharded file with hard-links #####
+############################################################
+
TEST unlink $M0/dir/dst
TEST touch $M0/dir/src
# Create a 9M sharded file
@@ -276,7 +286,10 @@ TEST ! stat $B0/${V0}1/dir/src2
EXPECT_WITHIN $FILE_COUNT_TIME 0 get_file_count $B0/${V0}0/.shard/.remove_me/$gfid_dst
EXPECT_WITHIN $FILE_COUNT_TIME 0 get_file_count $B0/${V0}1/.shard/.remove_me/$gfid_dst
-# Rename with non-existent dst and a sharded src
+##############################################################
+##### 11. Rename with non-existent dst and a sharded src #####
+##############################################################l
+
TEST touch $M0/dir/src
TEST dd if=/dev/zero of=$M0/dir/src bs=1024 count=9216
gfid_src=$(get_gfid_string $M0/dir/src)
@@ -286,7 +299,7 @@ TEST stat $B0/${V0}1/.shard/$gfid_src.1
TEST stat $B0/${V0}0/.shard/$gfid_src.2
TEST stat $B0/${V0}1/.shard/$gfid_src.2
# Now rename src to the dst.
-TEST mv $M0/dir/src $M0/dir/dst
+TEST mv $M0/dir/src $M0/dir/dst2
TEST stat $B0/${V0}0/.shard/$gfid_src.1
TEST stat $B0/${V0}1/.shard/$gfid_src.1
@@ -295,23 +308,26 @@ TEST stat $B0/${V0}1/.shard/$gfid_src.2
TEST ! stat $M0/dir/src
TEST ! stat $B0/${V0}0/dir/src
TEST ! stat $B0/${V0}1/dir/src
-TEST stat $M0/dir/dst
-TEST stat $B0/${V0}0/dir/dst
-TEST stat $B0/${V0}1/dir/dst
+TEST stat $M0/dir/dst2
+TEST stat $B0/${V0}0/dir/dst2
+TEST stat $B0/${V0}1/dir/dst2
+
+#############################################################################
+##### 12. Rename with non-existent dst and a sharded src with no shards #####
+#############################################################################
-# Rename with non-existent dst and a sharded src with no shards
TEST touch $M0/dir/src
TEST dd if=/dev/zero of=$M0/dir/src bs=1024 count=1024
gfid_src=$(get_gfid_string $M0/dir/src)
TEST ! stat $B0/${V0}0/.shard/$gfid_src.1
TEST ! stat $B0/${V0}1/.shard/$gfid_src.1
# Now rename src to the dst.
-TEST mv $M0/dir/src $M0/dir/dst
+TEST mv $M0/dir/src $M0/dir/dst1
TEST ! stat $M0/dir/src
TEST ! stat $B0/${V0}0/dir/src
TEST ! stat $B0/${V0}1/dir/src
-TEST stat $M0/dir/dst
-TEST stat $B0/${V0}0/dir/dst
-TEST stat $B0/${V0}1/dir/dst
+TEST stat $M0/dir/dst1
+TEST stat $B0/${V0}0/dir/dst1
+TEST stat $B0/${V0}1/dir/dst1
cleanup
diff --git a/xlators/features/shard/src/shard.c b/xlators/features/shard/src/shard.c
index 2faf711..6066a54 100644
--- a/xlators/features/shard/src/shard.c
+++ b/xlators/features/shard/src/shard.c
@@ -4324,9 +4324,12 @@ shard_rename_src_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
}
/* Set ctx->refresh to TRUE to force a lookup on disk when
* shard_lookup_base_file() is called next to refresh the hard link
- * count in ctx
+ * count in ctx. Note that this is applicable only to the case where
+ * the rename dst is already existent and sharded.
*/
- shard_inode_ctx_set_refresh_flag (local->int_inodelk.loc.inode, this);
+ if ((local->dst_block_size) && (!local->cleanup_required))
+ shard_inode_ctx_set_refresh_flag (local->int_inodelk.loc.inode,
+ this);
local->prebuf = *buf;
local->preoldparent = *preoldparent;
--
1.8.3.1

153
0433-mgmt-glusterd-use-proper-path-to-the-volfile.patch Normal file
View File

@ -0,0 +1,153 @@
From fd9e0103cd5c3f2962e063dbc3083c451b7e592b Mon Sep 17 00:00:00 2001
From: Raghavendra Bhat <raghavendra@redhat.com>
Date: Thu, 4 Oct 2018 14:27:45 -0400
Subject: [PATCH 433/444] mgmt/glusterd: use proper path to the volfile
> Upstream: https://review.gluster.org/#/c/glusterfs/+/21314/
> BUG: 1635050
> Change-Id: I28b2dfa5d9b379fe943db92c2fdfea879a6a594e
NOTE: This patch is actually directly applied from the patch that
was sent to the release-4.1 branch. The master branch patch
will have merge conflicts due to the clang format changes done
there. This is the patch which this commit is a backport of.
upstream(4.1): https://review.gluster.org/#/c/glusterfs/+/21348/
Chane-ID: I28b2dfa5d9b379fe943db92c2fdfea879a6a594e
Till now, glusterd was generating the volfile path for the snapshot
volume's bricks like this.
/snaps/<snap name>/<brick volfile>
But in reality, the path to the brick volfile for a snapshot volume is
/snaps/<snap name>/<snap volume name>/<brick volfile>
The above workaround was used to distinguish between a mount command used
to mount the snapshot volume, and a brick of the snapshot volume, so that
based on what is actually happening, glusterd can return the proper volfile
(client volfile for the former and the brick volfile for the latter). But,
this was causing problems for snapshot restore when brick multiplexing is
enabled. Because, with brick multiplexing, it tries to find the volfile
and sends GETSPEC rpc call to glusterd using the 2nd style of path i.e.
/snaps/<snap name>/<snap volume name>/<brick volfile>
So, when the snapshot brick (which is multiplexed) sends a GETSPEC rpc
request to glusterd for obtaining the brick volume file, glusterd was
returning the client volume file of the snapshot volume instead of the
brick volume file.
Change-Id: I28b2dfa5d9b379fe943db92c2fdfea879a6a594e
BUG: 1636291
Signed-off-by: Raghavendra Bhat <raghavendra@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/155129
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Sunil Kumar Heggodu Gopala Acharya <sheggodu@redhat.com>
---
.../snapview-server/src/snapview-server-helpers.c | 5 +++--
xlators/mgmt/glusterd/src/glusterd-handshake.c | 20 ++++++++++++++++++--
xlators/mgmt/glusterd/src/glusterd-utils.c | 9 +++++----
3 files changed, 26 insertions(+), 8 deletions(-)
diff --git a/xlators/features/snapview-server/src/snapview-server-helpers.c b/xlators/features/snapview-server/src/snapview-server-helpers.c
index 2ad74ef..4c2edc6 100644
--- a/xlators/features/snapview-server/src/snapview-server-helpers.c
+++ b/xlators/features/snapview-server/src/snapview-server-helpers.c
@@ -481,8 +481,9 @@ __svs_initialise_snapshot_volume (xlator_t *this, const char *name,
goto out;
}
- snprintf (volname, sizeof (volname), "/snaps/%s/%s",
- dirent->name, dirent->snap_volname);
+ snprintf (volname, sizeof (volname), "/snaps/%s/%s/%s",
+ dirent->name, dirent->snap_volname,
+ dirent->snap_volname);
fs = glfs_new (volname);
diff --git a/xlators/mgmt/glusterd/src/glusterd-handshake.c b/xlators/mgmt/glusterd/src/glusterd-handshake.c
index d5594d0..b2a9b20 100644
--- a/xlators/mgmt/glusterd/src/glusterd-handshake.c
+++ b/xlators/mgmt/glusterd/src/glusterd-handshake.c
@@ -52,6 +52,7 @@ get_snap_volname_and_volinfo (const char *volpath, char **volname,
char *vol = NULL;
glusterd_snap_t *snap = NULL;
xlator_t *this = NULL;
+ char *volfile_token = NULL;
this = THIS;
GF_ASSERT (this);
@@ -101,12 +102,27 @@ get_snap_volname_and_volinfo (const char *volpath, char **volname,
*/
ret = glusterd_volinfo_find (volname_token, volinfo);
if (ret) {
- *volname = gf_strdup (volname_token);
+ gf_msg (this->name, GF_LOG_WARNING, 0, GD_MSG_VOLINFO_GET_FAIL,
+ "failed to get the volinfo for the volume %s",
+ volname_token);
+
+ /* Get the actual volfile name */
+ volfile_token = strtok_r (NULL, "/", &save_ptr);
+ *volname = gf_strdup (volfile_token);
if (NULL == *volname) {
ret = -1;
goto out;
}
+ /*
+ * Ideally, this should succeed as volname_token now
+ * contains the name of the snap volume (i.e. name of
+ * the volume that represents the snapshot).
+ * But, if for some reason, volinfo for the snap volume
+ * is not found, then try to get from the name of the
+ * volfile. Name of the volfile is like this.
+ * <snap volume name>.<hostname>.<brick path>.vol
+ */
ret = glusterd_snap_volinfo_find (volname_token, snap,
volinfo);
if (ret) {
@@ -115,7 +131,7 @@ get_snap_volname_and_volinfo (const char *volpath, char **volname,
if (!vol) {
gf_msg (this->name, GF_LOG_ERROR, EINVAL,
GD_MSG_INVALID_ENTRY, "Invalid "
- "volname (%s)", volname_token);
+ "volname (%s)", volfile_token);
goto out;
}
diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.c b/xlators/mgmt/glusterd/src/glusterd-utils.c
index 04fae63..7179a68 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.c
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.c
@@ -2068,10 +2068,10 @@ retry:
}
if (volinfo->is_snap_volume) {
- snprintf (volfile, PATH_MAX,"/%s/%s/%s.%s.%s",
+ snprintf (volfile, PATH_MAX, "/%s/%s/%s/%s.%s.%s",
GLUSTERD_VOL_SNAP_DIR_PREFIX,
volinfo->snapshot->snapname, volinfo->volname,
- brickinfo->hostname, exp_path);
+ volinfo->volname, brickinfo->hostname, exp_path);
} else {
snprintf (volfile, PATH_MAX, "%s.%s.%s", volinfo->volname,
brickinfo->hostname, exp_path);
@@ -5676,10 +5676,11 @@ attach_brick (xlator_t *this,
GLUSTERD_GET_BRICK_PIDFILE (pidfile2, volinfo, brickinfo, conf);
if (volinfo->is_snap_volume) {
- snprintf (full_id, sizeof(full_id), "/%s/%s/%s.%s.%s",
+ snprintf (full_id, sizeof(full_id), "/%s/%s/%s/%s.%s.%s",
GLUSTERD_VOL_SNAP_DIR_PREFIX,
volinfo->snapshot->snapname,
- volinfo->volname, brickinfo->hostname, unslashed);
+ volinfo->volname, volinfo->volname,
+ brickinfo->hostname, unslashed);
} else {
snprintf (full_id, sizeof(full_id), "%s.%s.%s",
volinfo->volname, brickinfo->hostname, unslashed);
--
1.8.3.1

32
0434-cluster-afr-s-uuid_is_null-gf_uuid_is_null.patch Normal file
View File

@ -0,0 +1,32 @@
From f0914172f481bb32b202612b080f7902ac31ad30 Mon Sep 17 00:00:00 2001
From: Pranith Kumar K <pkarampu@redhat.com>
Date: Tue, 6 Nov 2018 21:32:55 +0530
Subject: [PATCH 434/444] cluster/afr: s/uuid_is_null/gf_uuid_is_null
BUG: 1619357
Upstream-patch: https://review.gluster.org/c/glusterfs/+/21571
Change-Id: I006116d329ac96268db132ae3aac06cc2be70e75
Signed-off-by: Pranith Kumar K <pkarampu@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/155128
Reviewed-by: Atin Mukherjee <amukherj@redhat.com>
Tested-by: RHGS Build Bot <nigelb@redhat.com>
---
xlators/cluster/afr/src/afr-common.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/xlators/cluster/afr/src/afr-common.c b/xlators/cluster/afr/src/afr-common.c
index ce2b17a..10d9620 100644
--- a/xlators/cluster/afr/src/afr-common.c
+++ b/xlators/cluster/afr/src/afr-common.c
@@ -2768,7 +2768,7 @@ afr_lookup_entry_heal (call_frame_t *frame, xlator_t *this)
continue;
if (replies[i].op_ret == 0) {
- if (uuid_is_null (gfid)) {
+ if (gf_uuid_is_null (gfid)) {
gf_uuid_copy (gfid,
replies[i].poststat.ia_gfid);
}
--
1.8.3.1

93
0435-geo-rep-Fix-traceback-with-symlink-metadata-sync.patch Normal file
View File

@ -0,0 +1,93 @@
From 7e7ffc4cc56b6b6ed460a49344082c3c25c1a23d Mon Sep 17 00:00:00 2001
From: Kotresh HR <khiremat@redhat.com>
Date: Mon, 5 Nov 2018 11:46:41 +0530
Subject: [PATCH 435/444] geo-rep: Fix traceback with symlink metadata sync
While syncing metadata, 'os.chmod', 'os.chown',
'os.utime' should be used without de-reference.
But python supports only 'os.chown' without
de-reference. That's mostly because Linux
doesn't support 'chmod' on symlink file itself
but it does support 'chown'.
So while syncing metadata ops, if it's symlink
we should only sync 'chown' and not do 'chmod'
and 'utime'. It will lead to tracebacks with
errors like EROFS, EPERM, ACCESS, ENOENT.
All the three errors (EPERM, ACCESS, ENOENT)
were handled except EROFS. But the way it was
handled was not fool proof. The operation is
tried and failure was handled based on the errors.
All the errors with symlink file for 'chown',
'utime' had to be passed to safe errors list of
'errno_wrap'. This patch handles it better by
avoiding 'chmod' and 'utime' if it's symlink
file.
Backport of:
> Patch: https://review.gluster.org/21546
> fixes: bz#1646104
> Change-Id: Ic354206455cdc7ab2a87d741d81f4efe1f19d77d
> Signed-off-by: Kotresh HR <khiremat@redhat.com>
BUG: 1645916
Change-Id: Ic354206455cdc7ab2a87d741d81f4efe1f19d77d
Signed-off-by: Kotresh HR <khiremat@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/155049
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Aravinda Vishwanathapura Krishna Murthy <avishwan@redhat.com>
Reviewed-by: Sunny Kumar <sunkumar@redhat.com>
Reviewed-by: Sunil Kumar Heggodu Gopala Acharya <sheggodu@redhat.com>
---
geo-replication/syncdaemon/resource.py | 26 +++++++++++---------------
1 file changed, 11 insertions(+), 15 deletions(-)
diff --git a/geo-replication/syncdaemon/resource.py b/geo-replication/syncdaemon/resource.py
index eb696f3..b289b3b 100644
--- a/geo-replication/syncdaemon/resource.py
+++ b/geo-replication/syncdaemon/resource.py
@@ -790,10 +790,8 @@ class Server(object):
# 'lchown' 'lchmod' 'utime with no-deference' blindly.
# But since 'lchmod' and 'utime with no de-reference' is
# not supported in python3, we have to rely on 'chmod'
- # and 'utime with de-reference'. But 'chmod'
- # de-reference the symlink and gets ENOENT, EACCES,
- # EPERM errors, hence ignoring those errors if it's on
- # symlink file.
+ # and 'utime with de-reference'. Hence avoiding 'chmod'
+ # and 'utime' if it's symlink file.
is_symlink = False
cmd_ret = errno_wrap(os.lchown, [go, uid, gid], [ENOENT],
@@ -801,19 +799,17 @@ class Server(object):
if isinstance(cmd_ret, int):
continue
- cmd_ret = errno_wrap(os.chmod, [go, mode],
- [ENOENT, EACCES, EPERM], [ESTALE, EINVAL])
- if isinstance(cmd_ret, int):
- is_symlink = os.path.islink(go)
- if not is_symlink:
+ is_symlink = os.path.islink(go)
+
+ if not is_symlink:
+ cmd_ret = errno_wrap(os.chmod, [go, mode],
+ [ENOENT, EACCES, EPERM], [ESTALE, EINVAL])
+ if isinstance(cmd_ret, int):
failures.append((e, cmd_ret, "chmod"))
- cmd_ret = errno_wrap(os.utime, [go, (atime, mtime)],
- [ENOENT, EACCES, EPERM], [ESTALE, EINVAL])
- if isinstance(cmd_ret, int):
- if not is_symlink:
- is_symlink = os.path.islink(go)
- if not is_symlink:
+ cmd_ret = errno_wrap(os.utime, [go, (atime, mtime)],
+ [ENOENT, EACCES, EPERM], [ESTALE, EINVAL])
+ if isinstance(cmd_ret, int):
failures.append((e, cmd_ret, "utime"))
return failures
--
1.8.3.1

204
0436-geo-rep-Fix-issue-in-gfid-conflict-resolution.patch Normal file
View File

@ -0,0 +1,204 @@
From f42b8789cdcd93cb9fa93f35ed067268ce75f789 Mon Sep 17 00:00:00 2001
From: Kotresh HR <khiremat@redhat.com>
Date: Thu, 25 Oct 2018 03:23:56 -0400
Subject: [PATCH 436/444] geo-rep: Fix issue in gfid-conflict-resolution
Problem:
During gfid-conflict-resolution, geo-rep crashes
with 'ValueError: list.remove(x): x not in list'
Cause and Analysis:
During gfid-conflict-resolution, the entry blob is
passed back to master along with additional
information to verify it's integrity. If everything
looks fine, the entry creation is ignored and is
deleted from the original list. But it is crashing
during removal of entry from the list saying entry
not in list. The reason is that the stat information
in the entry blob was modified and sent back to
master if present.
Fix:
Send back the correct stat information for
gfid-conflict-resolution.
Backport of:
> Patch: https://review.gluster.org/21483
> fixes: bz#1642865
> Change-Id: I47a6aa60b2a495465aa9314eebcb4085f0b1c4fd
> Signed-off-by: Kotresh HR <khiremat@redhat.com>
BUG: 1640347
Change-Id: I47a6aa60b2a495465aa9314eebcb4085f0b1c4fd
Signed-off-by: Kotresh HR <khiremat@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/155038
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Aravinda Vishwanathapura Krishna Murthy <avishwan@redhat.com>
Reviewed-by: Sunny Kumar <sunkumar@redhat.com>
Reviewed-by: Sunil Kumar Heggodu Gopala Acharya <sheggodu@redhat.com>
---
geo-replication/syncdaemon/resource.py | 42 +++++++++++++++++++---------------
1 file changed, 24 insertions(+), 18 deletions(-)
diff --git a/geo-replication/syncdaemon/resource.py b/geo-replication/syncdaemon/resource.py
index b289b3b..f16066e 100644
--- a/geo-replication/syncdaemon/resource.py
+++ b/geo-replication/syncdaemon/resource.py
@@ -456,7 +456,7 @@ class Server(object):
st['uid'], st['gid'],
gf, st['mode'], bn, lnk)
- def entry_purge(op, entry, gfid, e):
+ def entry_purge(op, entry, gfid, e, uid, gid):
# This is an extremely racy code and needs to be fixed ASAP.
# The GFID check here is to be sure that the pargfid/bname
# to be purged is the GFID gotten from the changelog.
@@ -470,7 +470,7 @@ class Server(object):
return
if not matching_disk_gfid(gfid, entry):
- collect_failure(e, EEXIST)
+ collect_failure(e, EEXIST, uid, gid)
return
if op == 'UNLINK':
@@ -486,7 +486,7 @@ class Server(object):
if er == ENOTEMPTY:
return er
- def collect_failure(e, cmd_ret, dst=False):
+ def collect_failure(e, cmd_ret, uid, gid, dst=False):
slv_entry_info = {}
slv_entry_info['gfid_mismatch'] = False
slv_entry_info['name_mismatch'] = False
@@ -499,6 +499,11 @@ class Server(object):
if cmd_ret is None:
return False
+ if e.get("stat", {}):
+ # Copy actual UID/GID value back to entry stat
+ e['stat']['uid'] = uid
+ e['stat']['gid'] = gid
+
if cmd_ret == EEXIST:
if dst:
en = e['entry1']
@@ -559,7 +564,7 @@ class Server(object):
errno_wrap(os.rmdir, [path], [ENOENT, ESTALE], [EBUSY])
- def rename_with_disk_gfid_confirmation(gfid, entry, en):
+ def rename_with_disk_gfid_confirmation(gfid, entry, en, uid, gid):
if not matching_disk_gfid(gfid, entry):
logging.error(lf("RENAME ignored: source entry does not match "
"with on-disk gfid",
@@ -567,14 +572,13 @@ class Server(object):
gfid=gfid,
disk_gfid=get_gfid_from_mnt(entry),
target=en))
- collect_failure(e, EEXIST)
+ collect_failure(e, EEXIST, uid, gid)
return
cmd_ret = errno_wrap(os.rename,
[entry, en],
[ENOENT, EEXIST], [ESTALE, EBUSY])
- collect_failure(e, cmd_ret)
-
+ collect_failure(e, cmd_ret, uid, gid)
for e in entries:
blob = None
@@ -595,7 +599,7 @@ class Server(object):
if op in ['RMDIR', 'UNLINK']:
# Try once, if rmdir failed with ENOTEMPTY
# then delete recursively.
- er = entry_purge(op, entry, gfid, e)
+ er = entry_purge(op, entry, gfid, e, uid, gid)
if isinstance(er, int):
if er == ENOTEMPTY and op == 'RMDIR':
# Retry if ENOTEMPTY, ESTALE
@@ -632,7 +636,7 @@ class Server(object):
cmd_ret = errno_wrap(os.link,
[slink, entry],
[ENOENT, EEXIST], [ESTALE])
- collect_failure(e, cmd_ret)
+ collect_failure(e, cmd_ret, uid, gid)
elif op == 'MKDIR':
en = e['entry']
slink = os.path.join(pfx, gfid)
@@ -676,7 +680,7 @@ class Server(object):
cmd_ret = errno_wrap(os.link,
[slink, entry],
[ENOENT, EEXIST], [ESTALE])
- collect_failure(e, cmd_ret)
+ collect_failure(e, cmd_ret, uid, gid)
elif op == 'SYMLINK':
en = e['entry']
st = lstat(entry)
@@ -684,7 +688,7 @@ class Server(object):
blob = entry_pack_symlink(gfid, bname, e['link'],
e['stat'])
elif not matching_disk_gfid(gfid, en):
- collect_failure(e, EEXIST)
+ collect_failure(e, EEXIST, uid, gid)
elif op == 'RENAME':
en = e['entry1']
# The matching disk gfid check validates two things
@@ -704,7 +708,7 @@ class Server(object):
blob = entry_pack_symlink(gfid, bname,
e['link'], e['stat'])
elif not matching_disk_gfid(gfid, en):
- collect_failure(e, EEXIST, True)
+ collect_failure(e, EEXIST, uid, gid, True)
else:
slink = os.path.join(pfx, gfid)
st = lstat(slink)
@@ -716,12 +720,13 @@ class Server(object):
else:
cmd_ret = errno_wrap(os.link, [slink, en],
[ENOENT, EEXIST], [ESTALE])
- collect_failure(e, cmd_ret)
+ collect_failure(e, cmd_ret, uid, gid)
else:
st = lstat(entry)
st1 = lstat(en)
if isinstance(st1, int):
- rename_with_disk_gfid_confirmation(gfid, entry, en)
+ rename_with_disk_gfid_confirmation(gfid, entry, en,
+ uid, gid)
else:
if st.st_ino == st1.st_ino:
# we have a hard link, we can now unlink source
@@ -746,15 +751,16 @@ class Server(object):
else:
raise
elif not matching_disk_gfid(gfid, en):
- collect_failure(e, EEXIST, True)
+ collect_failure(e, EEXIST, uid, gid, True)
else:
- rename_with_disk_gfid_confirmation(gfid, entry, en)
+ rename_with_disk_gfid_confirmation(gfid, entry, en,
+ uid, gid)
if blob:
cmd_ret = errno_wrap(Xattr.lsetxattr,
[pg, 'glusterfs.gfid.newfile', blob],
[EEXIST, ENOENT],
[ESTALE, EINVAL, EBUSY])
- failed = collect_failure(e, cmd_ret)
+ collect_failure(e, cmd_ret, uid, gid)
# If UID/GID is different than zero that means we are trying
# create Entry with different UID/GID. Create Entry with
@@ -763,7 +769,7 @@ class Server(object):
path = os.path.join(pfx, gfid)
cmd_ret = errno_wrap(os.lchown, [path, uid, gid], [ENOENT],
[ESTALE, EINVAL])
- collect_failure(e, cmd_ret)
+ collect_failure(e, cmd_ret, uid, gid)
return failures
--
1.8.3.1

144
0437-geo-rep-Add-more-intelligence-to-automatic-error-han.patch Normal file
View File

@ -0,0 +1,144 @@
From 85da98b9c54889139822b5c3d351a0249abf75b0 Mon Sep 17 00:00:00 2001
From: Kotresh HR <khiremat@redhat.com>
Date: Fri, 26 Oct 2018 03:45:46 -0400
Subject: [PATCH 437/444] geo-rep: Add more intelligence to automatic error
handling
Geo-rep's automatic error handling does gfid conflict
resolution. But if there are ENOENT errors because the
parent is not synced to slave, it doesn' handle them.
This patch adds the intelligence to create missing
parent directories on slave. It can create the missing
directories upto the depth of 10.
Backport of:
> Patch: https://review.gluster.org/21498
> fixes: bz#1643402
> Change-Id: Ic97ed1fa5899c087e404d559e04f7963ed7bb54c
> Signed-off-by: Kotresh HR <khiremat@redhat.com>
BUG: 1638069
Change-Id: Ic97ed1fa5899c087e404d559e04f7963ed7bb54c
Signed-off-by: Kotresh HR <khiremat@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/155039
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Aravinda Vishwanathapura Krishna Murthy <avishwan@redhat.com>
Reviewed-by: Sunny Kumar <sunkumar@redhat.com>
Reviewed-by: Sunil Kumar Heggodu Gopala Acharya <sheggodu@redhat.com>
---
geo-replication/syncdaemon/master.py | 68 ++++++++++++++++++++++++------------
1 file changed, 46 insertions(+), 22 deletions(-)
diff --git a/geo-replication/syncdaemon/master.py b/geo-replication/syncdaemon/master.py
index cd135df..bdb4da2 100644
--- a/geo-replication/syncdaemon/master.py
+++ b/geo-replication/syncdaemon/master.py
@@ -693,7 +693,7 @@ class GMasterChangelogMixin(GMasterCommon):
TYPE_ENTRY = "E "
MAX_EF_RETRIES = 10
- MAX_OE_RETRIES = 5
+ MAX_OE_RETRIES = 10
# flat directory hierarchy for gfid based access
FLAT_DIR_HIERARCHY = '.'
@@ -836,11 +836,12 @@ class GMasterChangelogMixin(GMasterCommon):
# The file exists on master but with different name.
# Probably renamed and got missed during xsync crawl.
elif failure[2]['slave_isdir']:
- realpath = os.readlink(os.path.join(gconf.local_path,
- ".glusterfs",
- slave_gfid[0:2],
- slave_gfid[2:4],
- slave_gfid))
+ realpath = os.readlink(os.path.join(
+ gconf.local_path,
+ ".glusterfs",
+ slave_gfid[0:2],
+ slave_gfid[2:4],
+ slave_gfid))
dst_entry = os.path.join(pfx, realpath.split('/')[-2],
realpath.split('/')[-1])
src_entry = pbname
@@ -881,25 +882,37 @@ class GMasterChangelogMixin(GMasterCommon):
gfid=failure[2]['slave_gfid'],
entry=pbname))
elif failure[1] == ENOENT:
- # Ignore ENOENT error for fix_entry_ops aka retry_count > 1
- if retry_count > 1:
- logging.info(lf('ENOENT error while fixing entry ops. '
- 'Safe to ignore, take out entry',
+ if op in ['RENAME']:
+ pbname = failure[0]['entry1']
+ else:
+ pbname = failure[0]['entry']
+
+ pargfid = pbname.split('/')[1]
+ st = lstat(os.path.join(pfx, pargfid))
+ # Safe to ignore the failure as master doesn't contain
+ # parent directory.
+ if isinstance(st, int):
+ logging.info(lf('Fixing ENOENT error in slave. Parent '
+ 'does not exist on master. Safe to '
+ 'ignore, take out entry',
retry_count=retry_count,
entry=repr(failure)))
entries.remove(failure[0])
- elif op in ('MKNOD', 'CREATE', 'MKDIR'):
- pargfid = pbname.split('/')[1]
- st = lstat(os.path.join(pfx, pargfid))
- # Safe to ignore the failure as master doesn't contain
- # parent directory.
- if isinstance(st, int):
- logging.info(lf('Fixing ENOENT error in slave. Parent '
- 'does not exist on master. Safe to '
- 'ignore, take out entry',
- retry_count=retry_count,
- entry=repr(failure)))
- entries.remove(failure[0])
+ else:
+ logging.info(lf('Fixing ENOENT error in slave. Create '
+ 'parent directory on slave.',
+ retry_count=retry_count,
+ entry=repr(failure)))
+ realpath = os.readlink(os.path.join(gconf.local_path,
+ ".glusterfs",
+ pargfid[0:2],
+ pargfid[2:4],
+ pargfid))
+ dir_entry = os.path.join(pfx, realpath.split('/')[-2],
+ realpath.split('/')[-1])
+ fix_entry_ops.append(
+ edct('MKDIR', gfid=pargfid, entry=dir_entry,
+ mode=st.st_mode, uid=st.st_uid, gid=st.st_gid))
if fix_entry_ops:
# Process deletions of entries whose gfids are mismatched
@@ -1077,6 +1090,11 @@ class GMasterChangelogMixin(GMasterCommon):
os.path.join(pfx, ec[self.POS_ENTRY1 - 1]))
entries.append(edct(ty, gfid=gfid, entry=e1, entry1=en,
stat=st, link=rl))
+ # If src doesn't exist while doing rename, destination
+ # is created. If data is not followed by rename, this
+ # remains zero byte file on slave. Hence add data entry
+ # for renames
+ datas.add(os.path.join(pfx, gfid))
else:
# stat() to get mode and other information
if not matching_disk_gfid(gfid, en):
@@ -1100,6 +1118,12 @@ class GMasterChangelogMixin(GMasterCommon):
rl = None
entries.append(edct(ty, stat=st, entry=en, gfid=gfid,
link=rl))
+ # If src doesn't exist while doing link, destination
+ # is created based on file type. If data is not
+ # followed by link, this remains zero byte file on
+ # slave. Hence add data entry for links
+ if rl is None:
+ datas.add(os.path.join(pfx, gfid))
elif ty == 'SYMLINK':
rl = errno_wrap(os.readlink, [en], [ENOENT],
[ESTALE, EINTR])
--
1.8.3.1

365
0438-cluster-dht-In-rename-unlink-after-creating-linkto-f.patch Normal file
View File

@ -0,0 +1,365 @@
From e76a777f3820e62948256a45a38d5e97f3eb08a9 Mon Sep 17 00:00:00 2001
From: N Balachandran <nbalacha@redhat.com>
Date: Tue, 28 Aug 2018 12:00:33 +0530
Subject: [PATCH 438/444] cluster/dht: In rename, unlink after creating linkto
file
The linkto file creation for the dst was done in parallel with
the unlink of the old src linkto. If these operations reached
the brick out of order, we end up with a dst linkto file without
a .glusterfs handle.
Fixed by unlinking only after the linkto file creation has
completed.
upstream patch: https://review.gluster.org/#/c/glusterfs/+/21023/
> Change-Id: I4246f7655f5bc180f5ded7fd34d263b7828a8110
> fixes: bz#1621981
> Signed-off-by: N Balachandran <nbalacha@redhat.com>
Change-Id: Ia845a68bb314997cadab57887a84dff9373400c4
BUG: 1622001
Signed-off-by: N Balachandran <nbalacha@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/154933
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Raghavendra Gowdappa <rgowdapp@redhat.com>
---
tests/bugs/posix/bug-1619720.t | 1 +
xlators/cluster/dht/src/dht-rename.c | 254 ++++++++++++++++++-----------------
2 files changed, 133 insertions(+), 122 deletions(-)
diff --git a/tests/bugs/posix/bug-1619720.t b/tests/bugs/posix/bug-1619720.t
index 5e0d0f7..bfd304d 100755
--- a/tests/bugs/posix/bug-1619720.t
+++ b/tests/bugs/posix/bug-1619720.t
@@ -48,6 +48,7 @@ TEST mv $M0/tmp/file-2 $M0/tmp/file-3
TEST mv -f $M0/tmp/file-1 $M0/tmp/file-3
+
TEST getfattr -n $pgfid_xattr_name $B0/${V0}0/tmp/file-3
TEST getfattr -n $pgfid_xattr_name $B0/${V0}1/tmp/file-3
diff --git a/xlators/cluster/dht/src/dht-rename.c b/xlators/cluster/dht/src/dht-rename.c
index 1d0c2bb..378cb0a 100644
--- a/xlators/cluster/dht/src/dht-rename.c
+++ b/xlators/cluster/dht/src/dht-rename.c
@@ -849,8 +849,8 @@ dht_rename_cleanup (call_frame_t *frame)
if (src_cached == dst_cached)
goto nolinks;
- if (local->linked && (dst_hashed != src_hashed )&&
- (dst_hashed != src_cached)) {
+ if (local->linked && (dst_hashed != src_hashed) &&
+ (dst_hashed != src_cached)) {
call_cnt++;
}
@@ -935,6 +935,120 @@ nolinks:
int
+dht_rename_unlink (call_frame_t *frame, xlator_t *this)
+{
+ dht_local_t *local = NULL;
+ xlator_t *src_hashed = NULL;
+ xlator_t *src_cached = NULL;
+ xlator_t *dst_hashed = NULL;
+ xlator_t *dst_cached = NULL;
+ xlator_t *rename_subvol = NULL;
+ dict_t *xattr = NULL;
+
+ local = frame->local;
+
+ src_hashed = local->src_hashed;
+ src_cached = local->src_cached;
+ dst_hashed = local->dst_hashed;
+ dst_cached = local->dst_cached;
+
+ local->call_cnt = 0;
+
+ /* NOTE: rename_subvol is the same subvolume from which dht_rename_cbk
+ * is called. since rename has already happened on rename_subvol,
+ * unlink shouldn't be sent for oldpath (either linkfile or cached-file)
+ * on rename_subvol. */
+ if (src_cached == dst_cached)
+ rename_subvol = src_cached;
+ else
+ rename_subvol = dst_hashed;
+
+ /* TODO: delete files in background */
+
+ if (src_cached != dst_hashed && src_cached != dst_cached)
+ local->call_cnt++;
+
+ if (src_hashed != rename_subvol && src_hashed != src_cached)
+ local->call_cnt++;
+
+ if (dst_cached && dst_cached != dst_hashed && dst_cached != src_cached)
+ local->call_cnt++;
+
+ if (local->call_cnt == 0)
+ goto unwind;
+
+ DHT_MARK_FOP_INTERNAL (xattr);
+
+ if (src_cached != dst_hashed && src_cached != dst_cached) {
+ dict_t *xattr_new = NULL;
+
+ xattr_new = dict_copy_with_ref (xattr, NULL);
+
+ gf_msg_trace (this->name, 0,
+ "deleting old src datafile %s @ %s",
+ local->loc.path, src_cached->name);
+
+ if (gf_uuid_compare (local->loc.pargfid,
+ local->loc2.pargfid) == 0) {
+ DHT_MARKER_DONT_ACCOUNT(xattr_new);
+ }
+
+ DHT_CHANGELOG_TRACK_AS_RENAME(xattr_new, &local->loc,
+ &local->loc2);
+ STACK_WIND_COOKIE (frame, dht_rename_unlink_cbk, src_cached,
+ src_cached, src_cached->fops->unlink,
+ &local->loc, 0, xattr_new);
+
+ dict_unref (xattr_new);
+ xattr_new = NULL;
+ }
+
+ if (src_hashed != rename_subvol && src_hashed != src_cached) {
+ dict_t *xattr_new = NULL;
+
+ xattr_new = dict_copy_with_ref (xattr, NULL);
+
+ gf_msg_trace (this->name, 0,
+ "deleting old src linkfile %s @ %s",
+ local->loc.path, src_hashed->name);
+
+ DHT_MARKER_DONT_ACCOUNT(xattr_new);
+
+ STACK_WIND_COOKIE (frame, dht_rename_unlink_cbk, src_hashed,
+ src_hashed, src_hashed->fops->unlink,
+ &local->loc, 0, xattr_new);
+
+ dict_unref (xattr_new);
+ xattr_new = NULL;
+ }
+
+ if (dst_cached &&
+ (dst_cached != dst_hashed) &&
+ (dst_cached != src_cached)) {
+ gf_msg_trace (this->name, 0,
+ "deleting old dst datafile %s @ %s",
+ local->loc2.path, dst_cached->name);
+
+ STACK_WIND_COOKIE (frame, dht_rename_unlink_cbk, dst_cached,
+ dst_cached, dst_cached->fops->unlink,
+ &local->loc2, 0, xattr);
+ }
+ if (xattr)
+ dict_unref (xattr);
+ return 0;
+
+unwind:
+ WIPE (&local->preoldparent);
+ WIPE (&local->postoldparent);
+ WIPE (&local->preparent);
+ WIPE (&local->postparent);
+
+ dht_rename_done (frame, this);
+
+ return 0;
+}
+
+int
dht_rename_links_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno,
inode_t *inode, struct iatt *stbuf,
@@ -947,6 +1061,7 @@ dht_rename_links_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
prev = cookie;
local = frame->local;
+ /* TODO: Handle this case in lookup-optimize */
if (op_ret == -1) {
gf_msg (this->name, GF_LOG_WARNING, op_errno,
DHT_MSG_CREATE_LINK_FAILED,
@@ -958,8 +1073,8 @@ dht_rename_links_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local->linked = _gf_false;
dht_linkfile_attr_heal (frame, this);
}
- DHT_STACK_DESTROY (frame);
+ dht_rename_unlink (frame, this);
return 0;
}
@@ -973,19 +1088,14 @@ dht_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
{
dht_local_t *local = NULL;
xlator_t *prev = NULL;
- xlator_t *src_hashed = NULL;
xlator_t *src_cached = NULL;
xlator_t *dst_hashed = NULL;
xlator_t *dst_cached = NULL;
- xlator_t *rename_subvol = NULL;
- call_frame_t *link_frame = NULL;
- dht_local_t *link_local = NULL;
- dict_t *xattr = NULL;
+ loc_t link_loc = {0};
local = frame->local;
prev = cookie;
- src_hashed = local->src_hashed;
src_cached = local->src_cached;
dst_hashed = local->dst_hashed;
dst_cached = local->dst_cached;
@@ -1043,31 +1153,6 @@ dht_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local->xattr = dict_copy_with_ref (xdata, local->xattr);
}
- if ((src_cached == dst_cached) && (dst_hashed != dst_cached)) {
- link_frame = copy_frame (frame);
- if (!link_frame) {
- goto err;
- }
-
- /* fop value sent as maxvalue because it is not used
- anywhere in this case */
- link_local = dht_local_init (link_frame, &local->loc2, NULL,
- GF_FOP_MAXVALUE);
- if (!link_local) {
- goto err;
- }
-
- if (link_local->loc.inode)
- inode_unref (link_local->loc.inode);
- link_local->loc.inode = inode_ref (local->loc.inode);
- gf_uuid_copy (link_local->gfid, local->loc.inode->gfid);
-
- dht_linkfile_create (link_frame, dht_rename_links_create_cbk,
- this, src_cached, dst_hashed,
- &link_local->loc);
- }
-
-err:
/* Merge attrs only from src_cached. In case there of src_cached !=
* dst_hashed, this ignores linkfile attrs. */
if (prev == src_cached) {
@@ -1080,98 +1165,23 @@ err:
dht_iatt_merge (this, &local->postparent, postnewparent, prev);
}
+ /* Create the linkto file for the dst file */
+ if ((src_cached == dst_cached) && (dst_hashed != dst_cached)) {
- /* NOTE: rename_subvol is the same subvolume from which dht_rename_cbk
- * is called. since rename has already happened on rename_subvol,
- * unlink should not be sent for oldpath (either linkfile or cached-file)
- * on rename_subvol. */
- if (src_cached == dst_cached)
- rename_subvol = src_cached;
- else
- rename_subvol = dst_hashed;
-
- /* TODO: delete files in background */
-
- if (src_cached != dst_hashed && src_cached != dst_cached)
- local->call_cnt++;
-
- if (src_hashed != rename_subvol && src_hashed != src_cached)
- local->call_cnt++;
-
- if (dst_cached && dst_cached != dst_hashed && dst_cached != src_cached)
- local->call_cnt++;
-
- if (local->call_cnt == 0)
- goto unwind;
-
- DHT_MARK_FOP_INTERNAL (xattr);
-
- if (src_cached != dst_hashed && src_cached != dst_cached) {
- dict_t *xattr_new = NULL;
-
- xattr_new = dict_copy_with_ref (xattr, NULL);
-
- gf_msg_trace (this->name, 0,
- "deleting old src datafile %s @ %s",
- local->loc.path, src_cached->name);
-
- if (gf_uuid_compare (local->loc.pargfid,
- local->loc2.pargfid) == 0) {
- DHT_MARKER_DONT_ACCOUNT(xattr_new);
- }
-
- DHT_CHANGELOG_TRACK_AS_RENAME(xattr_new, &local->loc,
- &local->loc2);
- STACK_WIND_COOKIE (frame, dht_rename_unlink_cbk, src_cached,
- src_cached, src_cached->fops->unlink,
- &local->loc, 0, xattr_new);
-
- dict_unref (xattr_new);
- xattr_new = NULL;
- }
-
- if (src_hashed != rename_subvol && src_hashed != src_cached) {
- dict_t *xattr_new = NULL;
-
- xattr_new = dict_copy_with_ref (xattr, NULL);
-
- gf_msg_trace (this->name, 0,
- "deleting old src linkfile %s @ %s",
- local->loc.path, src_hashed->name);
-
- DHT_MARKER_DONT_ACCOUNT(xattr_new);
-
- STACK_WIND_COOKIE (frame, dht_rename_unlink_cbk, src_hashed,
- src_hashed, src_hashed->fops->unlink,
- &local->loc, 0, xattr_new);
-
- dict_unref (xattr_new);
- xattr_new = NULL;
- }
+ loc_copy (&link_loc, &local->loc2);
+ if (link_loc.inode)
+ inode_unref (link_loc.inode);
+ link_loc.inode = inode_ref (local->loc.inode);
+ gf_uuid_copy (local->gfid, local->loc.inode->gfid);
+ gf_uuid_copy (link_loc.gfid, local->loc.inode->gfid);
- if (dst_cached
- && (dst_cached != dst_hashed)
- && (dst_cached != src_cached)) {
- gf_msg_trace (this->name, 0,
- "deleting old dst datafile %s @ %s",
- local->loc2.path, dst_cached->name);
-
- STACK_WIND_COOKIE (frame, dht_rename_unlink_cbk, dst_cached,
- dst_cached, dst_cached->fops->unlink,
- &local->loc2, 0, xattr);
+ dht_linkfile_create (frame, dht_rename_links_create_cbk,
+ this, src_cached, dst_hashed,
+ &link_loc);
+ return 0;
}
- if (xattr)
- dict_unref (xattr);
- return 0;
-
-unwind:
- WIPE (&local->preoldparent);
- WIPE (&local->postoldparent);
- WIPE (&local->preparent);
- WIPE (&local->postparent);
-
- dht_rename_done (frame, this);
+ dht_rename_unlink (frame, this);
return 0;
cleanup:
--
1.8.3.1

70
0439-cluster-dht-fixes-to-unlinking-invalid-linkto-file.patch Normal file
View File

@ -0,0 +1,70 @@
From 07ae526af10de814d174189ff41709cf781ace9c Mon Sep 17 00:00:00 2001
From: Raghavendra Gowdappa <rgowdapp@redhat.com>
Date: Tue, 30 Oct 2018 12:15:35 +0530
Subject: [PATCH 439/444] cluster/dht: fixes to unlinking invalid linkto file
If unlinking of an invalid linkto file failed in lookup-everywhere
codepath, lookup was failed with EIO. The rational as per the comment
was,
<snip>
/* When dht_lookup_everywhere is performed, one cached
*and one hashed file was found and hashed file does
*not point to the above mentioned cached node. So it
*was considered as stale and an unlink was performed.
*But unlink fails. So may be rebalance is in progress.
*now ideally we have two data-files. One obtained during
*lookup_everywhere and one where unlink-failed. So
*at this point in time we cannot decide which one to
*choose because there are chances of first cached
*file is truncated after rebalance and if it is chosen
*as cached node, application will fail. So return EIO.
*/
</snip>
However, this reasoning is only valid when
* op_errno is EBUSY, indicating rebalance is in progress
* op_errno is ENOTCONN as wecannot determine what was the status of
file on brick.
Hence this patch doesn't fail lookup unless unlink fails with an
either EBUSY or ENOTCONN
>Change-Id: Ife55f3d97fe557f3db05beae0c2d786df31e8e55
>Fixes: bz#1635145
>Signed-off-by: Raghavendra Gowdappa <rgowdapp@redhat.com>
Change-Id: Ife55f3d97fe557f3db05beae0c2d786df31e8e55
BUG: 1634649
upstream patch: http://review.gluster.org/r/Ife55f3d97fe557f3db05beae0c2d786df31e8e55
Signed-off-by: Raghavendra Gowdappa <rgowdapp@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/155102
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Nithya Balachandran <nbalacha@redhat.com>
---
xlators/cluster/dht/src/dht-common.c | 7 +++----
1 file changed, 3 insertions(+), 4 deletions(-)
diff --git a/xlators/cluster/dht/src/dht-common.c b/xlators/cluster/dht/src/dht-common.c
index 0984f8f..d3a0c8b 100644
--- a/xlators/cluster/dht/src/dht-common.c
+++ b/xlators/cluster/dht/src/dht-common.c
@@ -2104,10 +2104,9 @@ dht_lookup_unlink_of_false_linkto_cbk (call_frame_t *frame, void *cookie,
this_call_cnt = dht_frame_return (frame);
if (is_last_call (this_call_cnt)) {
-
- if (op_ret == 0) {
- dht_lookup_everywhere_done (frame, this);
- } else {
+ if ((op_ret == 0) || ((op_errno != EBUSY) && (op_errno != ENOTCONN))) {
+ dht_lookup_everywhere_done (frame, this);
+ } else {
/*When dht_lookup_everywhere is performed, one cached
*and one hashed file was found and hashed file does
*not point to the above mentioned cached node. So it
--
1.8.3.1

34
0440-features-locks-Use-pthread_mutex_unlock-instead-of-p.patch Normal file
View File

@ -0,0 +1,34 @@
From efd713e5f9067a743f532923c529416fc5f5189e Mon Sep 17 00:00:00 2001
From: Susant Palai <spalai@redhat.com>
Date: Wed, 7 Nov 2018 13:06:07 +0530
Subject: [PATCH 440/444] features/locks:Use pthread_mutex_unlock() instead of
pthread_mutex_lock()
Upstream patch: https://review.gluster.org/#/c/glusterfs/+/21579/
Change-Id: I85ea6e351f07cc289245cfb501f027942ee3176c
BUG: 1647675
Signed-off-by: Susant Palai <spalai@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/155326
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Sunil Kumar Heggodu Gopala Acharya <sheggodu@redhat.com>
---
xlators/features/locks/src/posix.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/xlators/features/locks/src/posix.c b/xlators/features/locks/src/posix.c
index b434a08..2cc2837 100644
--- a/xlators/features/locks/src/posix.c
+++ b/xlators/features/locks/src/posix.c
@@ -2966,7 +2966,7 @@ pl_metalk (call_frame_t *frame, xlator_t *this, inode_t *inode)
ret = -1;
}
}
- pthread_mutex_lock (&pl_inode->mutex);
+ pthread_mutex_unlock (&pl_inode->mutex);
if (ret == -1) {
goto out;
--
1.8.3.1

201
0441-features-shard-Make-lru-limit-of-inode-list-configur.patch Normal file
View File

@ -0,0 +1,201 @@
From 7b12a7ea7a6b4945ad52f218b187ca440dfbef63 Mon Sep 17 00:00:00 2001
From: Krutika Dhananjay <kdhananj@redhat.com>
Date: Fri, 20 Jul 2018 10:52:22 +0530
Subject: [PATCH 441/444] features/shard: Make lru limit of inode list
configurable
> Upstream: https://review.gluster.org/20544
> BUG: 1605056
> Change-Id: Ifdcc2099f634314fafe8444e2d676e192e89e295
Currently this lru limit is hard-coded to 16384. This patch makes it
configurable to make it easier to hit the lru limit and enable testing
of different cases that arise when the limit is reached.
The option is features.shard-lru-limit. It is by design allowed to
be configured only in init() but not in reconfigure(). This is to avoid
all the complexity associated with eviction of least recently used shards
when the list is shrunk.
Change-Id: Ifdcc2099f634314fafe8444e2d676e192e89e295
BUG: 1603118
Signed-off-by: Krutika Dhananjay <kdhananj@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/155126
Reviewed-by: Xavi Hernandez <xhernandez@redhat.com>
Tested-by: RHGS Build Bot <nigelb@redhat.com>
---
libglusterfs/src/globals.h | 4 ++-
tests/bugs/shard/configure-lru-limit.t | 48 +++++++++++++++++++++++++
xlators/features/shard/src/shard.c | 19 ++++++++--
xlators/features/shard/src/shard.h | 3 +-
xlators/mgmt/glusterd/src/glusterd-volume-set.c | 6 ++++
5 files changed, 75 insertions(+), 5 deletions(-)
create mode 100644 tests/bugs/shard/configure-lru-limit.t
diff --git a/libglusterfs/src/globals.h b/libglusterfs/src/globals.h
index 97c4fad..555f44b 100644
--- a/libglusterfs/src/globals.h
+++ b/libglusterfs/src/globals.h
@@ -43,7 +43,7 @@
*/
#define GD_OP_VERSION_MIN 1 /* MIN is the fresh start op-version, mostly
should not change */
-#define GD_OP_VERSION_MAX GD_OP_VERSION_3_13_3 /* MAX VERSION is the maximum
+#define GD_OP_VERSION_MAX GD_OP_VERSION_4_2_0 /* MAX VERSION is the maximum
count in VME table, should
keep changing with
introduction of newer
@@ -111,6 +111,8 @@
#define GD_OP_VERSION_3_13_3 31303 /* Op-version for GlusterFS 3.13.3 */
+#define GD_OP_VERSION_4_2_0 40200 /* Op-version for GlusterFS 4.2.0 */
+
/* Downstream only change */
#define GD_OP_VERSION_3_11_2 31102 /* Op-version for RHGS 3.3.1-async */
#define GD_OP_VERSION_3_13_3 31303 /* Op-version for RHGS-3.4-Batch Update-1*/
diff --git a/tests/bugs/shard/configure-lru-limit.t b/tests/bugs/shard/configure-lru-limit.t
new file mode 100644
index 0000000..a8ba8ed
--- /dev/null
+++ b/tests/bugs/shard/configure-lru-limit.t
@@ -0,0 +1,48 @@
+#!/bin/bash
+
+. $(dirname $0)/../../include.rc
+. $(dirname $0)/../../volume.rc
+cleanup
+
+TEST glusterd
+TEST pidof glusterd
+TEST $CLI volume create $V0 replica 3 $H0:$B0/${V0}{0,1,2}
+TEST $CLI volume set $V0 features.shard on
+TEST $CLI volume set $V0 features.shard-block-size 4MB
+TEST $CLI volume set $V0 features.shard-lru-limit 25
+TEST $CLI volume start $V0
+
+TEST $GFS --volfile-id=$V0 --volfile-server=$H0 $M0
+
+# Perform a write that would cause 25 shards to be created, 24 of them under .shard
+TEST dd if=/dev/zero of=$M0/foo bs=1M count=100
+
+statedump=$(generate_mount_statedump $V0)
+sleep 1
+EXPECT "25" echo $(grep "lru-max-limit" $statedump | cut -f2 -d'=' | tail -1)
+
+# Base shard is never added to this list. So all other shards should make up for 24 inodes in lru list
+EXPECT "24" echo $(grep "inode-count" $statedump | cut -f2 -d'=' | tail -1)
+
+rm -f $statedump
+
+# Test to ensure there's no "reconfiguration" of the value once set.
+TEST $CLI volume set $V0 features.shard-lru-limit 30
+statedump=$(generate_mount_statedump $V0)
+sleep 1
+EXPECT "25" echo $(grep "lru-max-limit" $statedump | cut -f2 -d'=' | tail -1)
+rm -f $statedump
+
+EXPECT_WITHIN $UMOUNT_TIMEOUT "Y" force_umount $M0
+
+TEST $GFS --volfile-id=$V0 --volfile-server=$H0 $M0
+statedump=$(generate_mount_statedump $V0)
+sleep 1
+EXPECT "30" echo $(grep "lru-max-limit" $statedump | cut -f2 -d'=' | tail -1)
+rm -f $statedump
+
+EXPECT_WITHIN $UMOUNT_TIMEOUT "Y" force_umount $M0
+TEST $CLI volume stop $V0
+TEST $CLI volume delete $V0
+
+cleanup
diff --git a/xlators/features/shard/src/shard.c b/xlators/features/shard/src/shard.c
index 6066a54..eb32168 100644
--- a/xlators/features/shard/src/shard.c
+++ b/xlators/features/shard/src/shard.c
@@ -668,7 +668,7 @@ __shard_update_shards_inode_list (inode_t *linked_inode, xlator_t *this,
shard_inode_ctx_get (linked_inode, this, &ctx);
if (list_empty (&ctx->ilist)) {
- if (priv->inode_count + 1 <= SHARD_MAX_INODES) {
+ if (priv->inode_count + 1 <= priv->lru_limit) {
/* If this inode was linked here for the first time (indicated
* by empty list), and if there is still space in the priv list,
* add this ctx to the tail of the list.
@@ -6690,6 +6690,8 @@ init (xlator_t *this)
GF_OPTION_INIT ("shard-deletion-rate", priv->deletion_rate, uint32, out);
+ GF_OPTION_INIT ("shard-lru-limit", priv->lru_limit, uint64, out);
+
this->local_pool = mem_pool_new (shard_local_t, 128);
if (!this->local_pool) {
ret = -1;
@@ -6808,7 +6810,7 @@ shard_priv_dump (xlator_t *this)
gf_uint64_2human_readable (priv->block_size));
gf_proc_dump_write ("inode-count", "%d", priv->inode_count);
gf_proc_dump_write ("ilist_head", "%p", &priv->ilist_head);
- gf_proc_dump_write ("lru-max-limit", "%d", SHARD_MAX_INODES);
+ gf_proc_dump_write ("lru-max-limit", "%d", priv->lru_limit);
return 0;
}
@@ -6877,5 +6879,18 @@ struct volume_options options[] = {
.max = INT_MAX,
.description = "The number of shards to send deletes on at a time",
},
+ { .key = {"shard-lru-limit"},
+ .type = GF_OPTION_TYPE_INT,
+ .default_value = "16384",
+ .min = 20,
+ .max = INT_MAX,
+ .description = "The number of resolved shard inodes to keep in "
+ "memory. A higher number means shards that are "
+ "resolved will remain in memory longer, avoiding "
+ "frequent lookups on them when they participate in "
+ "file operations. The option also has a bearing on "
+ "amount of memory consumed by these inodes and their "
+ "internal metadata",
+ },
{ .key = {NULL} },
};
diff --git a/xlators/features/shard/src/shard.h b/xlators/features/shard/src/shard.h
index 5de098a..ac3813c 100644
--- a/xlators/features/shard/src/shard.h
+++ b/xlators/features/shard/src/shard.h
@@ -23,8 +23,6 @@
#define SHARD_MAX_BLOCK_SIZE (4 * GF_UNIT_TB)
#define SHARD_XATTR_PREFIX "trusted.glusterfs.shard."
#define GF_XATTR_SHARD_BLOCK_SIZE "trusted.glusterfs.shard.block-size"
-#define SHARD_INODE_LRU_LIMIT 4096
-#define SHARD_MAX_INODES 16384
/**
* Bit masks for the valid flag, which is used while updating ctx
**/
@@ -216,6 +214,7 @@ typedef struct shard_priv {
struct list_head ilist_head;
uint32_t deletion_rate;
shard_first_lookup_state_t first_lookup;
+ uint64_t lru_limit;
} shard_priv_t;
typedef struct {
diff --git a/xlators/mgmt/glusterd/src/glusterd-volume-set.c b/xlators/mgmt/glusterd/src/glusterd-volume-set.c
index a825f52..d442fe0 100644
--- a/xlators/mgmt/glusterd/src/glusterd-volume-set.c
+++ b/xlators/mgmt/glusterd/src/glusterd-volume-set.c
@@ -3298,6 +3298,12 @@ struct volopt_map_entry glusterd_volopt_map[] = {
.op_version = GD_OP_VERSION_3_7_0,
.flags = OPT_FLAG_CLIENT_OPT
},
+ { .key = "features.shard-lru-limit",
+ .voltype = "features/shard",
+ .op_version = GD_OP_VERSION_4_2_0,
+ .flags = OPT_FLAG_CLIENT_OPT,
+ .type = NO_DOC,
+ },
{ .key = "features.shard-deletion-rate",
.voltype = "features/shard",
.op_version = GD_OP_VERSION_3_13_4,
--
1.8.3.1

63
0442-glusterd-Reset-op-version-for-features.shard-lru-lim.patch Normal file
View File

@ -0,0 +1,63 @@
From dd1d565505d1f9c41dd6f151341f9337d89aa7cf Mon Sep 17 00:00:00 2001
From: Krutika Dhananjay <kdhananj@redhat.com>
Date: Tue, 6 Nov 2018 18:44:55 +0530
Subject: [PATCH 442/444] glusterd: Reset op-version for
"features.shard-lru-limit"
The op-version for the "features.shard-lru-limit" option was set to
4.2.0 in the upstream patch and backported at
41e7e33c6512e98a1567e5a5532d3898b59cfa98
This commit reverts the op-version for this option to 3.13.4.
Label: DOWNSTREAM ONLY
Change-Id: I7d3ed6b373851267c78fc6815a83bee2c0906413
BUG: 1603118
Signed-off-by: Krutika Dhananjay <kdhananj@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/155127
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Xavi Hernandez <xhernandez@redhat.com>
---
libglusterfs/src/globals.h | 4 +---
xlators/mgmt/glusterd/src/glusterd-volume-set.c | 2 +-
2 files changed, 2 insertions(+), 4 deletions(-)
diff --git a/libglusterfs/src/globals.h b/libglusterfs/src/globals.h
index 555f44b..1bede2e 100644
--- a/libglusterfs/src/globals.h
+++ b/libglusterfs/src/globals.h
@@ -43,7 +43,7 @@
*/
#define GD_OP_VERSION_MIN 1 /* MIN is the fresh start op-version, mostly
should not change */
-#define GD_OP_VERSION_MAX GD_OP_VERSION_4_2_0 /* MAX VERSION is the maximum
+#define GD_OP_VERSION_MAX GD_OP_VERSION_3_13_4 /* MAX VERSION is the maximum
count in VME table, should
keep changing with
introduction of newer
@@ -111,8 +111,6 @@
#define GD_OP_VERSION_3_13_3 31303 /* Op-version for GlusterFS 3.13.3 */
-#define GD_OP_VERSION_4_2_0 40200 /* Op-version for GlusterFS 4.2.0 */
-
/* Downstream only change */
#define GD_OP_VERSION_3_11_2 31102 /* Op-version for RHGS 3.3.1-async */
#define GD_OP_VERSION_3_13_3 31303 /* Op-version for RHGS-3.4-Batch Update-1*/
diff --git a/xlators/mgmt/glusterd/src/glusterd-volume-set.c b/xlators/mgmt/glusterd/src/glusterd-volume-set.c
index d442fe0..1175f1d 100644
--- a/xlators/mgmt/glusterd/src/glusterd-volume-set.c
+++ b/xlators/mgmt/glusterd/src/glusterd-volume-set.c
@@ -3300,7 +3300,7 @@ struct volopt_map_entry glusterd_volopt_map[] = {
},
{ .key = "features.shard-lru-limit",
.voltype = "features/shard",
- .op_version = GD_OP_VERSION_4_2_0,
+ .op_version = GD_OP_VERSION_3_13_4,
.flags = OPT_FLAG_CLIENT_OPT,
.type = NO_DOC,
},
--
1.8.3.1

367
0443-features-shard-Hold-a-ref-on-base-inode-when-adding-.patch Normal file
View File

@ -0,0 +1,367 @@
From 72ce80749fca03ab97a63af79d4e6bc76a49ab64 Mon Sep 17 00:00:00 2001
From: Krutika Dhananjay <kdhananj@redhat.com>
Date: Fri, 5 Oct 2018 11:32:21 +0530
Subject: [PATCH 443/444] features/shard: Hold a ref on base inode when adding
a shard to lru list
> Upstream: https://review.gluster.org/21454
> BUG: 1605056
> Change-Id: Ic15ca41444dd04684a9458bd4a526b1d3e160499
In __shard_update_shards_inode_list(), previously shard translator
was not holding a ref on the base inode whenever a shard was added to
the lru list. But if the base shard is forgotten and destroyed either
by fuse due to memory pressure or due to the file being deleted at some
point by a different client with this client still containing stale
shards in its lru list, the client would crash at the time of locking
lru_base_inode->lock owing to illegal memory access.
So now the base shard is ref'd into the inode ctx of every shard that
is added to lru list until it gets lru'd out.
The patch also handles the case where none of the shards associated
with a file that is about to be deleted are part of the LRU list and
where an unlink at the beginning of the operation destroys the base
inode (because there are no refkeepers) and hence all of the shards
that are about to be deleted will be resolved without the existence
of a base shard in-memory. This, if not handled properly, could lead
to a crash.
Change-Id: Ic15ca41444dd04684a9458bd4a526b1d3e160499
BUG: 1603118
Signed-off-by: Krutika Dhananjay <kdhananj@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/155318
Reviewed-by: Xavi Hernandez <xhernandez@redhat.com>
Tested-by: RHGS Build Bot <nigelb@redhat.com>
---
tests/bugs/shard/bug-1605056-2.t | 34 +++++++++++++++
tests/bugs/shard/bug-1605056.t | 63 ++++++++++++++++++++++++++++
tests/bugs/shard/shard-inode-refcount-test.t | 2 +-
tests/volume.rc | 12 ++++--
xlators/features/shard/src/shard.c | 48 +++++++++++++++------
5 files changed, 141 insertions(+), 18 deletions(-)
create mode 100644 tests/bugs/shard/bug-1605056-2.t
create mode 100644 tests/bugs/shard/bug-1605056.t
diff --git a/tests/bugs/shard/bug-1605056-2.t b/tests/bugs/shard/bug-1605056-2.t
new file mode 100644
index 0000000..a9c10fe
--- /dev/null
+++ b/tests/bugs/shard/bug-1605056-2.t
@@ -0,0 +1,34 @@
+#!/bin/bash
+
+. $(dirname $0)/../../include.rc
+. $(dirname $0)/../../volume.rc
+
+cleanup
+
+TEST glusterd
+TEST pidof glusterd
+TEST $CLI volume create $V0 replica 3 $H0:$B0/${V0}{0,1,2}
+TEST $CLI volume set $V0 features.shard on
+TEST $CLI volume set $V0 features.shard-block-size 4MB
+TEST $CLI volume set $V0 features.shard-lru-limit 25
+TEST $CLI volume set $V0 performance.write-behind off
+
+TEST $CLI volume start $V0
+
+TEST $GFS --volfile-id=$V0 --volfile-server=$H0 $M0
+
+# Perform a write that would cause 25 shards to be created under .shard
+TEST dd if=/dev/zero of=$M0/foo bs=1M count=104
+
+# Write into another file bar to ensure all of foo's shards are evicted from lru list of $M0
+TEST dd if=/dev/zero of=$M0/bar bs=1M count=104
+
+# Delete foo from $M0. If there's a bug, the mount will crash.
+TEST unlink $M0/foo
+
+EXPECT_WITHIN $UMOUNT_TIMEOUT "Y" force_umount $M0
+
+TEST $CLI volume stop $V0
+TEST $CLI volume delete $V0
+
+cleanup
diff --git a/tests/bugs/shard/bug-1605056.t b/tests/bugs/shard/bug-1605056.t
new file mode 100644
index 0000000..c2329ea
--- /dev/null
+++ b/tests/bugs/shard/bug-1605056.t
@@ -0,0 +1,63 @@
+#!/bin/bash
+
+. $(dirname $0)/../../include.rc
+. $(dirname $0)/../../volume.rc
+
+SHARD_COUNT_TIME=5
+
+cleanup
+
+TEST glusterd
+TEST pidof glusterd
+TEST $CLI volume create $V0 replica 3 $H0:$B0/${V0}{0,1,2}
+TEST $CLI volume set $V0 features.shard on
+TEST $CLI volume set $V0 features.shard-block-size 4MB
+TEST $CLI volume set $V0 features.shard-lru-limit 25
+TEST $CLI volume set $V0 performance.write-behind off
+
+TEST $CLI volume start $V0
+
+TEST $GFS --volfile-id=$V0 --volfile-server=$H0 $M0
+TEST $GFS --volfile-id=$V0 --volfile-server=$H0 $M1
+
+# Perform a write that would cause 25 shards to be created under .shard
+TEST dd if=/dev/zero of=$M0/foo bs=1M count=104
+
+# Read the file from $M1, indirectly filling up the lru list.
+TEST `cat $M1/foo > /dev/null`
+statedump=$(generate_mount_statedump $V0 $M1)
+sleep 1
+EXPECT "25" echo $(grep "inode-count" $statedump | cut -f2 -d'=' | tail -1)
+rm -f $statedump
+
+# Delete foo from $M0.
+TEST unlink $M0/foo
+
+# Send stat on foo from $M1 to force $M1 to "forget" inode associated with foo.
+# Now the ghost shards associated with "foo" are still in lru list of $M1.
+TEST ! stat $M1/foo
+
+# Let's force the ghost shards of "foo" out of lru list by looking up more shards
+# through I/O on a file named "bar" from $M1. This should crash if the base inode
+# had been destroyed by now.
+
+TEST dd if=/dev/zero of=$M1/bar bs=1M count=104
+
+###############################################
+#### Now for some inode ref-leak tests ... ####
+###############################################
+
+# Expect there to be 29 active inodes - 26 belonging to "bar", 1 for .shard,
+# 1 for .shard/remove_me and 1 for '/'
+EXPECT_WITHIN $SHARD_COUNT_TIME `expr 26 + 3` get_mount_active_size_value $V0 $M1
+
+TEST rm -f $M1/bar
+EXPECT_WITHIN $SHARD_COUNT_TIME 3 get_mount_active_size_value $V0 $M1
+
+EXPECT_WITHIN $UMOUNT_TIMEOUT "Y" force_umount $M0
+EXPECT_WITHIN $UMOUNT_TIMEOUT "Y" force_umount $M1
+
+TEST $CLI volume stop $V0
+TEST $CLI volume delete $V0
+
+cleanup
diff --git a/tests/bugs/shard/shard-inode-refcount-test.t b/tests/bugs/shard/shard-inode-refcount-test.t
index 087c8ba..3fd181b 100644
--- a/tests/bugs/shard/shard-inode-refcount-test.t
+++ b/tests/bugs/shard/shard-inode-refcount-test.t
@@ -21,7 +21,7 @@ TEST dd if=/dev/zero conv=fsync of=$M0/one-plus-five-shards bs=1M count=23
ACTIVE_INODES_BEFORE=$(get_mount_active_size_value $V0)
TEST rm -f $M0/one-plus-five-shards
# Expect 5 inodes less. But one inode more than before because .remove_me would be created.
-EXPECT_WITHIN $SHARD_COUNT_TIME `expr $ACTIVE_INODES_BEFORE - 5 + 1` get_mount_active_size_value $V0
+EXPECT_WITHIN $SHARD_COUNT_TIME `expr $ACTIVE_INODES_BEFORE - 5 + 1` get_mount_active_size_value $V0 $M0
EXPECT_WITHIN $UMOUNT_TIMEOUT "Y" force_umount $M0
TEST $CLI volume stop $V0
diff --git a/tests/volume.rc b/tests/volume.rc
index bba7e4e..6a983fd 100644
--- a/tests/volume.rc
+++ b/tests/volume.rc
@@ -93,7 +93,8 @@ function remove_brick_status_completed_field {
function get_mount_process_pid {
local vol=$1
- ps auxww | grep glusterfs | grep -E "volfile-id[ =]/?$vol " | awk '{print $2}' | head -1
+ local mnt=$2
+ ps auxww | grep glusterfs | grep -E "volfile-id[ =]/?$vol .*$mnt" | awk '{print $2}' | head -1
}
function get_nfs_pid ()
@@ -126,7 +127,8 @@ function generate_statedump {
function generate_mount_statedump {
local vol=$1
- generate_statedump $(get_mount_process_pid $vol)
+ local mnt=$2
+ generate_statedump $(get_mount_process_pid $vol $mnt)
}
function cleanup_mount_statedump {
@@ -850,7 +852,8 @@ function get_active_fd_count {
function get_mount_active_size_value {
local vol=$1
- local statedump=$(generate_mount_statedump $vol)
+ local mount=$2
+ local statedump=$(generate_mount_statedump $vol $mount)
sleep 1
local val=$(grep "active_size" $statedump | cut -f2 -d'=' | tail -1)
rm -f $statedump
@@ -859,7 +862,8 @@ function get_mount_active_size_value {
function get_mount_lru_size_value {
local vol=$1
- local statedump=$(generate_mount_statedump $vol)
+ local mount=$2
+ local statedump=$(generate_mount_statedump $vol $mount)
sleep 1
local val=$(grep "lru_size" $statedump | cut -f2 -d'=' | tail -1)
rm -f $statedump
diff --git a/xlators/features/shard/src/shard.c b/xlators/features/shard/src/shard.c
index eb32168..fb88315 100644
--- a/xlators/features/shard/src/shard.c
+++ b/xlators/features/shard/src/shard.c
@@ -651,7 +651,8 @@ out:
inode_t *
__shard_update_shards_inode_list (inode_t *linked_inode, xlator_t *this,
- inode_t *base_inode, int block_num)
+ inode_t *base_inode, int block_num,
+ uuid_t gfid)
{
char block_bname[256] = {0,};
inode_t *lru_inode = NULL;
@@ -679,10 +680,13 @@ __shard_update_shards_inode_list (inode_t *linked_inode, xlator_t *this,
inode_ref (linked_inode);
if (base_inode)
gf_uuid_copy (ctx->base_gfid, base_inode->gfid);
+ else
+ gf_uuid_copy(ctx->base_gfid, gfid);
ctx->block_num = block_num;
list_add_tail (&ctx->ilist, &priv->ilist_head);
priv->inode_count++;
- ctx->base_inode = base_inode;
+ if (base_inode)
+ ctx->base_inode = inode_ref (base_inode);
} else {
/*If on the other hand there is no available slot for this inode
* in the list, delete the lru inode from the head of the list,
@@ -701,6 +705,8 @@ __shard_update_shards_inode_list (inode_t *linked_inode, xlator_t *this,
* deleted from fsync list and fsync'd in a new frame,
* and then unlinked in memory and forgotten.
*/
+ if (!lru_base_inode)
+ goto after_fsync_check;
LOCK (&lru_base_inode->lock);
LOCK (&lru_inode->lock);
{
@@ -715,6 +721,7 @@ __shard_update_shards_inode_list (inode_t *linked_inode, xlator_t *this,
UNLOCK (&lru_inode->lock);
UNLOCK (&lru_base_inode->lock);
+after_fsync_check:
if (!do_fsync) {
shard_make_block_bname (lru_inode_ctx->block_num,
lru_inode_ctx->base_gfid,
@@ -729,20 +736,31 @@ __shard_update_shards_inode_list (inode_t *linked_inode, xlator_t *this,
inode_forget (lru_inode, 0);
} else {
fsync_inode = lru_inode;
- inode_unref (lru_base_inode);
+ if (lru_base_inode)
+ inode_unref (lru_base_inode);
}
/* The following unref corresponds to the ref
* held by inode_find() above.
*/
inode_unref (lru_inode);
+
+ /* The following unref corresponds to the ref held on
+ * the base shard at the time of adding shard inode to
+ * lru list
+ */
+ if (lru_base_inode)
+ inode_unref (lru_base_inode);
/* For as long as an inode is in lru list, we try to
* keep it alive by holding a ref on it.
*/
inode_ref (linked_inode);
if (base_inode)
gf_uuid_copy (ctx->base_gfid, base_inode->gfid);
+ else
+ gf_uuid_copy (ctx->base_gfid, gfid);
ctx->block_num = block_num;
- ctx->base_inode = base_inode;
+ if (base_inode)
+ ctx->base_inode = inode_ref (base_inode);
list_add_tail (&ctx->ilist, &priv->ilist_head);
}
} else {
@@ -1027,7 +1045,7 @@ shard_common_resolve_shards (call_frame_t *frame, xlator_t *this,
fsync_inode = __shard_update_shards_inode_list (inode,
this,
res_inode,
- shard_idx_iter);
+ shard_idx_iter, gfid);
}
UNLOCK(&priv->lock);
shard_idx_iter++;
@@ -2173,7 +2191,8 @@ shard_link_block_inode (shard_local_t *local, int block_num, inode_t *inode,
fsync_inode = __shard_update_shards_inode_list (linked_inode,
this,
local->loc.inode,
- block_num);
+ block_num,
+ gfid);
}
UNLOCK(&priv->lock);
if (fsync_inode)
@@ -2881,6 +2900,7 @@ void
shard_unlink_block_inode (shard_local_t *local, int shard_block_num)
{
char block_bname[256] = {0,};
+ int unref_base_inode = 0;
uuid_t gfid = {0,};
inode_t *inode = NULL;
inode_t *base_inode = NULL;
@@ -2894,11 +2914,12 @@ shard_unlink_block_inode (shard_local_t *local, int shard_block_num)
priv = this->private;
inode = local->inode_list[shard_block_num - local->first_block];
- base_inode = local->resolver_base_inode;
+ shard_inode_ctx_get (inode, this, &ctx);
+ base_inode = ctx->base_inode;
if (base_inode)
gf_uuid_copy (gfid, base_inode->gfid);
else
- gf_uuid_copy (gfid, local->base_gfid);
+ gf_uuid_copy (gfid, ctx->base_gfid);
shard_make_block_bname (shard_block_num, gfid,
block_bname, sizeof (block_bname));
@@ -2912,17 +2933,16 @@ shard_unlink_block_inode (shard_local_t *local, int shard_block_num)
if (!list_empty (&ctx->ilist)) {
list_del_init (&ctx->ilist);
priv->inode_count--;
+ unref_base_inode++;
GF_ASSERT (priv->inode_count >= 0);
unlink_unref_forget = _gf_true;
}
if (ctx->fsync_needed) {
- if (base_inode)
- inode_unref (base_inode);
+ unref_base_inode++;
list_del_init (&ctx->to_fsync_list);
- if (base_inode) {
+ if (base_inode)
__shard_inode_ctx_get (base_inode, this, &base_ictx);
- base_ictx->fsync_count--;
- }
+ base_ictx->fsync_count--;
}
}
UNLOCK(&inode->lock);
@@ -2933,6 +2953,8 @@ shard_unlink_block_inode (shard_local_t *local, int shard_block_num)
inode_unref (inode);
inode_forget (inode, 0);
}
+ if (base_inode && unref_base_inode)
+ inode_ref_reduce_by_n (base_inode, unref_base_inode);
UNLOCK(&priv->lock);
}
--
1.8.3.1

31
0444-features-shard-fix-formatting-warning.patch Normal file
View File

@ -0,0 +1,31 @@
From 6dbeac0371f3f2b42c0b428ba9f95b4ac3bc889d Mon Sep 17 00:00:00 2001
From: Xavi Hernandez <xhernandez@redhat.com>
Date: Thu, 8 Nov 2018 18:42:26 +0100
Subject: [PATCH 444/444] features/shard: fix formatting warning
BUG: 1603118
Change-Id: I6191351f824901a45416ffe7610ad2b964645012
Signed-off-by: Xavi Hernandez <xhernandez@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/155395
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Sunil Kumar Heggodu Gopala Acharya <sheggodu@redhat.com>
---
xlators/features/shard/src/shard.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/xlators/features/shard/src/shard.c b/xlators/features/shard/src/shard.c
index fb88315..5b72399 100644
--- a/xlators/features/shard/src/shard.c
+++ b/xlators/features/shard/src/shard.c
@@ -6832,7 +6832,7 @@ shard_priv_dump (xlator_t *this)
gf_uint64_2human_readable (priv->block_size));
gf_proc_dump_write ("inode-count", "%d", priv->inode_count);
gf_proc_dump_write ("ilist_head", "%p", &priv->ilist_head);
- gf_proc_dump_write ("lru-max-limit", "%d", priv->lru_limit);
+ gf_proc_dump_write ("lru-max-limit", "%" PRIu64, priv->lru_limit);
return 0;
}
--
1.8.3.1

45
glusterfs.spec
View File

@ -192,7 +192,7 @@ Release: 0.1%{?prereltag:.%{prereltag}}%{?dist}
%else %else
Name: glusterfs Name: glusterfs
Version: 3.12.2 Version: 3.12.2
Release: 25%{?dist} Release: 26%{?dist}
%endif %endif
License: GPLv2 or LGPLv3+ License: GPLv2 or LGPLv3+
Group: System Environment/Base Group: System Environment/Base
@ -672,6 +672,43 @@ Patch0404: 0404-core-Resolve-some-warnings-to-release-a-build.patch
Patch0405: 0405-glusterfsd-add-missing-UNLOCK.patch Patch0405: 0405-glusterfsd-add-missing-UNLOCK.patch
Patch0406: 0406-glusterd-improve-logging-for-stage_deleted-flag.patch Patch0406: 0406-glusterd-improve-logging-for-stage_deleted-flag.patch
Patch0407: 0407-spec-update-RHGS-version-for-RHGSWA.patch Patch0407: 0407-spec-update-RHGS-version-for-RHGSWA.patch
# Patch0408: 0408-Update-rfc.sh-to-rhgs-3.4.2.patch
Patch0409: 0409-Update-database-profile-group.patch
Patch0410: 0410-cli-fix-glusterd-memory-leak-cause-by-gluster-v-stat.patch
Patch0411: 0411-glusterd-ensure-volinfo-caps-is-set-to-correct-value.patch
Patch0412: 0412-glusterd-set-fsid-while-performing-replace-brick.patch
Patch0413: 0413-glusterfind-add-logs-to-identify-parsing-phases.patch
Patch0414: 0414-logrotate-utilize-the-new-maxsize-option.patch
Patch0415: 0415-statedump-fix-clang-null-dereference-error.patch
Patch0416: 0416-glusterd-ignore-RPC-events-when-glusterd-is-shutting.patch
Patch0417: 0417-cli-Add-warning-message-while-converting-to-replica-.patch
Patch0418: 0418-cli-correct-rebalance-status-elapsed-check.patch
Patch0419: 0419-glusterfs-During-reconfigure-set-log-level-per-xlato.patch
Patch0420: 0420-Modify-log-message-DH-ciphers-are-disabled-from-ERRO.patch
Patch0421: 0421-rpc-handle-EAGAIN-when-SSL_ERROR_SYSCALL-is-returned.patch
Patch0422: 0422-glusterd-raise-default-transport.listen-backlog.patch
Patch0423: 0423-glusterd-acquire-lock-to-update-volinfo-structure.patch
Patch0424: 0424-cluster-afr-Delegate-metadata-heal-with-pending-xatt.patch
Patch0425: 0425-cluster-afr-Delegate-name-heal-when-possible.patch
Patch0426: 0426-features-shard-Make-operations-on-internal-directori.patch
Patch0427: 0427-features-shard-Add-option-to-barrier-parallel-lookup.patch
Patch0428: 0428-libglusterfs-syncop-Handle-barrier_-init-destroy-in-.patch
Patch0429: 0429-features-shard-Introducing-.shard-.remove_me-for-ato.patch
Patch0430: 0430-features-shard-Perform-shards-deletion-in-the-backgr.patch
Patch0431: 0431-glusterd-Reset-op-version-for-features.shard-deletio.patch
Patch0432: 0432-features-shard-Fix-crash-and-test-case-in-RENAME-fop.patch
Patch0433: 0433-mgmt-glusterd-use-proper-path-to-the-volfile.patch
Patch0434: 0434-cluster-afr-s-uuid_is_null-gf_uuid_is_null.patch
Patch0435: 0435-geo-rep-Fix-traceback-with-symlink-metadata-sync.patch
Patch0436: 0436-geo-rep-Fix-issue-in-gfid-conflict-resolution.patch
Patch0437: 0437-geo-rep-Add-more-intelligence-to-automatic-error-han.patch
Patch0438: 0438-cluster-dht-In-rename-unlink-after-creating-linkto-f.patch
Patch0439: 0439-cluster-dht-fixes-to-unlinking-invalid-linkto-file.patch
Patch0440: 0440-features-locks-Use-pthread_mutex_unlock-instead-of-p.patch
Patch0441: 0441-features-shard-Make-lru-limit-of-inode-list-configur.patch
Patch0442: 0442-glusterd-Reset-op-version-for-features.shard-lru-lim.patch
Patch0443: 0443-features-shard-Hold-a-ref-on-base-inode-when-adding-.patch
Patch0444: 0444-features-shard-fix-formatting-warning.patch
%description %description
GlusterFS is a distributed file-system capable of scaling to several GlusterFS is a distributed file-system capable of scaling to several
@ -2620,6 +2657,12 @@ fi
%endif %endif
%changelog %changelog
* Fri Nov 09 2018 Milind Changire <mchangir@redhat.com> - 3.12.2-26
- fixes bugs bz#1479446 bz#1520882 bz#1579758 bz#1598407 bz#1599808
bz#1603118 bz#1619357 bz#1622001 bz#1622308 bz#1631166 bz#1631418 bz#1632563
bz#1634649 bz#1635071 bz#1635100 bz#1635136 bz#1636291 bz#1638069 bz#1640347
bz#1642854 bz#1643035 bz#1644120 bz#1644279 bz#1645916 bz#1647675
* Thu Oct 25 2018 Milind Changire <mchangir@redhat.com> - 3.12.2-25 * Thu Oct 25 2018 Milind Changire <mchangir@redhat.com> - 3.12.2-25
- fixes bugs bz#1641586 - fixes bugs bz#1641586