rpms/sbd
7
0
Fork 0
Code Issues Pull Requests Releases Activity

import sbd-1.4.0-10.el8

Browse Source
This commit is contained in:
CentOS Sources 2019-08-02 12:17:32 -04:00 committed by Stepan Oksanichenko
commit 2b0db1e2ae
18 changed files with 1640 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
.gitignore vendored Normal file
View File

@ -0,0 +1 @@
SOURCES/sbd-7f33d1a409d0a4e2cd69946688c48eaa8f3c5d26.tar.gz

1
.sbd.metadata Normal file
View File

@ -0,0 +1 @@
7bf4ad26875cafa743ca96aec9dae1739bd6281b SOURCES/sbd-7f33d1a409d0a4e2cd69946688c48eaa8f3c5d26.tar.gz

79
SOURCES/0001-Fix-sbd-cluster-finalize-cmap-connection-if-disconne.patch Normal file
View File

@ -0,0 +1,79 @@
From f8f980340256ab5bef5385cd3bc082fdfb7613ed Mon Sep 17 00:00:00 2001
From: "Gao,Yan" <ygao@suse.com>
Date: Thu, 7 Mar 2019 15:01:26 +0100
Subject: [PATCH] Fix: sbd-cluster: finalize cmap connection if disconnected
from cluster
Previously if sbd cluster servant anyhow got dis-/reconnected from the
cluster, it'd start hogging CPU keeping polling the main loop source
from the old cmap connection.
---
src/sbd-cluster.c | 38 ++++++++++++++++++++++++++------------
1 file changed, 26 insertions(+), 12 deletions(-)
diff --git a/src/sbd-cluster.c b/src/sbd-cluster.c
index 51bb456..541212f 100644
--- a/src/sbd-cluster.c
+++ b/src/sbd-cluster.c
@@ -174,6 +174,25 @@ cmap_dispatch_callback (gpointer user_data)
return TRUE;
}
+static void
+cmap_destroy(void)
+{
+ if (cmap_source) {
+ g_source_destroy(cmap_source);
+ cmap_source = NULL;
+ }
+
+ if (track_handle) {
+ cmap_track_delete(cmap_handle, track_handle);
+ track_handle = 0;
+ }
+
+ if (cmap_handle) {
+ cmap_finalize(cmap_handle);
+ cmap_handle = 0;
+ }
+}
+
static gboolean
sbd_get_two_node(void)
{
@@ -217,18 +236,7 @@ sbd_get_two_node(void)
return TRUE;
out:
- if (cmap_source) {
- g_source_destroy(cmap_source);
- cmap_source = NULL;
- }
- if (track_handle) {
- cmap_track_delete(cmap_handle, track_handle);
- track_handle = 0;
- }
- if (cmap_handle) {
- cmap_finalize(cmap_handle);
- cmap_handle = 0;
- }
+ cmap_destroy();
return FALSE;
}
@@ -327,6 +335,12 @@ sbd_membership_destroy(gpointer user_data)
{
cl_log(LOG_WARNING, "Lost connection to %s", name_for_cluster_type(get_cluster_type()));
+ if (get_cluster_type() != pcmk_cluster_unknown) {
+#if SUPPORT_COROSYNC && CHECK_TWO_NODE
+ cmap_destroy();
+#endif
+ }
+
set_servant_health(pcmk_health_unclean, LOG_ERR, "Cluster connection terminated");
notify_parent();
--
1.8.3.1

302
SOURCES/0002-Fix-sbd-pacemaker-make-handling-of-cib-connection-lo.patch Normal file
View File

@ -0,0 +1,302 @@
From a716a8ddd3df615009bcff3bd96dd9ae64cb5f68 Mon Sep 17 00:00:00 2001
From: Klaus Wenninger <klaus.wenninger@aon.at>
Date: Tue, 19 Mar 2019 21:36:15 +0100
Subject: [PATCH] Fix: sbd-pacemaker: make handling of cib-connection loss more
robust
Exit pcmk-servant on graceful pacemaker shutdown and go back
to state before pacemaker was detected initially.
Purge all cib-traces otherwise and try to reconnect within timeout.
---
src/sbd-inquisitor.c | 24 ++++++++++++++++++++----
src/sbd-md.c | 30 +++++++++++++++---------------
src/sbd-pacemaker.c | 38 +++++++++++++++++++++++++++++---------
src/sbd.h | 11 +++++++----
4 files changed, 71 insertions(+), 32 deletions(-)
diff --git a/src/sbd-inquisitor.c b/src/sbd-inquisitor.c
index 9be6c99..77c6e4f 100644
--- a/src/sbd-inquisitor.c
+++ b/src/sbd-inquisitor.c
@@ -490,19 +490,19 @@ void inquisitor_child(void)
if (sbd_is_disk(s)) {
if (WIFEXITED(status)) {
switch(WEXITSTATUS(status)) {
- case EXIT_MD_IO_FAIL:
+ case EXIT_MD_SERVANT_IO_FAIL:
DBGLOG(LOG_INFO, "Servant for %s requests to be disowned",
s->devname);
break;
- case EXIT_MD_REQUEST_RESET:
+ case EXIT_MD_SERVANT_REQUEST_RESET:
cl_log(LOG_WARNING, "%s requested a reset", s->devname);
do_reset();
break;
- case EXIT_MD_REQUEST_SHUTOFF:
+ case EXIT_MD_SERVANT_REQUEST_SHUTOFF:
cl_log(LOG_WARNING, "%s requested a shutoff", s->devname);
do_off();
break;
- case EXIT_MD_REQUEST_CRASHDUMP:
+ case EXIT_MD_SERVANT_REQUEST_CRASHDUMP:
cl_log(LOG_WARNING, "%s requested a crashdump", s->devname);
do_crashdump();
break;
@@ -510,6 +510,22 @@ void inquisitor_child(void)
break;
}
}
+ } else if (sbd_is_pcmk(s)) {
+ if (WIFEXITED(status)) {
+ switch(WEXITSTATUS(status)) {
+ case EXIT_PCMK_SERVANT_GRACEFUL_SHUTDOWN:
+ DBGLOG(LOG_INFO, "PCMK-Servant has exited gracefully");
+ /* revert to state prior to pacemaker-detection */
+ s->restarts = 0;
+ s->restart_blocked = 0;
+ cluster_appeared = 0;
+ s->outdated = 1;
+ s->t_last.tv_sec = 0;
+ break;
+ default:
+ break;
+ }
+ }
}
cleanup_servant_by_pid(pid);
}
diff --git a/src/sbd-md.c b/src/sbd-md.c
index ba2c34d..c51d381 100644
--- a/src/sbd-md.c
+++ b/src/sbd-md.c
@@ -1061,19 +1061,19 @@ int servant_md(const char *diskname, int mode, const void* argp)
st = open_device(diskname, LOG_WARNING);
if (!st) {
- exit(EXIT_MD_IO_FAIL);
+ exit(EXIT_MD_SERVANT_IO_FAIL);
}
s_header = header_get(st);
if (!s_header) {
cl_log(LOG_ERR, "Not a valid header on %s", diskname);
- exit(EXIT_MD_IO_FAIL);
+ exit(EXIT_MD_SERVANT_IO_FAIL);
}
if (servant_check_timeout_inconsistent(s_header) < 0) {
cl_log(LOG_ERR, "Timeouts on %s do not match first device",
diskname);
- exit(EXIT_MD_IO_FAIL);
+ exit(EXIT_MD_SERVANT_IO_FAIL);
}
if (s_header->minor_version > 0) {
@@ -1086,14 +1086,14 @@ int servant_md(const char *diskname, int mode, const void* argp)
cl_log(LOG_ERR,
"No slot allocated, and automatic allocation failed for disk %s.",
diskname);
- rc = EXIT_MD_IO_FAIL;
+ rc = EXIT_MD_SERVANT_IO_FAIL;
goto out;
}
s_node = sector_alloc();
if (slot_read(st, mbox, s_node) < 0) {
cl_log(LOG_ERR, "Unable to read node entry on %s",
diskname);
- exit(EXIT_MD_IO_FAIL);
+ exit(EXIT_MD_SERVANT_IO_FAIL);
}
cl_log(LOG_NOTICE, "Monitoring slot %d on disk %s", mbox, diskname);
@@ -1109,7 +1109,7 @@ int servant_md(const char *diskname, int mode, const void* argp)
if (mode > 0) {
if (mbox_read(st, mbox, s_mbox) < 0) {
cl_log(LOG_ERR, "mbox read failed during start-up in servant.");
- rc = EXIT_MD_IO_FAIL;
+ rc = EXIT_MD_SERVANT_IO_FAIL;
goto out;
}
if (s_mbox->cmd != SBD_MSG_EXIT &&
@@ -1125,7 +1125,7 @@ int servant_md(const char *diskname, int mode, const void* argp)
DBGLOG(LOG_INFO, "First servant start - zeroing inbox");
memset(s_mbox, 0, sizeof(*s_mbox));
if (mbox_write(st, mbox, s_mbox) < 0) {
- rc = EXIT_MD_IO_FAIL;
+ rc = EXIT_MD_SERVANT_IO_FAIL;
goto out;
}
}
@@ -1154,28 +1154,28 @@ int servant_md(const char *diskname, int mode, const void* argp)
s_header_retry = header_get(st);
if (!s_header_retry) {
cl_log(LOG_ERR, "No longer found a valid header on %s", diskname);
- exit(EXIT_MD_IO_FAIL);
+ exit(EXIT_MD_SERVANT_IO_FAIL);
}
if (memcmp(s_header, s_header_retry, sizeof(*s_header)) != 0) {
cl_log(LOG_ERR, "Header on %s changed since start-up!", diskname);
- exit(EXIT_MD_IO_FAIL);
+ exit(EXIT_MD_SERVANT_IO_FAIL);
}
free(s_header_retry);
s_node_retry = sector_alloc();
if (slot_read(st, mbox, s_node_retry) < 0) {
cl_log(LOG_ERR, "slot read failed in servant.");
- exit(EXIT_MD_IO_FAIL);
+ exit(EXIT_MD_SERVANT_IO_FAIL);
}
if (memcmp(s_node, s_node_retry, sizeof(*s_node)) != 0) {
cl_log(LOG_ERR, "Node entry on %s changed since start-up!", diskname);
- exit(EXIT_MD_IO_FAIL);
+ exit(EXIT_MD_SERVANT_IO_FAIL);
}
free(s_node_retry);
if (mbox_read(st, mbox, s_mbox) < 0) {
cl_log(LOG_ERR, "mbox read failed in servant.");
- exit(EXIT_MD_IO_FAIL);
+ exit(EXIT_MD_SERVANT_IO_FAIL);
}
if (s_mbox->cmd > 0) {
@@ -1190,14 +1190,14 @@ int servant_md(const char *diskname, int mode, const void* argp)
sigqueue(ppid, SIG_TEST, signal_value);
break;
case SBD_MSG_RESET:
- exit(EXIT_MD_REQUEST_RESET);
+ exit(EXIT_MD_SERVANT_REQUEST_RESET);
case SBD_MSG_OFF:
- exit(EXIT_MD_REQUEST_SHUTOFF);
+ exit(EXIT_MD_SERVANT_REQUEST_SHUTOFF);
case SBD_MSG_EXIT:
sigqueue(ppid, SIG_EXITREQ, signal_value);
break;
case SBD_MSG_CRASHDUMP:
- exit(EXIT_MD_REQUEST_CRASHDUMP);
+ exit(EXIT_MD_SERVANT_REQUEST_CRASHDUMP);
default:
/* FIXME:
An "unknown" message might result
diff --git a/src/sbd-pacemaker.c b/src/sbd-pacemaker.c
index aac355a..c69fc55 100644
--- a/src/sbd-pacemaker.c
+++ b/src/sbd-pacemaker.c
@@ -103,6 +103,9 @@ static pe_working_set_t *data_set = NULL;
static long last_refresh = 0;
+static int pcmk_clean_shutdown = 0;
+static int pcmk_shutdown = 0;
+
static gboolean
mon_timer_reconnect(gpointer data)
{
@@ -128,10 +131,26 @@ mon_cib_connection_destroy(gpointer user_data)
{
if (cib) {
cib->cmds->signoff(cib);
+ /* retrigger as last one might have been skipped */
+ mon_refresh_state(NULL);
+ if (pcmk_clean_shutdown) {
+ /* assume a graceful pacemaker-shutdown */
+ clean_up(EXIT_PCMK_SERVANT_GRACEFUL_SHUTDOWN);
+ }
+ /* getting here we aren't sure about the pacemaker-state
+ so try to use the timeout to reconnect and get
+ everything sorted out again
+ */
+ pcmk_shutdown = 0;
set_servant_health(pcmk_health_transient, LOG_WARNING, "Disconnected from CIB");
timer_id_reconnect = g_timeout_add(reconnect_msec, mon_timer_reconnect, NULL);
}
cib_connected = 0;
+ /* no sense in looking into outdated cib, trying to apply patch, ... */
+ if (current_cib) {
+ free_xml(current_cib);
+ current_cib = NULL;
+ }
return;
}
@@ -171,7 +190,7 @@ static gboolean
mon_timer_notify(gpointer data)
{
static int counter = 0;
- int counter_max = timeout_watchdog / timeout_loop;
+ int counter_max = timeout_watchdog / timeout_loop / 2;
if (timer_id_notify > 0) {
g_source_remove(timer_id_notify);
@@ -280,11 +299,6 @@ compute_status(pe_working_set_t * data_set)
} else if (node->details->pending) {
set_servant_health(pcmk_health_pending, LOG_WARNING, "Node state: pending");
-#if 0
- } else if (node->details->shutdown) {
- set_servant_health(pcmk_health_shutdown, LOG_WARNING, "Node state: shutting down");
-#endif
-
} else if (data_set->flags & pe_flag_have_quorum) {
set_servant_health(pcmk_health_online, LOG_INFO, "Node state: online");
ever_had_quorum = TRUE;
@@ -315,6 +329,12 @@ compute_status(pe_working_set_t * data_set)
}
}
+ if (node->details->shutdown) {
+ pcmk_shutdown = 1;
+ }
+ if (pcmk_shutdown && !(node->details->running_rsc)) {
+ pcmk_clean_shutdown = 1;
+ }
notify_parent();
return;
}
@@ -339,7 +359,7 @@ crm_diff_update(const char *event, xmlNode * msg)
static mainloop_timer_t *refresh_timer = NULL;
if(refresh_timer == NULL) {
- refresh_timer = mainloop_timer_add("refresh", 2000, FALSE, mon_trigger_refresh, NULL);
+ refresh_timer = mainloop_timer_add("refresh", reconnect_msec, FALSE, mon_trigger_refresh, NULL);
refresh_trigger = mainloop_add_trigger(G_PRIORITY_LOW, mon_refresh_state, refresh_timer);
}
@@ -369,9 +389,9 @@ crm_diff_update(const char *event, xmlNode * msg)
}
/* Refresh
- * - immediately if the last update was more than 5s ago
+ * - immediately if the last update was more than 1s ago
* - every 10 updates
- * - at most 2s after the last update
+ * - at most 1s after the last update
*/
if (updates > 10 || (now - last_refresh) > (reconnect_msec / 1000)) {
mon_refresh_state(refresh_timer);
diff --git a/src/sbd.h b/src/sbd.h
index 6fe07f9..3b05a11 100644
--- a/src/sbd.h
+++ b/src/sbd.h
@@ -54,10 +54,13 @@
/* FIXME: should add dynamic check of SIG_XX >= SIGRTMAX */
/* exit status for disk-servant */
-#define EXIT_MD_IO_FAIL 20
-#define EXIT_MD_REQUEST_RESET 21
-#define EXIT_MD_REQUEST_SHUTOFF 22
-#define EXIT_MD_REQUEST_CRASHDUMP 23
+#define EXIT_MD_SERVANT_IO_FAIL 20
+#define EXIT_MD_SERVANT_REQUEST_RESET 21
+#define EXIT_MD_SERVANT_REQUEST_SHUTOFF 22
+#define EXIT_MD_SERVANT_REQUEST_CRASHDUMP 23
+
+/* exit status for pcmk-servant */
+#define EXIT_PCMK_SERVANT_GRACEFUL_SHUTDOWN 30
#define HOG_CHAR 0xff
#define SECTOR_NAME_MAX 63
--
1.8.3.1

45
SOURCES/0003-Fix-sbd-pacemaker-bail-out-of-status-earlier.patch Normal file
View File

@ -0,0 +1,45 @@
From 79b778debfee5b4ab2d099b2bfc7385f45597f70 Mon Sep 17 00:00:00 2001
From: Klaus Wenninger <klaus.wenninger@aon.at>
Date: Tue, 26 Mar 2019 11:17:45 +0100
Subject: [PATCH] Fix: sbd-pacemaker: bail out of status earlier
Prevents possible subsequent null-pointer access and avoids
unnecessary search for node.
---
src/sbd-pacemaker.c | 10 +++++++---
1 file changed, 7 insertions(+), 3 deletions(-)
diff --git a/src/sbd-pacemaker.c b/src/sbd-pacemaker.c
index c69fc55..9a8b95f 100644
--- a/src/sbd-pacemaker.c
+++ b/src/sbd-pacemaker.c
@@ -276,7 +276,7 @@ compute_status(pe_working_set_t * data_set)
static int updates = 0;
static int ever_had_quorum = FALSE;
- node_t *node = pe_find_node(data_set->nodes, local_uname);
+ node_t *node = NULL;
updates++;
@@ -286,11 +286,15 @@ compute_status(pe_working_set_t * data_set)
return;
}
+ node = pe_find_node(data_set->nodes, local_uname);
- if (node == NULL) {
+ if ((node == NULL) || (node->details == NULL)) {
set_servant_health(pcmk_health_unknown, LOG_WARNING, "Node state: %s is UNKNOWN", local_uname);
+ notify_parent();
+ return;
+ }
- } else if (node->details->online == FALSE) {
+ if (node->details->online == FALSE) {
set_servant_health(pcmk_health_unknown, LOG_WARNING, "Node state: OFFLINE");
} else if (node->details->unclean) {
--
1.8.3.1

26
SOURCES/0004-Doc-sbd.8.pod-use-the-generic-term-cluster-services-.patch Normal file
View File

@ -0,0 +1,26 @@
From 6d4289655dacad4b72fb64373c37bd1ad33649e6 Mon Sep 17 00:00:00 2001
From: "Gao,Yan" <ygao@suse.com>
Date: Tue, 22 Jan 2019 16:33:04 +0100
Subject: [PATCH] Doc: sbd.8.pod: use the generic term "cluster services"
instead of the specific "openais"
---
man/sbd.8.pod | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/man/sbd.8.pod b/man/sbd.8.pod
index ffd01c2..fed6bd3 100644
--- a/man/sbd.8.pod
+++ b/man/sbd.8.pod
@@ -476,7 +476,7 @@ storage (with internal redundancy) anyway; the SBD device does not
introduce an additional single point of failure then.
If the SBD device is not accessible, the daemon will fail to start and
-inhibit openais startup.
+inhibit startup of cluster services.
=item Two devices
--
1.8.3.1

27
SOURCES/0005-Doc-sbd.sysconfig-watchdog-timeout-set-in-the-on-dis.patch Normal file
View File

@ -0,0 +1,27 @@
From 091e10ae3f62239251b53bf7d81d47a57a9b82f2 Mon Sep 17 00:00:00 2001
From: "Gao,Yan" <ygao@suse.com>
Date: Wed, 23 Jan 2019 17:21:15 +0100
Subject: [PATCH] Doc: sbd.sysconfig: watchdog timeout set in the on-disk
metadata takes precedence
---
src/sbd.sysconfig | 3 +++
1 file changed, 3 insertions(+)
diff --git a/src/sbd.sysconfig b/src/sbd.sysconfig
index e661f96..f163f21 100644
--- a/src/sbd.sysconfig
+++ b/src/sbd.sysconfig
@@ -68,6 +68,9 @@ SBD_WATCHDOG_DEV=/dev/watchdog
# If your sbd device(s) reside on a multipath setup or iSCSI, this
# should be the time required to detect a path failure.
#
+# Be aware that watchdog timeout set in the on-disk metadata takes
+# precedence.
+#
SBD_WATCHDOG_TIMEOUT=5
## Type: string
--
1.8.3.1

142
SOURCES/0006-Refactor-fail-earlier-on-invalid-servants.patch Normal file
View File

@ -0,0 +1,142 @@
From 8301cbafed191f30656a22876941cc7c9189b623 Mon Sep 17 00:00:00 2001
From: Klaus Wenninger <klaus.wenninger@aon.at>
Date: Thu, 31 Jan 2019 14:42:01 +0100
Subject: [PATCH] Refactor: fail earlier on invalid servants
---
src/sbd-inquisitor.c | 51 ++++++++++++++++++++++++++++++++-------------------
src/sbd-md.c | 7 +------
src/sbd.h | 2 +-
3 files changed, 34 insertions(+), 26 deletions(-)
diff --git a/src/sbd-inquisitor.c b/src/sbd-inquisitor.c
index 8e0bc87..9be6c99 100644
--- a/src/sbd-inquisitor.c
+++ b/src/sbd-inquisitor.c
@@ -42,19 +42,36 @@ void recruit_servant(const char *devname, pid_t pid)
struct servants_list_item *newbie;
if (lookup_servant_by_dev(devname)) {
- cl_log(LOG_DEBUG, "Servant %s already exists", devname);
- return;
+ cl_log(LOG_DEBUG, "Servant %s already exists", devname);
+ return;
}
newbie = malloc(sizeof(*newbie));
- if (!newbie) {
- fprintf(stderr, "malloc failed in recruit_servant.\n");
- exit(1);
+ if (newbie) {
+ memset(newbie, 0, sizeof(*newbie));
+ newbie->devname = strdup(devname);
+ newbie->pid = pid;
+ newbie->first_start = 1;
+ }
+ if (!newbie || !newbie->devname) {
+ fprintf(stderr, "heap allocation failed in recruit_servant.\n");
+ exit(1);
+ }
+
+ /* some sanity-check on our newbie */
+ if (sbd_is_disk(newbie)) {
+ cl_log(LOG_INFO, "Monitoring %s", devname);
+ disk_count++;
+ } else if (sbd_is_pcmk(newbie) || sbd_is_cluster(newbie)) {
+ /* alive just after pcmk and cluster servants have shown up */
+ newbie->outdated = 1;
+ } else {
+ /* toss our newbie */
+ cl_log(LOG_ERR, "Refusing to recruit unrecognized servant %s", devname);
+ free((void *) newbie->devname);
+ free(newbie);
+ return;
}
- memset(newbie, 0, sizeof(*newbie));
- newbie->devname = strdup(devname);
- newbie->pid = pid;
- newbie->first_start = 1;
if (!s) {
servants_leader = newbie;
@@ -65,12 +82,6 @@ void recruit_servant(const char *devname, pid_t pid)
}
servant_count++;
- if(sbd_is_disk(newbie)) {
- cl_log(LOG_INFO, "Monitoring %s", devname);
- disk_count++;
- } else {
- newbie->outdated = 1;
- }
}
int assign_servant(const char* devname, functionp_t functionp, int mode, const void* argp)
@@ -148,7 +159,7 @@ void servant_start(struct servants_list_item *s)
if (sbd_is_disk(s)) {
#if SUPPORT_SHARED_DISK
DBGLOG(LOG_INFO, "Starting servant for device %s", s->devname);
- s->pid = assign_servant(s->devname, servant, start_mode, s);
+ s->pid = assign_servant(s->devname, servant_md, start_mode, s);
#else
cl_log(LOG_ERR, "Shared disk functionality not supported");
return;
@@ -785,12 +796,14 @@ parse_device_line(const char *line)
if (lpc > last) {
entry = calloc(1, 1 + lpc - last);
+ if (!entry) {
+ fprintf(stderr, "heap allocation failed parsing device-line.\n");
+ exit(1);
+ }
rc = sscanf(line + last, "%[^;]", entry);
}
- if (entry == NULL) {
- /* Skip */
- } else if (rc != 1) {
+ if (rc != 1) {
cl_log(LOG_WARNING, "Could not parse (%d %d): %s", last, lpc, line + last);
} else {
cl_log(LOG_DEBUG, "Adding '%s'", entry);
diff --git a/src/sbd-md.c b/src/sbd-md.c
index 579d273..ba2c34d 100644
--- a/src/sbd-md.c
+++ b/src/sbd-md.c
@@ -1031,7 +1031,7 @@ static int servant_check_timeout_inconsistent(struct sector_header_s *hdr)
return 0;
}
-int servant(const char *diskname, int mode, const void* argp)
+int servant_md(const char *diskname, int mode, const void* argp)
{
struct sector_mbox_s *s_mbox = NULL;
struct sector_node_s *s_node = NULL;
@@ -1046,11 +1046,6 @@ int servant(const char *diskname, int mode, const void* argp)
char uuid[37];
const struct servants_list_item *s = argp;
- if (!diskname) {
- cl_log(LOG_ERR, "Empty disk name %s.", diskname);
- return -1;
- }
-
cl_log(LOG_INFO, "Servant starting for device %s", diskname);
/* Block most of the signals */
diff --git a/src/sbd.h b/src/sbd.h
index 386c85c..6fe07f9 100644
--- a/src/sbd.h
+++ b/src/sbd.h
@@ -175,7 +175,7 @@ int ping_via_slots(const char *name, struct servants_list_item *servants);
int dump_headers(struct servants_list_item *servants);
unsigned long get_first_msgwait(struct servants_list_item *servants);
int messenger(const char *name, const char *msg, struct servants_list_item *servants);
-int servant(const char *diskname, int mode, const void* argp);
+int servant_md(const char *diskname, int mode, const void* argp);
#endif
int servant_pcmk(const char *diskname, int mode, const void* argp);
--
1.8.3.1

112
SOURCES/0007-Fix-sbd-inquisitor-overhaul-device-list-parser.patch Normal file
View File

@ -0,0 +1,112 @@
From d3be2caffb9edbb6bfe0e2658c66a1826f4e9c3a Mon Sep 17 00:00:00 2001
From: Klaus Wenninger <klaus.wenninger@aon.at>
Date: Mon, 15 Apr 2019 17:41:51 +0200
Subject: [PATCH] Fix: sbd-inquisitor: overhaul device-list-parser
for readability and robustness
---
src/sbd-inquisitor.c | 60 ++++++++++++++++++++++++++--------------------------
1 file changed, 30 insertions(+), 30 deletions(-)
diff --git a/src/sbd-inquisitor.c b/src/sbd-inquisitor.c
index 715e978..b4b5585 100644
--- a/src/sbd-inquisitor.c
+++ b/src/sbd-inquisitor.c
@@ -780,56 +780,56 @@ int inquisitor(void)
int
parse_device_line(const char *line)
{
- int lpc = 0;
- int last = 0;
- int max = 0;
+ size_t lpc = 0;
+ size_t last = 0;
+ size_t max = 0;
int found = 0;
+ bool skip_space = true;
+ int space_run = 0;
- if(line) {
- max = strlen(line);
+ if (!line) {
+ return 0;
}
- if (max <= 0) {
- return found;
- }
+ max = strlen(line);
- cl_log(LOG_DEBUG, "Processing %d bytes: [%s]", max, line);
- /* Skip initial whitespace */
- for (lpc = 0; lpc <= max && isspace(line[lpc]); lpc++) {
- last = lpc + 1;
- }
+ cl_log(LOG_DEBUG, "Processing %d bytes: [%s]", (int) max, line);
- /* Now the actual content */
for (lpc = 0; lpc <= max; lpc++) {
- int a_space = isspace(line[lpc]);
-
- if (a_space && lpc < max && isspace(line[lpc + 1])) {
- /* fast-forward to the end of the spaces */
-
- } else if (a_space || line[lpc] == ';' || line[lpc] == 0) {
- int rc = 1;
- char *entry = NULL;
+ if (isspace(line[lpc])) {
+ if (skip_space) {
+ last = lpc + 1;
+ } else {
+ space_run++;
+ }
+ continue;
+ }
+ skip_space = false;
+ if (line[lpc] == ';' || line[lpc] == 0) {
+ int rc = 0;
+ char *entry = calloc(1, 1 + lpc - last);
- if (lpc > last) {
- entry = calloc(1, 1 + lpc - last);
- if (!entry) {
- fprintf(stderr, "heap allocation failed parsing device-line.\n");
- exit(1);
- }
+ if (entry) {
rc = sscanf(line + last, "%[^;]", entry);
+ } else {
+ fprintf(stderr, "Heap allocation failed parsing device-line.\n");
+ exit(1);
}
if (rc != 1) {
- cl_log(LOG_WARNING, "Could not parse (%d %d): %s", last, lpc, line + last);
+ cl_log(LOG_WARNING, "Could not parse: '%s'", line + last);
} else {
+ entry[strlen(entry)-space_run] = '\0';
cl_log(LOG_DEBUG, "Adding '%s'", entry);
recruit_servant(entry, 0);
found++;
}
free(entry);
+ skip_space = true;
last = lpc + 1;
}
+ space_run = 0;
}
return found;
}
@@ -890,7 +890,7 @@ int main(int argc, char **argv, char **envp)
int devices = parse_device_line(value);
if(devices < 1) {
fprintf(stderr, "Invalid device line: %s\n", value);
- exit_status = -2;
+ exit_status = -2;
goto out;
}
#else
--
1.8.3.1

47
SOURCES/0008-Refactor-sbd-common-no-reason-for-stack-hogger-havin.patch Normal file
View File

@ -0,0 +1,47 @@
From 8e94781169fc2f36eb49078de1978ceb53df6b6c Mon Sep 17 00:00:00 2001
From: Klaus Wenninger <klaus.wenninger@aon.at>
Date: Mon, 15 Apr 2019 17:40:26 +0200
Subject: [PATCH] Refactor: sbd-common: no reason for stack-hogger having
retval
---
src/sbd-common.c | 10 +++++-----
1 file changed, 5 insertions(+), 5 deletions(-)
diff --git a/src/sbd-common.c b/src/sbd-common.c
index 3966f25..873a76e 100644
--- a/src/sbd-common.c
+++ b/src/sbd-common.c
@@ -568,13 +568,13 @@ enum {
#define IOPRIO_PRIO_DATA(mask) ((mask) & IOPRIO_PRIO_MASK)
#define IOPRIO_PRIO_VALUE(class, data) (((class) << IOPRIO_CLASS_SHIFT) | data)
-static unsigned char
+static void
sbd_stack_hogger(unsigned char * inbuf, int kbytes)
{
unsigned char buf[1024];
if(kbytes <= 0) {
- return HOG_CHAR;
+ return;
}
if (inbuf == NULL) {
@@ -584,10 +584,10 @@ sbd_stack_hogger(unsigned char * inbuf, int kbytes)
}
if (kbytes > 0) {
- return sbd_stack_hogger(buf, kbytes-1);
- } else {
- return buf[sizeof(buf)-1];
+ sbd_stack_hogger(buf, kbytes-1);
}
+
+ return;
}
static void
--
1.8.3.1

36
SOURCES/0009-Sanity-sbd-inquisitor-free-timeout-action-on-bail-ou.patch Normal file
View File

@ -0,0 +1,36 @@
From 5c80753afb4abc2b5b024f4a5f2fc78669bda70b Mon Sep 17 00:00:00 2001
From: Klaus Wenninger <klaus.wenninger@aon.at>
Date: Mon, 15 Apr 2019 17:39:12 +0200
Subject: [PATCH] Sanity: sbd-inquisitor: free timeout action on bail out
---
src/sbd-inquisitor.c | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/src/sbd-inquisitor.c b/src/sbd-inquisitor.c
index 77c6e4f..715e978 100644
--- a/src/sbd-inquisitor.c
+++ b/src/sbd-inquisitor.c
@@ -1088,7 +1088,8 @@ int main(int argc, char **argv, char **envp)
break;
case 'h':
usage();
- return (0);
+ goto out;
+ break;
default:
exit_status = -2;
goto out;
@@ -1241,6 +1242,9 @@ int main(int argc, char **argv, char **envp)
}
out:
+ if (timeout_action) {
+ free(timeout_action);
+ }
if (exit_status < 0) {
if (exit_status == -2) {
usage();
--
1.8.3.1

29
SOURCES/0010-Sanity-sbd-md-prevent-unrealistic-overflow-on-sector.patch Normal file
View File

@ -0,0 +1,29 @@
From f6af36a0fb05b5a37b3dfb153677e28ca5cb3fd8 Mon Sep 17 00:00:00 2001
From: Klaus Wenninger <klaus.wenninger@aon.at>
Date: Mon, 15 Apr 2019 17:37:42 +0200
Subject: [PATCH] Sanity: sbd-md: prevent unrealistic overflow on sector io
calc
---
src/sbd-md.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/src/sbd-md.c b/src/sbd-md.c
index 60a1873..f437c41 100644
--- a/src/sbd-md.c
+++ b/src/sbd-md.c
@@ -162,9 +162,9 @@ sector_io(struct sbd_context *st, int sector, void *data, int rw)
memset(&st->io, 0, sizeof(struct iocb));
if (rw) {
- io_prep_pwrite(&st->io, st->devfd, data, sector_size, sector_size * sector);
+ io_prep_pwrite(&st->io, st->devfd, data, sector_size, (long long) sector_size * sector);
} else {
- io_prep_pread(&st->io, st->devfd, data, sector_size, sector_size * sector);
+ io_prep_pread(&st->io, st->devfd, data, sector_size, (long long) sector_size * sector);
}
if (io_submit(st->ioctx, 1, ios) != 1) {
--
1.8.3.1

35
SOURCES/0011-Sanity-sbd-md-remove-some-left-over-code.patch Normal file
View File

@ -0,0 +1,35 @@
From a80fe9392fd910074eccc4733ff2cd3e1625e48e Mon Sep 17 00:00:00 2001
From: Klaus Wenninger <klaus.wenninger@aon.at>
Date: Mon, 15 Apr 2019 17:36:12 +0200
Subject: [PATCH] Sanity: sbd-md: remove some left over code
---
src/sbd-md.c | 5 -----
1 file changed, 5 deletions(-)
diff --git a/src/sbd-md.c b/src/sbd-md.c
index c51d381..60a1873 100644
--- a/src/sbd-md.c
+++ b/src/sbd-md.c
@@ -373,7 +373,6 @@ init_device(struct sbd_context *st)
struct sector_header_s *s_header;
struct sector_node_s *s_node;
struct sector_mbox_s *s_mbox;
- struct stat s;
char uuid[37];
int i;
int rc = 0;
@@ -394,10 +393,6 @@ init_device(struct sbd_context *st)
uuid_generate(s_header->uuid);
uuid_unparse_lower(s_header->uuid, uuid);
- fstat(st->devfd, &s);
- /* printf("st_size = %ld, st_blksize = %ld, st_blocks = %ld\n",
- s.st_size, s.st_blksize, s.st_blocks); */
-
cl_log(LOG_INFO, "Creating version %d.%d header on device %d (uuid: %s)",
s_header->version, s_header->minor_version,
st->devfd, uuid);
--
1.8.3.1

215
SOURCES/0012-Fix-sbd-common-query-rt-budget-0-otherwise-try-movin.patch Normal file
View File

@ -0,0 +1,215 @@
From eaeed6cca46a0223617ead834aaa576dd5ad07ff Mon Sep 17 00:00:00 2001
From: Klaus Wenninger <klaus.wenninger@aon.at>
Date: Fri, 31 May 2019 16:11:16 +0200
Subject: [PATCH] Fix: sbd-common: query rt-budget > 0 otherwise try moving to
root-slice
---
src/sbd-common.c | 110 +++++++++++++++++++++++++++++++++++++++++++++++++++
src/sbd-inquisitor.c | 15 +++++++
src/sbd.h | 2 +
src/sbd.sysconfig | 14 +++++++
4 files changed, 141 insertions(+)
diff --git a/src/sbd-common.c b/src/sbd-common.c
index 873a76e..ebfdaa3 100644
--- a/src/sbd-common.c
+++ b/src/sbd-common.c
@@ -662,6 +662,112 @@ static void sbd_memlock(int stackgrowK, int heapgrowK)
#endif
}
+static int get_realtime_budget(void)
+{
+ FILE *f;
+ char fname[PATH_MAX];
+ int res = -1, lnum = 0;
+ char *cgroup = NULL, *namespecs = NULL;
+
+ snprintf(fname, PATH_MAX, "/proc/%jd/cgroup", (intmax_t)getpid());
+ f = fopen(fname, "rt");
+ if (f == NULL) {
+ cl_log(LOG_WARNING, "Can't open cgroup file for pid=%jd",
+ (intmax_t)getpid());
+ goto exit_res;
+ }
+ while( fscanf(f, "%d:%m[^:]:%m[^\n]", &lnum, &namespecs, &cgroup) !=EOF ) {
+ if (namespecs && strstr(namespecs, "cpuacct")) {
+ free(namespecs);
+ break;
+ }
+ if (cgroup) {
+ free(cgroup);
+ cgroup = NULL;
+ }
+ if (namespecs) {
+ free(namespecs);
+ namespecs = NULL;
+ }
+ }
+ fclose(f);
+ if (cgroup == NULL) {
+ cl_log(LOG_WARNING, "Failed getting cgroup for pid=%jd",
+ (intmax_t)getpid());
+ goto exit_res;
+ }
+ snprintf(fname, PATH_MAX, "/sys/fs/cgroup/cpu%s/cpu.rt_runtime_us",
+ cgroup);
+ f = fopen(fname, "rt");
+ if (f == NULL) {
+ cl_log(LOG_WARNING, "cpu.rt_runtime_us existed for root-slice but "
+ "doesn't for '%s'", cgroup);
+ goto exit_res;
+ }
+ if (fscanf(f, "%d", &res) != 1) {
+ cl_log(LOG_WARNING, "failed reading rt-budget from %s", fname);
+ } else {
+ cl_log(LOG_INFO, "slice='%s' has rt-budget=%d", cgroup, res);
+ }
+ fclose(f);
+
+exit_res:
+ if (cgroup) {
+ free(cgroup);
+ }
+ return res;
+}
+
+/* stolen from corosync */
+static int sbd_move_to_root_cgroup(bool enforce_root_cgroup) {
+ FILE *f;
+ int res = -1;
+
+ /*
+ * /sys/fs/cgroup is hardcoded, because most of Linux distributions are now
+ * using systemd and systemd uses hardcoded path of cgroup mount point.
+ *
+ * This feature is expected to be removed as soon as systemd gets support
+ * for managing RT configuration.
+ */
+ f = fopen("/sys/fs/cgroup/cpu/cpu.rt_runtime_us", "rt");
+ if (f == NULL) {
+ cl_log(LOG_DEBUG, "cpu.rt_runtime_us doesn't exist -> "
+ "system without cgroup or with disabled CONFIG_RT_GROUP_SCHED");
+ res = 0;
+ goto exit_res;
+ }
+ fclose(f);
+
+ if ((!enforce_root_cgroup) && (get_realtime_budget() > 0)) {
+ cl_log(LOG_DEBUG, "looks as if we have rt-budget in the slice we are "
+ "-> skip moving to root-slice");
+ res = 0;
+ goto exit_res;
+ }
+
+ f = fopen("/sys/fs/cgroup/cpu/tasks", "w");
+ if (f == NULL) {
+ cl_log(LOG_WARNING, "Can't open cgroups tasks file for writing");
+
+ goto exit_res;
+ }
+
+ if (fprintf(f, "%jd\n", (intmax_t)getpid()) <= 0) {
+ cl_log(LOG_WARNING, "Can't write sbd pid into cgroups tasks file");
+ goto close_and_exit_res;
+ }
+
+close_and_exit_res:
+ if (fclose(f) != 0) {
+ cl_log(LOG_WARNING, "Can't close cgroups tasks file");
+ goto exit_res;
+ }
+
+exit_res:
+ return (res);
+}
+
void
sbd_make_realtime(int priority, int stackgrowK, int heapgrowK)
{
@@ -670,6 +776,10 @@ sbd_make_realtime(int priority, int stackgrowK, int heapgrowK)
}
#ifdef SCHED_RR
+ if (move_to_root_cgroup) {
+ sbd_move_to_root_cgroup(enforce_moving_to_root_cgroup);
+ }
+
{
int pcurrent = 0;
int pmin = sched_get_priority_min(SCHED_RR);
diff --git a/src/sbd-inquisitor.c b/src/sbd-inquisitor.c
index abde4e5..cef5cc7 100644
--- a/src/sbd-inquisitor.c
+++ b/src/sbd-inquisitor.c
@@ -33,6 +33,8 @@ int start_mode = 0;
char* pidfile = NULL;
bool do_flush = true;
char timeout_sysrq_char = 'b';
+bool move_to_root_cgroup = true;
+bool enforce_moving_to_root_cgroup = false;
int parse_device_line(const char *line);
@@ -965,6 +967,19 @@ int main(int argc, char **argv, char **envp)
timeout_action = strdup(value);
}
+ value = getenv("SBD_MOVE_TO_ROOT_CGROUP");
+ if(value) {
+ move_to_root_cgroup = crm_is_true(value);
+
+ if (move_to_root_cgroup) {
+ enforce_moving_to_root_cgroup = true;
+ } else {
+ if (strcmp(value, "auto") == 0) {
+ move_to_root_cgroup = true;
+ }
+ }
+ }
+
while ((c = getopt(argc, argv, "czC:DPRTWZhvw:d:n:p:1:2:3:4:5:t:I:F:S:s:r:")) != -1) {
switch (c) {
case 'D':
diff --git a/src/sbd.h b/src/sbd.h
index 3b05a11..ac30ec7 100644
--- a/src/sbd.h
+++ b/src/sbd.h
@@ -159,6 +159,8 @@ extern bool watchdogdev_is_default;
extern char* local_uname;
extern bool do_flush;
extern char timeout_sysrq_char;
+extern bool move_to_root_cgroup;
+extern bool enforce_moving_to_root_cgroup;
/* Global, non-tunable variables: */
extern int sector_size;
diff --git a/src/sbd.sysconfig b/src/sbd.sysconfig
index f163f21..e1a60ed 100644
--- a/src/sbd.sysconfig
+++ b/src/sbd.sysconfig
@@ -91,6 +91,20 @@ SBD_WATCHDOG_TIMEOUT=5
#
SBD_TIMEOUT_ACTION=flush,reboot
+## Type: yesno / auto
+## Default: auto
+#
+# If CPUAccounting is enabled default is not to assign any RT-budget
+# to the system.slice which prevents sbd from running RR-scheduled.
+#
+# One way to escape that issue is to move sbd-processes from the
+# slice they were originally started to root-slice.
+# Of course starting sbd in a certain slice might be intentional.
+# Thus in auto-mode sbd will check if the slice has RT-budget assigned.
+# If that is the case sbd will stay in that slice while it will
+# be moved to root-slice otherwise.
+SBD_MOVE_TO_ROOT_CGROUP=auto
+
## Type: string
## Default: ""
#
--
1.8.3.1

60
SOURCES/0013-Fix-sbd-pacemaker-assume-graceful-exit-if-leftovers-.patch Normal file
View File

@ -0,0 +1,60 @@
From 824fe834c67fb7bae7feb87607381f9fa8fa2945 Mon Sep 17 00:00:00 2001
From: Klaus Wenninger <klaus.wenninger@aon.at>
Date: Fri, 7 Jun 2019 19:09:06 +0200
Subject: [PATCH] Fix: sbd-pacemaker: assume graceful exit if leftovers are
unmanged
---
src/sbd-pacemaker.c | 32 +++++++++++++++++++++++++++++++-
1 file changed, 31 insertions(+), 1 deletion(-)
diff --git a/src/sbd-pacemaker.c b/src/sbd-pacemaker.c
index 9a8b95f..2b35ff6 100644
--- a/src/sbd-pacemaker.c
+++ b/src/sbd-pacemaker.c
@@ -333,11 +333,41 @@ compute_status(pe_working_set_t * data_set)
}
}
+ /* If we are in shutdown-state once this will go on till the end.
+ * If we've on top reached a state of 0 locally running resources
+ * we can assume a clean shutdown.
+ * Tricky are the situations where the node is in maintenance-mode
+ * or resources are unmanaged. So if the node is in maintenance or
+ * all left-over running resources are unmanaged we assume intention.
+ */
if (node->details->shutdown) {
pcmk_shutdown = 1;
}
- if (pcmk_shutdown && !(node->details->running_rsc)) {
+ if (pcmk_shutdown)
+ {
pcmk_clean_shutdown = 1;
+ if (!(node->details->maintenance)) {
+ GListPtr iter;
+
+ for (iter = node->details->running_rsc;
+ iter != NULL; iter = iter->next) {
+ resource_t *rsc = (resource_t *) iter->data;
+
+
+ if (is_set(rsc->flags, pe_rsc_managed)) {
+ pcmk_clean_shutdown = 0;
+ crm_debug("not clean as %s managed and still running",
+ rsc->id);
+ break;
+ }
+ }
+ if (pcmk_clean_shutdown) {
+ crm_debug("pcmk_clean_shutdown because "
+ "all managed resources down");
+ }
+ } else {
+ crm_debug("pcmk_clean_shutdown because node is in maintenance");
+ }
}
notify_parent();
return;
--
1.8.3.1

123
SOURCES/0014-Fix-sbd-cluster-periodically-check-corosync-daemon-l.patch Normal file
View File

@ -0,0 +1,123 @@
From 1387ed890e3a9e246e9b9f780b2a7cb5379459ab Mon Sep 17 00:00:00 2001
From: Klaus Wenninger <klaus.wenninger@aon.at>
Date: Wed, 5 Jun 2019 11:32:49 +0200
Subject: [PATCH] Fix: sbd-cluster: periodically check corosync-daemon liveness
using votequorum_getinfo.
---
configure.ac | 12 +++++++++++-
src/sbd-cluster.c | 36 ++++++++++++++++++++++++++++++++++--
2 files changed, 45 insertions(+), 3 deletions(-)
diff --git a/configure.ac b/configure.ac
index fac26a8..c44e747 100644
--- a/configure.ac
+++ b/configure.ac
@@ -33,6 +33,7 @@ PKG_CHECK_MODULES(glib, [glib-2.0])
dnl PKG_CHECK_MODULES(libcoroipcc, [libcoroipcc])
PKG_CHECK_MODULES(cmap, [libcmap], HAVE_cmap=1, HAVE_cmap=0)
+PKG_CHECK_MODULES(votequorum, [libvotequorum], HAVE_votequorum=1, HAVE_votequorum=0)
dnl pacemaker > 1.1.8
PKG_CHECK_MODULES(pacemaker, [pacemaker, pacemaker-cib], HAVE_pacemaker=1, HAVE_pacemaker=0)
@@ -49,7 +50,12 @@ elif test $HAVE_pacemaker = 1; then
if test $HAVE_cmap = 0; then
AC_MSG_NOTICE(No package 'cmap' found)
else
- CPPFLAGS="$CPPFLAGS $cmap_CFLAGS"
+ CPPFLAGS="$CPPFLAGS $cmap_CFLAGS"
+ fi
+ if test $HAVE_votequorum = 0; then
+ AC_MSG_NOTICE(No library 'votequorum' found)
+ else
+ CPPFLAGS="$CPPFLAGS $votequorum_CFLAGS"
fi
fi
@@ -66,6 +72,7 @@ AC_CHECK_LIB(pe_rules, test_rule, , missing="yes")
AC_CHECK_LIB(crmcluster, crm_peer_init, , missing="yes")
AC_CHECK_LIB(uuid, uuid_unparse, , missing="yes")
AC_CHECK_LIB(cmap, cmap_initialize, , HAVE_cmap=0)
+AC_CHECK_LIB(votequorum, votequorum_getinfo, , HAVE_votequorum=0)
dnl pacemaker >= 1.1.8
AC_CHECK_HEADERS(pacemaker/crm/cluster.h)
@@ -107,6 +114,9 @@ fi
AC_DEFINE_UNQUOTED(CHECK_TWO_NODE, $HAVE_cmap, Turn on checking for 2-node cluster)
AM_CONDITIONAL(CHECK_TWO_NODE, test "$HAVE_cmap" = "1")
+AC_DEFINE_UNQUOTED(CHECK_VOTEQUORUM_HANDLE, $HAVE_votequorum, Turn on periodic checking of votequorum-handle)
+AM_CONDITIONAL(CHECK_VOTEQUORUM_HANDLE, test "$HAVE_votequorum" = "1")
+
CONFIGDIR=""
AC_ARG_WITH(configdir,
[ --with-configdir=DIR
diff --git a/src/sbd-cluster.c b/src/sbd-cluster.c
index 541212f..9fb6224 100644
--- a/src/sbd-cluster.c
+++ b/src/sbd-cluster.c
@@ -80,6 +80,12 @@ sbd_plugin_membership_dispatch(cpg_handle_t handle,
#if SUPPORT_COROSYNC
+#if CHECK_VOTEQUORUM_HANDLE
+#include <corosync/votequorum.h>
+
+static votequorum_handle_t votequorum_handle = 0;
+#endif
+
static bool two_node = false;
static bool ever_seen_both = false;
static int cpg_membership_entries = -1;
@@ -261,12 +267,32 @@ notify_timer_cb(gpointer data)
#endif
case pcmk_cluster_corosync:
+ do {
+#if SUPPORT_COROSYNC && CHECK_VOTEQUORUM_HANDLE
+ struct votequorum_info info;
+
+ if (votequorum_getinfo(votequorum_handle, 0, &info) != CS_OK) {
+
+ votequorum_finalize(votequorum_handle);
+ if (votequorum_initialize(&votequorum_handle, NULL) != CS_OK) {
+ votequorum_handle = 0;
+ break;
+ }
+ if (votequorum_getinfo(votequorum_handle, 0, &info) != CS_OK) {
+ break;
+ }
+ }
+#endif
+ notify_parent();
+ } while (0);
+ break;
+
#if HAVE_DECL_PCMK_CLUSTER_CMAN
case pcmk_cluster_cman:
-#endif
- /* TODO - Make a CPG call and only call notify_parent() when we get a reply */
+
notify_parent();
break;
+#endif
default:
break;
@@ -533,6 +559,12 @@ find_pacemaker_remote(void)
static void
clean_up(int rc)
{
+#if CHECK_VOTEQUORUM_HANDLE
+ votequorum_finalize(votequorum_handle);
+ votequorum_handle = 0; /* there isn't really an invalid handle value
+ * just to be back where we started
+ */
+#endif
return;
}
--
1.8.3.1

26
SOURCES/0015-build-say-library-when-missing-cmap-not-package-to-a.patch Normal file
View File

@ -0,0 +1,26 @@
From 0de14256fc873aee735117955662503b773bf71c Mon Sep 17 00:00:00 2001
From: Klaus Wenninger <klaus.wenninger@aon.at>
Date: Tue, 11 Jun 2019 08:05:33 +0200
Subject: [PATCH] build: say library when missing cmap not package to avoid
confusion
---
configure.ac | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/configure.ac b/configure.ac
index c44e747..1c55094 100644
--- a/configure.ac
+++ b/configure.ac
@@ -48,7 +48,7 @@ if test $HAVE_pacemaker = 0 -a $HAVE_pcmk = 0; then
elif test $HAVE_pacemaker = 1; then
CPPFLAGS="$CPPFLAGS $glib_CFLAGS $pacemaker_CFLAGS"
if test $HAVE_cmap = 0; then
- AC_MSG_NOTICE(No package 'cmap' found)
+ AC_MSG_NOTICE(No library 'cmap' found)
else
CPPFLAGS="$CPPFLAGS $cmap_CFLAGS"
fi
--
1.8.3.1

334
SPECS/sbd.spec Normal file
View File

@ -0,0 +1,334 @@
#
# spec file for package sbd
#
# Copyright (c) 2014 SUSE LINUX Products GmbH, Nuernberg, Germany.
# Copyright (c) 2013 Lars Marowsky-Bree
#
# All modifications and additions to the file contributed by third parties
# remain the property of their copyright owners, unless otherwise agreed
# upon. The license for this file, and modifications and additions to the
# file, is the same license as for the pristine package itself (unless the
# license for the pristine package is not an Open Source License, in which
# case the license is the MIT License). An "Open Source License" is a
# license that conforms to the Open Source Definition (Version 1.9)
# published by the Open Source Initiative.
# Please submit bugfixes or comments via http://bugs.opensuse.org/
#
%global commit 7f33d1a409d0a4e2cd69946688c48eaa8f3c5d26
%global shortcommit %(c=%{commit}; echo ${c:0:7})
%global github_owner Clusterlabs
%global buildnum 10
Name: sbd
Summary: Storage-based death
License: GPLv2+
Group: System Environment/Daemons
Version: 1.4.0
Release: %{buildnum}%{?dist}
Url: https://github.com/%{github_owner}/%{name}
Source0: https://github.com/%{github_owner}/%{name}/archive/%{commit}/%{name}-%{commit}.tar.gz
Patch0: 0001-Fix-sbd-cluster-finalize-cmap-connection-if-disconne.patch
Patch1: 0002-Fix-sbd-pacemaker-make-handling-of-cib-connection-lo.patch
Patch2: 0003-Fix-sbd-pacemaker-bail-out-of-status-earlier.patch
Patch3: 0004-Doc-sbd.8.pod-use-the-generic-term-cluster-services-.patch
Patch4: 0005-Doc-sbd.sysconfig-watchdog-timeout-set-in-the-on-dis.patch
Patch5: 0006-Refactor-fail-earlier-on-invalid-servants.patch
Patch6: 0007-Fix-sbd-inquisitor-overhaul-device-list-parser.patch
Patch7: 0008-Refactor-sbd-common-no-reason-for-stack-hogger-havin.patch
Patch8: 0009-Sanity-sbd-inquisitor-free-timeout-action-on-bail-ou.patch
Patch9: 0010-Sanity-sbd-md-prevent-unrealistic-overflow-on-sector.patch
Patch10: 0011-Sanity-sbd-md-remove-some-left-over-code.patch
Patch11: 0012-Fix-sbd-common-query-rt-budget-0-otherwise-try-movin.patch
Patch12: 0013-Fix-sbd-pacemaker-assume-graceful-exit-if-leftovers-.patch
Patch13: 0014-Fix-sbd-cluster-periodically-check-corosync-daemon-l.patch
Patch14: 0015-build-say-library-when-missing-cmap-not-package-to-a.patch
BuildRoot: %{_tmppath}/%{name}-%{version}-build
BuildRequires: autoconf
BuildRequires: automake
BuildRequires: libuuid-devel
BuildRequires: glib2-devel
BuildRequires: libaio-devel
BuildRequires: corosynclib-devel
BuildRequires: pacemaker-libs-devel > 1.1.12
BuildRequires: libtool
BuildRequires: libuuid-devel
BuildRequires: libxml2-devel
BuildRequires: pkgconfig
BuildRequires: systemd
%if 0%{?rhel} > 0
ExclusiveArch: i686 x86_64 s390x ppc64le aarch64
%endif
%if %{defined systemd_requires}
%systemd_requires
%endif
%description
This package contains the storage-based death functionality.
###########################################################
%prep
%autosetup -n %{name}-%{commit} -p1
###########################################################
%build
autoreconf -i
export CFLAGS="$RPM_OPT_FLAGS -Wall -Werror"
%configure
make %{?_smp_mflags}
###########################################################
%install
make DESTDIR=$RPM_BUILD_ROOT LIBDIR=%{_libdir} install
rm -rf ${RPM_BUILD_ROOT}%{_libdir}/stonith
%if %{defined _unitdir}
install -D -m 0644 src/sbd.service $RPM_BUILD_ROOT/%{_unitdir}/sbd.service
install -D -m 0644 src/sbd_remote.service $RPM_BUILD_ROOT/%{_unitdir}/sbd_remote.service
%endif
mkdir -p ${RPM_BUILD_ROOT}%{_sysconfdir}/sysconfig
install -m 644 src/sbd.sysconfig ${RPM_BUILD_ROOT}%{_sysconfdir}/sysconfig/sbd
###########################################################
%clean
rm -rf %{buildroot}
%if %{defined _unitdir}
%post
%systemd_post sbd.service
%systemd_post sbd_remote.service
if [ $1 -ne 1 ] ; then
if systemctl --quiet is-enabled sbd.service 2>/dev/null
then
systemctl --quiet reenable sbd.service 2>/dev/null || :
fi
if systemctl --quiet is-enabled sbd_remote.service 2>/dev/null
then
systemctl --quiet reenable sbd_remote.service 2>/dev/null || :
fi
fi
%preun
%systemd_preun sbd.service
%systemd_preun sbd_remote.service
%postun
%systemd_postun sbd.service
%systemd_postun sbd_remote.service
%endif
%files
###########################################################
%defattr(-,root,root)
%config(noreplace) %{_sysconfdir}/sysconfig/sbd
%{_sbindir}/sbd
#%{_datadir}/sbd
%doc %{_mandir}/man8/sbd*
%if %{defined _unitdir}
%{_unitdir}/sbd.service
%{_unitdir}/sbd_remote.service
%endif
%doc COPYING
%changelog
* Wed Jun 12 2019 Klaus Wenninger <kwenning@redhat.com> - 1.4.0-10
- added missing patches to git
Resolves: rhbz#1702727
Resolves: rhbz#1718296
* Tue Jun 11 2019 Klaus Wenninger <kwenning@redhat.com> - 1.4.0-9
- assume graceful pacemaker exit if leftovers are unmanaged
- query corosync liveness via votequorum-api
Resolves: rhbz#1702727
Resolves: rhbz#1718296
* Mon Jun 3 2019 Klaus Wenninger <kwenning@redhat.com> - 1.4.0-8
- check for rt-budget > 0 and move to root-slice otherwise
Resolves: rhbz#1713021
* Wed Apr 10 2019 Klaus Wenninger <kwenning@redhat.com> - 1.4.0-7
- add some minor fixes from upstream found by coverity
Resolves: rhbz#1698056
* Wed Apr 10 2019 Klaus Wenninger <kwenning@redhat.com> - 1.4.0-6
- add decision-context to gating.yaml
Resolves: rhbz#1682137
* Mon Jan 14 2019 Klaus Wenninger <kwenning@redhat.com> - 1.4.0-5
- rebase to upstream v1.4.0
- finalize cmap connection if disconnected from cluster
- make handling of cib-connection loss more robust
- add ci test files
- use generic term cluster-services in doc
- stress in doc that on-disk metadata watchdog-timeout
takes precedence
- fail earlier on invalid servants to make gcc 9 happy
Resolves: rhbz#1698056
Resolves: rhbz#1682137
* Mon Dec 17 2018 Klaus Wenninger <kwenning@redhat.com> - 1.3.1-18
- make timeout-action executed by sbd configurable
Resolves: rhbz#1660147
* Mon Dec 3 2018 Klaus Wenninger <kwenning@redhat.com> - 1.3.1-17
- use pacemaker's new pe api with constructors/destructors
Resolves: rhbz#1650663
* Wed Sep 19 2018 Klaus Wenninger <kwenning@redhat.com> - 1.3.1-16
- avoid statting potential symlink-targets in /dev
Resolves: rhbz#1629020
* Wed Sep 19 2018 Klaus Wenninger <kwenning@redhat.com> - 1.3.1-15
- rebuild against new versions of libqb (1.0.3-7.el8),
corosync (2.99.3-4.el8) and pacemaker (2.0.0-9.el8)
Related: rhbz#1615945
* Fri Sep 14 2018 Klaus Wenninger <kwenning@redhat.com> - 1.3.1-14
- skip symlinks pointing to dev-nodes outside of /dev
Resolves: rhbz#1629020
* Wed Sep 5 2018 Klaus Wenninger <kwenning@redhat.com> - 1.3.1-13
- Require systemd-package during build to have the macros
Resolves: rhbz#1625553
* Mon Jul 30 2018 Florian Weimer <fweimer@redhat.com> - 1.3.1-12
- Rebuild with fixed binutils
* Tue Jul 3 2018 <kwenning@redhat.com> - 1.3.1-11
- replaced tarball by version downloaded from github
* Mon Jul 2 2018 <kwenning@redhat.com> - 1.3.1-10
- removed unneeded python build-dependency
- updated legacy corosync-devel to corosynclib-devel
Resolves: rhbz#1595856
* Fri May 4 2018 <kwenning@redhat.com> - 1.3.1-9
- use cib-api directly as get_cib_copy gone with
pacemaker 2.0.0
- add sys/sysmacros.h to build with glibc-2.25
- enlarge string buffer to satisfy newer gcc
- no corosync 1 support with pacemaker 2.0.0
- set default to LOG_NOTICE + overhaul levels
- refactor proc-parsing
- adaptions for daemon-names changed with
pacemaker 2.0.0 rc3
- added .do-not-sync-with-fedora
Resolves: rhbz#1571797
* Mon Apr 16 2018 <kwenning@redhat.com> - 1.3.1-8
- Added aarch64 target
Resolves: rhbz#1568029
* Mon Jan 15 2018 <kwenning@redhat.com> - 1.3.1-7
- reenable sbd on upgrade so that additional
links to make pacemaker properly depend on
sbd are created
Resolves: rhbz#1525981
* Wed Jan 10 2018 <kwenning@redhat.com> - 1.3.1-5
- add man sections for query- & test-watchdog
Resolves: rhbz#1462002
* Wed Dec 20 2017 <kwenning@redhat.com> - 1.3.1-3
- mention timeout caveat with SBD_DELAY_START
in configuration template
- make systemd wait for sbd-start to finish
before starting pacemaker or dlm
Resolves: rhbz#1525981
* Fri Nov 3 2017 <kwenning@redhat.com> - 1.3.1-2
- rebase to upstream v1.3.1
Resolves: rhbz#1499864
rhbz#1468580
rhbz#1462002
* Wed Jun 7 2017 <kwenning@redhat.com> - 1.3.0-3
- prevent creation of duplicate servants
- check 2Node flag in corosync to support
2-node-clusters with shared disk fencing
- move disk-triggered reboot/off/crashdump
to inquisitor to have sysrq observed by watchdog
Resolves: rhbz#1413951
* Sun Mar 26 2017 <kwenning@redhat.com> - 1.3.0-1
- rebase to upstream v1.3.0
- remove watchdog-limitation from description
Resolves: rhbz#1413951
* Mon Feb 27 2017 <kwenning@redhat.com> - 1.2.1-23
- if shared-storage enabled check for node-name <= 63 chars
Resolves: rhbz#1413951
* Tue Jan 31 2017 <kwenning@redhat.com> - 1.2.1-22
- Rebuild with shared-storage enabled
- Package original manpage
- Added ppc64le target
Resolves: rhbz#1413951
* Fri Apr 15 2016 <kwenning@redhat.com> - 1.2.1-21
- Rebuild for new pacemaker
Resolves: rhbz#1320400
* Fri Apr 15 2016 <kwenning@redhat.com> - 1.2.1-20
- tarball updated to c511b0692784a7085df4b1ae35748fb318fa79ee
from https://github.com/Clusterlabs/sbd
Resolves: rhbz#1324240
* Thu Jul 23 2015 <abeekhof@redhat.com> - 1.2.1-5
- Rebuild for pacemaker
* Tue Jun 02 2015 <abeekhof@redhat.com> - 1.2.1-4
- Include the dist tag in the release string
- Rebuild for new pacemaker
* Mon Jan 12 2015 <abeekhof@redhat.com> - 1.2.1-3
- Correctly parse SBD_WATCHDOG_TIMEOUT into seconds (not milliseconds)
* Mon Oct 27 2014 <abeekhof@redhat.com> - 1.2.1-2
- Correctly enable /proc/pid validation for sbd_lock_running()
* Fri Oct 24 2014 <abeekhof@redhat.com> - 1.2.1-1
- Further improve integration with the el7 environment
* Thu Oct 16 2014 <abeekhof@redhat.com> - 1.2.1-0.5.872e82f3.git
- Disable unsupported functionality (for now)
* Wed Oct 15 2014 <abeekhof@redhat.com> - 1.2.1-0.4.872e82f3.git
- Improved integration with the el7 environment
* Tue Sep 30 2014 <abeekhof@redhat.com> - 1.2.1-0.3.8f912945.git
- Only build on archs supported by the HA Add-on
* Fri Aug 29 2014 <abeekhof@redhat.com> - 1.2.1-0.2.8f912945.git
- Remove some additional SUSE-isms
* Fri Aug 29 2014 <abeekhof@redhat.com> - 1.2.1-0.1.8f912945.git
- Prepare for package review