Recommended update for cluster-glue and SBD
SUSE Recommended Update: Recommended update for cluster-glue and SBD
This update to the STONITH Block Device (SBD) adds support
for taking the quorum state and Pacemaker health into
account when the majority of SBD devices is unreachable.
This allows clusters running SBD to survive transient
faults in the storage layer as long as the cluster
partition remains quorate and the node is in a healthy
state. This new feature is activated using the "-P" flag
in /etc/sysconfig/sbd; please see the sbd man page for
further details.
IO to the devices is now fully asynchronous, and enforces
its own timeouts and retries. Previously, IO processes
could get stuck on certain multipath issues, and SBD had
to wait until the SCSI IO timed out. This allows SBD's
watchdog and msgwait timeouts to be reduced and thus
results in faster fail-overs in the cluster.
The restart logic of IO processes has been improved and
clarified in the man page.
It is now possible to specify a different watchdog timeout
for the crashdump support, so that crashdumps do not get
interrupted by the watchdog triggering.
Missing devices no longer fail the list or dump commands.
sbd is now provided as a separate RPM from cluster-glue;
both are automatically installed by this update.
Please refer to the changelog for further details.
https://bugzilla.novell.com/738295
https://bugzilla.novell.com/753559
https://bugzilla.novell.com/761797
https://bugzilla.novell.com/762300
https://bugzilla.novell.com/767293
http://download.suse.com/patch/finder/?keywords=3abae703e3caa9c58bc38f6d04a2387b
| Announcement ID: | SUSE-RU-2012:0839-1 |
| Rating: | low |
| References: | #738295 #753559 #761797 #762300 #767293 |
| Affected Products: |
An update that has 5 recommended fixes can now be installed. It includes one version update.
Description:
This update to the STONITH Block Device (SBD) adds support
for taking the quorum state and Pacemaker health into
account when the majority of SBD devices is unreachable.
This allows clusters running SBD to survive transient
faults in the storage layer as long as the cluster
partition remains quorate and the node is in a healthy
state. This new feature is activated using the "-P" flag
in /etc/sysconfig/sbd; please see the sbd man page for
further details.
IO to the devices is now fully asynchronous, and enforces
its own timeouts and retries. Previously, IO processes
could get stuck on certain multipath issues, and SBD had
to wait until the SCSI IO timed out. This allows SBD's
watchdog and msgwait timeouts to be reduced and thus
results in faster fail-overs in the cluster.
The restart logic of IO processes has been improved and
clarified in the man page.
It is now possible to specify a different watchdog timeout
for the crashdump support, so that crashdumps do not get
interrupted by the watchdog triggering.
Missing devices no longer fail the list or dump commands.
sbd is now provided as a separate RPM from cluster-glue;
both are automatically installed by this update.
Please refer to the changelog for further details.
Patch Instructions:
To install this SUSE Recommended Update use YaST online_update.
Alternatively you can run the command listed for your product:
- SUSE Linux Enterprise High Availability Extension 11 SP2:
zypper in -t patch sleshasp2-cluster-glue-sbd-6438
To bring your system up-to-date, use "zypper patch".
Package List:
- SUSE Linux Enterprise High Availability Extension 11 SP2 (i586 ia64 ppc64 s390x x86_64) [New Version: 1.0.9.1]:
- cluster-glue-1.0.9.1-0.36.1
- libglue-devel-1.0.9.1-0.36.1
- libglue2-1.0.9.1-0.36.1
- sbd-1.0-0.5.6