RHBA-2007:0580 - Bug Fix Advisory

Topic

Updated rgmanager packages that fix various bugs and add enhancements are
now available.

Description

The rgmanager package contains the Red Hat Resource Group Manager, which
provides high availability of critical server applications in the event of
planned or unplanned system downtime.

These updated packages address the following issues:

• running the "clustat" command when qdisk has been configured caused the

status of nodes to be reported incorrectly. Interaction between rgmanager
and qdisk has been resolved in this update.

• if the node processing a request from clusvcadm crashed, clusvcadm would

hang. This has been resolved in these updated packages.

• using rg_test to stop, start then stop a cluster service with an IP

address and no netmask specifically defined, caused the following warning
"Warning: Executing wildcard deletion to stay compatible with old scripts.
Explicitly specify the prefix length (<ip-address/prefix>) to avoid this
warning. This special behavior is likely to disappear in further releases,
fix your scripts!". This was due to a change in how iproute2 parses
parameters. These updated packages resolve this issue.

• a fail-restart loop occurred instead of failover when a link was

unplugged. In these updated packages ip.sh fails during the start phase
if a link is unplugged.

• previously rgmanager could stop resources in the wrong order. Implicit

ordering guarantees are now provided in cluster.conf, resolving this issue.

• rgmanager live migration for VMs is now enabled and functional.
• nodes were able to run multiple resources marked as "exclusive" which

could lead to data corruption. These updated packages change this
behavior, allowing only one service to run as "exclusive".

• subtrees can now be restarted without restarting the service.
• previously resource agents were not probing according to configured time

intervals. This behavior has been resolved in this update.

• restarting rgmanager no longer restarts the node erroneously.
• the rgmanager init script now provides correct exit codes.
• rgmanager now correctly mounts the Veritas File System (VxFS).
• previously stopping the rgmanager service on a node caused rgmanager to

exit uncleanly without stopping the Distributed Lock Manager (DLM)
service. This prevented the Cluster Manager (CMAN) service stopping on the
last node. Rgmanager's DLM lock space is no longer left open if the CMAN
service is stopped first, resolving this issue.

• rgmanager now prevents more than one status check from being queued,

preventing long status checks repeatedly being queued.

• rgmanager no longer restarts virtual machines when adding them as a

Cluster Managed service.

• rgmanager no longer causes a node to restart if it is shutdown while

processing a relocation request.

• a memory leak occurring during the simultaneous relocation of multiple

services has been fixed.

This update also adds the following enhancements:

• there is now a clusvcadm flag to respect failover domain priorities

when enabling services.

• rgmanager now detects virtual machine migration if non-cluster tools are

allowed.

• nested resources can now inherit attributes from all direct ancestors,

not just direct parents.

All rgmanager users should upgrade to these latest packages, which
resolve these issues and add these enhancements.

Solution

Before applying this update, make sure all previously released errata
relevant to your system have been applied.

This update is available via Red Hat Network. To use Red Hat Network,
launch the Red Hat Update Agent with the following command:

up2date

This will start an interactive process that will result in the appropriate
RPMs being upgraded on your system.

Affected Products

• Red Hat Enterprise Linux High Availability for x86_64 5 x86_64
• Red Hat Enterprise Linux High Availability for x86_64 5 ia64
• Red Hat Enterprise Linux High Availability for x86_64 5 i386
• Red Hat Enterprise Linux High Availability (for RHEL Server) from RHUI 5 x86_64
• Red Hat Enterprise Linux High Availability (for RHEL Server) from RHUI 5 i386

Fixes

• BZ - 209526 - clusvcadm hangs if node processing request dies
• BZ - 211469 - RFE: flag for clusvcadm to respect failover domain priorities (rhcs5)
• BZ - 212479 - ip.sh causes /sbin/ip to produce warnings
• BZ - 222445 - Thousands of clurgmgrd threads when gfs exported thru nfs
• BZ - 222484 - ip.sh does not fail during 'start' phase if link is unplugged
• BZ - 223230 - clustat has bizarre output when qdisk is in use
• BZ - 225421 - rgmanager stops the resources in wrong order [RHCS5]
• BZ - 229650 - Restart only the failed resource and its dependencies instead of the whole service.
• BZ - 231477 - clusvcadm getopt rejects -M (vm live migration)
• BZ - 231521 - attr inheritance should try all direct ancestors
• BZ - 231692 - rgmanager live migration of vms doesn't work
• BZ - 232300 - rgmanager should discover migration if someone migrates a vm out from under it
• BZ - 234249 - rgmanager should ignore backup files in /usr/share/cluster/
• BZ - 234589 - rgmanager not working when using a quorum disk
• BZ - 235178 - resource agents probe more often than configured
• BZ - 236204 - rgmanager restarts system when trying to stop it just after it's start
• BZ - 237144 - Multiple "exclusive" services are started on the same node
• BZ - 239590 - HA LVM can collide with CLVM
• BZ - 239608 - HA LVM should prevent users from running an invalid setup
• BZ - 241673 - BA LVM HA issues tracker
• BZ - 243691 - Wrong init script
• BZ - 244143 - rgmanager is exiting without stopping dlm service
• BZ - 247291 - shutdown while processing relocation request results in node reboot
• BZ - 247488 - status check count in queue is unbounded, but should be 1
• BZ - 248727 - Don't bounce Virtual Machine when adding as a managed VM service

CVEs

(none)

References

(none)