RHBA-2008:0791 - Bug Fix Advisory
rgmanager bug fix and enhancement update
Bug Fix Advisory
An updated rgmanager package that fixes several bugs and adds enhancements
is now available.
The rgmanager package contains the Red Hat Resource Group Manager, which
provides high availability for critical server applications in the event of
planned or unplanned system downtime.
This updated package fixes the following bugs:
- rgmanager restarted services when non-critical parameters were changed,
such as "autostart".
- relocating an NFS service to another node, and then back to the original
node, caused the service to fail. Relocating NFS services works correctly
in this updated package.
- even when "force_unmount" was enabled, a process holding a lock on the
file system prevented failover for NFS mount points. Now, when
"force_unmount" is enabled, a failover for NFS mount points occurs, even if
a process is holding a lock on the file system.
- the HA LVM service failed to relocate during heavy I/O operations. Now,
the HA LVM service can be relocated during heavy I/O.
- time outs occurred while waiting for responses from rgmanager, due to
unbounded thread counts. This caused rgmanager to be unavailable, and a
"Timed out waiting for a response from Resource Group Manager" error
- "ip.sh" did not set the correct netmask. Now, "ip.sh" uses the netmask
set in "cluster.conf", instead of "255.255.255.0".
- when at least two services used HA LVM, tags were incorrectly stripped
from all volume groups. Now, "lvm_by_vg.sh" only strips tags from
configured volume groups.
- the clusvcadm man page has been updated to remove references to
- rgmanager no longer hangs during shutdown because of a previous service
- in certain cases, the "quotaoff" command caused hangs. This may have
- symbolic links in mount points in "cluster.conf" may have caused service
failures and warnings.
- the "monitor_link" parameter was not honored when using VIP.
- clurgmgrd crashed when deeply-nested resources were configured.
- clulog did not honor rgmanager log levels. For example, if the log level
was set to 7, running "clulog -s 7 [test]" did not produce a log message.
- bonding InfiniBand devices, or network devices that did not start with
"eth", failed with "ip.sh".
- in certain cases, "service rgmanager stop" did not stop the rgmanager
- multiple logical volumes per volume group may have caused commands such
as "lvs" to hang. Accessing certain logical volumes may have caused a
system hang. Now, multiple logical volumes are allowed per volume group, as
long as they are part of a single service.
As well, this updated package adds the following enhancements:
- support for broadcasting to rgmanagers has been added.
- if a HA LVM server loses its connection to a SAN, the service is
relocated to another machine.
- during service recovery, counters and thresholds are now reset before
switching to relocation.
- follow logic can now be configured. For example, if a server is running
on node A, and a replication service is running on node B, if node A fails,
the server service is started on node B, before being moved to another
- rgmanager now supports "hard" and "soft" dependencies. For example, if
service "A" has a "hard" dependency on service "B", service "A" must be
stopped if service "B" fails. As well, if service "A" has a "soft"
dependency on service "B", then service "A" must be started after service
"B", but would be able to recover by itself if service "B" failed.
- netfs.sh now supports CIFS file systems.
- a resource agent for Oracle Database 10g has been added.
- parameter descriptions for the LVM failover script have been updated.
Users of rgmanager are advised to upgrade to this updated package, which
resolves these issues and adds these enhancements.
Before applying this update, make sure that all previously-released
errata relevant to your system have been applied.
This update is available via Red Hat Network. Details on how to use
the Red Hat Network to apply this update are available at
- Red Hat Enterprise Linux High Availability for x86_64 4 x86_64
- Red Hat Enterprise Linux High Availability for x86_64 4 ppc
- Red Hat Enterprise Linux High Availability for x86_64 4 ia64
- Red Hat Enterprise Linux High Availability for x86_64 4 i386
- BZ - 182423 - Doc. Notes + resource agent for Oracle 10g Database Failover
- BZ - 242798 - RFE: If HA lvm server lost connection to SAN, relocate to next machine
- BZ - 245381 - [RFE] Restart counters before a switch to relocate.
- BZ - 247159 - VIP parameters 'monitor_link' not apply
- BZ - 247772 - RFE: One service following another
- BZ - 247945 - [RFE] rgmanager restart services when autostart option is changed
- BZ - 247980 - RFE: (strong and weak) service dependencies
- BZ - 250101 - RG event API on rgmanager (RHCS4)
- BZ - 250679 - RFE: Mounting samba share in netfs.sh script in RHEL4 CS
- BZ - 270421 - clurgmgrd crashes when resources are deeply nested
- BZ - 289521 - clulog doesn't honor cluster configuration specification of log level in all cases
- BZ - 358161 - a failover of a nfs mount point does not occur if there is a process holding the filesystem even when force_umount option is enabled
- BZ - 360401 - rgmanager is stuck in a loop while rebooting a node.
- BZ - 427377 - HA LVM should allow multiple LVs/VG as long as they move together
- BZ - 428475 - HA LVM service fails to relocate when I/O is running
- BZ - 431705 - HA LVM should prevent users from running an invalid setup (2)
- BZ - 438413 - Bond slaves other than eth* are ignored by rgmanager
- BZ - 438997 - HA LVM: lvm_by_vg.sh is stripping tags from all the VGs and not from only the configured one.
- BZ - 439948 - clushutdown references in man pages
- BZ - 440006 - rgmanager stuck on stop
- BZ - 440645 - service failover hangs at quotaoff in /usr/share/cluster/fs.sh
- BZ - 441577 - symlinks in cluster-managed mount points cause erroneous failures
- BZ - 443658 - Insufficient parameter descriptions in LVM script
- BZ - 448108 - rgmanager can get stuck forever with gulm
Red Hat Enterprise Linux High Availability for x86_64 4