RHBA-2011:0537 - Bug Fix Advisory

Topic

Updated cluster and gfs2-utils packages that fix various defects are now
available for Red Hat Enterprise Linux 6.

Description

The cluster packages contain the core clustering libraries for Red Hat High
Availability as well as utilities to maintain GFS2 file systems for users of Red
Hat Resilient Storage.

This update provides fixes for the following defects:

• CMAN init script race condition has been fixed (BZ#595725)
• plock owner synchronization has been fixed (BZ#617306)
• plocks are now ignored until they written to their checkpoint (BZ#623810)
• plock signatures are now re-sent after a new totem ring forms (BZ#623816)
• post_join_delay now works after a loss and subsequent regain of quorum

(BZ#624844)

• "service cman stop remove" now functions correctly (BZ#634718)
• Active cluster nodes with higher configuration version numbers are no longer

killed when they join the cluster (BZ#639018)

• The ccs_tool man page no longer shows 'update' and 'upgrade' subcommands

(BZ#577874)

• ccs_tool cluster configuration editing has been dropped (BZ#614885)
• The interaction between corosync and cman restarting independently of one

another has been improved (BZ#617234)

• reporting of corosync's exit code has been improved (BZ#617247)
• cman_tool manual page no longer talks about "config version" as an argument to
• r (BZ#619874)
• Qdiskd now stops voting & exits if removed from the configuration (BZ#620679)
• gfs_controld: fix plock owner in unmount (BZ#624822)
• Qdiskd now reports to users when the quorumd "label" attribute overrides the

"device" attribute (BZ#635413)

• Qdiskd now has a hard limit on heuristic timeouts (BZ#636243)
• Pacemaker-specific versions of dlm_controld and gfs_controld have been removed

since they are no longer required (BZ#649021)

• cman now allows users to select udpu (UDP unicast) corosync transport

mechanism (BZ#657041)

• Qdiskd now assumes votes for each cluster node are 1 when not specified in

cluster.conf (BZ#663433)

• The cman init script can no longer include an incorrect sysconf file

(BZ#669340)

• DLM recovery no longer hangs after fencing completion in some instances

(BZ#678704)

• The cluster.rng schema has been updated (BZ#645830, BZ#618705, BZ#684020,

BZ#629017, BZ#680172)

• A memory leak in the XML parser has been fixed (BZ#680155)
• Heuristic checks are unreliable (BZ#688154)
• cman quorum timeout is too short (BZ#688201)

The following issues have been addressed in the gfs2-utils package:

• fsck.gfs2 was truncating directories with more than 100,000 entries

(BZ#628013)

• fsck.gfs2 was processing some files twice (BZ#621313)
• fsck.gfs2 no longer crashes if journals are missing (BZ#622576)
• When mounting a gfs2 file system, the same device requested on the command

line now appears in /proc/mounts and /etc/mtab (BZ#632595)

• gfs2_convert now resumes after an interrupted conversion (BZ#637913)
• fsck.gfs2 can now repair rgrps resulting from gfs_grow->gfs2_convert

(BZ#576640)

• mkfs.gfs2 no longer segfaults with 18.55TB and -b512 (BZ#624535)
• mkfs.gfs2 now supports discard request generation (BZ#656956)
• fsck.gfs2: reports master/root inodes as unused and fixes the bitmap

(BZ#663037)

• gfs2_convert no longer corrupts the file system if the di_height is too large.

(BZ#630005)

• gfs2_convert no longer exits success without doing anything (BZ#688734)

In addition, this update provides the following enhancements:

• gfs2_edit now outputs hexadecimal values in lower-case (BZ#634623)
• gfs2_edit now prints continuation blocks (BZ#634623)
• gfs2_edit's savemeta and restoremeta functions now report progress (BZ#634623)
• gfs2_edit has improved handling of corrupt file systems and enhanced usability

(BZ#674843)

• It is now possible to prevent the cluster software from starting at boot using

the kernel command line (BZ#563901)

• The cluster now has a script which can be used with the watchdog daemon to

reboot the host (BZ#560700)

• Fenced now sends notifications over DBus (BZ#592964)

All users of Red Hat High Availability and Red Hat Resilient Storage are advised
to upgrade to these updated packages, which resolve these issues.

Solution

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
https://access.redhat.com/kb/docs/DOC-11259

Affected Products

• Red Hat Enterprise Linux High Availability for x86_64 6 x86_64
• Red Hat Enterprise Linux High Availability for x86_64 6 i386
• Red Hat Enterprise Linux Resilient Storage for x86_64 6 x86_64
• Red Hat Enterprise Linux Resilient Storage for x86_64 6 i386
• Red Hat Enterprise Linux High Availability (for RHEL Server) from RHUI 6 x86_64
• Red Hat Enterprise Linux High Availability (for RHEL Server) from RHUI 6 i386
• Red Hat Enterprise Linux Resilient Storage (for RHEL Server) from RHUI 6 x86_64
• Red Hat Enterprise Linux Resilient Storage (for RHEL Server) from RHUI 6 i386

Fixes

• BZ - 595456 - Full support/testing for RHCS on RHEV guests
• BZ - 595725 - cman init script is not consistent in checking daemons startup and introduces possible race conditions
• BZ - 614885 - Deprecation of ccs_tool cluster.conf creation/editing capability
• BZ - 617234 - Starting or stopping corosync blocks cman from starting or stopping - cman part
• BZ - 617247 - cman_tool - display meaningful translation of corosync exit code on exit
• BZ - 617306 - dlm_controld: fix plock owner syncing
• BZ - 618705 - Update schema to pull in resource and fencing agent changes
• BZ - 619874 - Documentation: man page update, cman_tool
• BZ - 620679 - qdiskd should stop voting if no <quorumd config is available
• BZ - 621313 - GFS2: fsck.gfs2 seems to process large files twice
• BZ - 622576 - fsck.gfs2 segfaults if journals are missing
• BZ - 623810 - dlm_controld: ignore plocks until checkpoint time
• BZ - 623816 - dlm_controld: fix plock signature in stored message
• BZ - 624535 - GFS2: mkfs.gfs2 segfaults with 18.55TB and -b512
• BZ - 624822 - gfs_controld: fix plock owner in unmount
• BZ - 624844 - fenced post_join_delay doesn't help after quorum is restored
• BZ - 628013 - fsck.gfs2 truncates directories with more than 100,000 entries
• BZ - 629017 - fence_egenera missing options in cluster.rng
• BZ - 630005 - gfs2_convert: corrupts file system when directory has di_height 3
• BZ - 632595 - Updating /proc/mounts and /etc/mtab with mount args for GFS2 fs
• BZ - 634623 - GFS2: gfs2_edit enhancements for RHEL6.1
• BZ - 634718 - cman stop remove broken
• BZ - 635413 - quorum label overriding quorum device not made clear to user
• BZ - 636243 - qdisk should have a timer for heuristic "programs" instead of relying on "program" to provide timeout
• BZ - 637913 - gfs2_convert doesn't resume after interrupted conversion
• BZ - 639018 - [RHEL6] active cluster nodes with higher config version gets killed
• BZ - 645830 - Attributes missing from cluster.rng for the quorumd element
• BZ - 649021 - Drop -pcmk versions of dlm_controld and gfs_controld
• BZ - 656956 - GFS2: [RFE] mkfs.gfs2 should support discard request generation
• BZ - 657041 - Cman doesn't allow user to select udpu (UDP unicast) corosync transport mechanism
• BZ - 659123 - GFS2: gfs2_grow fails on a full file system
• BZ - 663037 - fsck.gfs2: reports master/root dinodes as unused and fixes the bitmap
• BZ - 663433 - qdiskd should assume votes="1" when not present for a node
• BZ - 669340 - cman init script could theoretically include wrong sysconf file
• BZ - 674843 - GFS2: gfs2_edit enhancements for RHEL6.1
• BZ - 678704 - dlm recovery did not complete after fencing
• BZ - 680155 - Memory leak in XML parser when reload config operation done
• BZ - 684020 - Add totem/ttl option to cluster relax-ng schema
• BZ - 688154 - Heuristic checks are unreliable
• BZ - 688201 - cman quorum timeout is too short
• BZ - 688368 - Support RHEL HA on VMware with iSCSI based storage + fence_scsi
• BZ - 688373 - Support RHEL HA on VMware with raw device mapping (physical mapping) based storage + fence_scsi
• BZ - 688395 - Support RHEL HA on RHEV 2.2 with iSCSI based storage + fence_scsi
• BZ - 688734 - gfs2_convert exits with success without doing anything

CVEs

(none)

References

(none)