3.7. cman

Bug Fixes

BZ#1043484

Due to a regression, the fence_bladecenter fence agent was not able to login to fence devices. The login code for fencing.py has been fixed, and a fence agent now logins to fence devices as expected.

BZ#1043358

Previously, under some circumstances on a two-node cluster, both cluster nodes could be granted a lock on the same file at the same time on a GFS2 (Global File System 2) file system. A patch has been provided to fix this bug. With this update, only one cluster node is able to lock the file, the other cluster node waits till the lock is released.

Bug Fix

BZ#1021811

Previously, the fence_cisco_ucs fence agent did not respect the "delay" attribute, which could lead to the situation in which two nodes fence each other. With this update, a fence agent waits for given amount of time as expected.

Bug Fix

BZ#1073821

Previously, the fence_vmware_soap fence agent did not respect the "delay" attribute, which could cause a situation where two nodes fenced each other at the same time. With this update, the agent waits for the given amount of time as expected, and race conditions no longer occur.

Bug Fixes

BZ#994187

Due to incorrect quorum disk configuration, some cluster nodes could display an incorrect "Member status" value if the cluster was restarted by the qdisk quorum daemon. This update amends the quorum disk configuration and cluster nodes now display the correct status in the scenario described.

BZ#994228

Previously, the fence_cisco_ucs fence agent did not respect the "delay" attribute, which could lead to a situation in which two nodes fenced each other. This bug has been fixed and fence_cisco_ucs now waits for the given amount of time as expected.

BZ#1017916

A node on a two-node cluster could, under some circumstances, fail to unlink a checkpoint. As a consequence, both cluster nodes on the cluster could simultaneously be granted a lock on the same file located on a GFS2 file system. With this update, checkpoints on two-node clusters are unlinked as intended, and only one cluster can now hold a lock to a file at a time.

BZ#1029191

Due to a configuration error, qdisk in some situations used an incorrect tko parameter for its wait period when initializing. Consequently, qdisk initialization could be significantly delayed and, under some circumstances, it failed entirely. With this update, the cluster configuration file has been amended and qdisk initialization now proceeds as expected.

BZ#1040099

Due to errors in the login code for the fencing.py file, the fence_bladecenter fence agent was not able to log into fence devices. The login code for fencing.py has been fixed, and fence_bladecenter now logs into fence devices as expected.

BZ#1075691

Prior to this update, fence agents that connected using the SSH protocol failed on login if an identity file was used as the method of authentication. The bug has been fixed and these fence agents now successfully authenticate using an identity file.

Enhancement

BZ#1072605

The support for the "--delay" option has been added to the fence_vmware_soap fencing agent. This option allows the user to avoid fence races, and thus prevents nodes from potentially fencing each other.