- Multiple parts of the system could notice a node failure at different times. Consequently, if more than one component requested a node to be fenced, the fencing components did so multiple times. This bug has been fixed by merging identical requests from different clients if the first one is still in progress, so the node is fenced only once.
- Canceled operations were incorrectly stored in the cluster status. As a consequence, the cluster detected those operations and tried to clarify the status that led to additional logging and other confusing behavior. The underlying code has been modified so that the canceled operations are no longer stored in the cluster status, and Pacemaker now works as expected.
- An improper definition in the spec file caused unexpected implicit dependencies between Pacemaker subpackages; a certain library was in the incorrect location. The libstonithd.so.2 library has been relocated and the dependencies between Pacemaker subpackages are now defined correctly.
- On the systems running on AMD64 or Intel 64 architectures, the pacemaker-cts subpackage depends on some libraries from the pacemaker.libs subpackage. However, pacemaker-cts did not specify explicit package version requirement, which could cause dependency problems between new and old subpackages. The version specification of pacemaker-libs has been added to pacemaker-cts to prevent these dependency problems.
- Previously, deleting a master or slave resource led to one of the nodes being fenced. This update applies a patch to fix this bug and nodes are no longer fenced in such a case.
- Previously, the crm_report package did not install the perl-TimeData package as a dependency. Consequently, an attempt to run the crm_report utility on a system without this package failed with an error. This update adds this missing dependency and the crm_report utility can now be run as expected.
- Previously, it was possible to introduce non-significant whitespace characters into the Pacemaker configuration file. Consequently, Pacemaker returned confusing error messages when reading the configuration file. With this update, a patch has been applied to filter the undesired characters from the configuration file and Pacemaker no longer returns such error messages.
- With this update, Pacemaker provides a simpler XML output, which allows the users easier parsing and querying of the status of cluster resources.
- With this update, Pacemaker indicates when a cluster resource is reported as running based on cached information about a node that is no longer connected.