pacemaker-1.1.18-11.el7.2

The Pacemaker cluster resource manager is a collection of technologies working together to maintain data integrity and application availability in the event of failures.

This update fixes the following bugs:

* Previously, Pacemaker scheduled notifications for clone actions which had not actually been run, but were just implied by fencing of the underlying node. In situations with a large number of such actions, these notifications led to long recovery times, saturation of the messaging mechanisms, loss of messages, and other consequential errors when using, for example, the bundle feature. Pacemaker no longer schedules notifications for implied events. As a result, recovery proceeds quickly after failure of a node hosting bundles with clone resources. (BZ#1563345)

* In certain situations, Pacemaker cleared the fail counter of a remote node before the node was successfully fenced. Thus, a subsequent reconnect failed when the attempt was tried between error cleaning and the actual fencing. A very low value set in the reconnect_interval option increased the chance of this situation. Consequently, Pacemaker was unable to reconnect to the remote host. With this update, Pacemaker now delays resource failure time outs until pending fencing actions are completed. As a result, the described scenario no longer appears and remote nodes are correctly recovered. (BZ#1566533)

* Previously, when using the crm_diff utility to compare two Cluster Information Base (CIB) files, crm_diff also reported differences in the order of XML attributes. As a consequence, changes that do not affect the operation of the cluster were always reported. This update adds the --cib option to the utility to ignore differences in the attribute order and to display only significant changes. For backward compatibility, when using the utility without the new option, crm_diff still displays all differences in CIB files. (BZ#1568720)

* To ignore the result of clone notify actions, Pacemaker reports them always as successfully completed. However, when the record-pending setting was set to true, which is the default, Pacemaker incorrectly reported pending clone notifications as successfully completed which led to "old event" warnings and unnecessary transition restarts. With this update, Pacemaker only records completed, not pending, clone notify actions as successful. As a result, the correct ordering of actions is preserved if the record-pending setting is enabled. (BZ#1570618)

Users of pacemaker are advised to upgrade to these updated packages, which fix these bugs.