Hide Forgot
Description of problem: When Pacemaker fences a node, it tries to let cman/fenced know. However if /var/run/cluster/fenced_override is not yet open for reading, our call to open will block until it is. This could be forever. Version-Release number of selected component (if applicable): pacemaker-1.1.5-8.el6 How reproducible: Very Steps to Reproduce: 1. Configure a 7 node physical cman/pacemaker cluster 2. Run CTS with fencing enabled 3. Watch things blow up during the initial StartOnebyOne test Actual results: The cluster logs will contain messages that "cluster nodes failed to respond to the join offer" and "Welcome reply not received from". Attaching to any of these nodes with gdb will show them blocked in tengine_stonith_notify() trying to open /var/run/cluster/fenced_override Expected results: The error logic is not triggered. Additional info:
Technical note added. If any revisions are required, please edit the "Technical Notes" field accordingly. All revisions will be proofread by the Engineering Content Services team. New Contents: The Pacemaker Cluster Resource Manager has been rebased to version [X] from upstream. This includes fixes from the open source community. Pacemaker continues to be offered as a Technology Preview.
Technical note updated. If any revisions are required, please edit the "Technical Notes" field accordingly. All revisions will be proofread by the Engineering Content Services team. Diffed Contents: @@ -1 +1 @@ -The Pacemaker Cluster Resource Manager has been rebased to version [X] from upstream. This includes fixes from the open source community. Pacemaker continues to be offered as a Technology Preview.+The Pacemaker Cluster Resource Manager has been rebased to version 1.1.6 from upstream. This includes fixes from the open source community. Pacemaker continues to be offered as a Technology Preview.
solved (probably) by this patch: https://github.com/ClusterLabs/pacemaker/commit/cd2d55f68642c86afe157e836811aa0239c974c1
Technical note updated. If any revisions are required, please edit the "Technical Notes" field accordingly. All revisions will be proofread by the Engineering Content Services team. Diffed Contents: @@ -1 +1 @@ -The Pacemaker Cluster Resource Manager has been rebased to version 1.1.6 from upstream. This includes fixes from the open source community. Pacemaker continues to be offered as a Technology Preview.+Under certain circumstances, an attempt to fence a node may have caused Pacemaker to stop responding when accessing the /var/run/cluster/fenced_override file. With this update, this error no longer occurs, and Pacemaker now works as expected in this scenario.
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. http://rhn.redhat.com/errata/RHBA-2011-1669.html