New configuration is not applied by rgmanager or displayed by clustat after fencing failed at some point in the past in RHEL 5
Issue
clustatdoesn't show changes applied tocluster.confon all nodesrgmanageris not applying configuration changes- No cluster configuration refresh after the creation of a new resource.
- We have created a new resource and updated the configuration with the command
ccs_tool update /etc/cluster/cluster.conf, but after the update we only see the new resource in the node where we executed the command. - We added a service in
/etc/cluster/cluster.conf, incremented the cluster version number, did accs_tool update /etc/cluster/cluster.conf...After this thecluster.confwas updated on both nodes (checked withcman_tool status- ok); but the service is not recognized on one node and theclurgmgrddoes not check any more the services. - A node rebooted and a node outside the
failoverdomaintook it over. When the node came back into the cluster, the service did not failback to either of the higher priority nodes in thefailoverdomain, but stayed outside the domain. It was later discovered from anrgmanager-debugdump that the node which took over the service was still using an old config version which had differentfailoverdomains, even thoughcmanshowed as using the latest version.
Environment
- Red Hat Enterprise Linux (RHEL) 5 with the High Availability Add On
rgmanagerprior to release2.0.52-47.el5- One or more nodes required fencing at some point in the past, fencing failed, and the node eventually rejoined the cluster without being fenced
rgmanagerwas never restarted on one or more nodes since that rejoin
- The configuration version reported by
cmanandrgmanager(clurgmgrd) are different (see Diagnostic Steps)
Subscriber exclusive content
A Red Hat subscription provides unlimited access to our knowledgebase, tools, and much more.