When starting clvmd with one or more nodes missing from the cluster, it fails to start with "Starting clvmd: clvmd startup timed out" in RHEL 6

Solution Unverified - Updated 2024-08-06T06:04:58+00:00 -

Issue

When starting clvmd after cman, with the other node missing, it fails to start with this message on the console:

Starting clvmd: clvmd startup timed out

clvmd timed out while starting if post_join_delay fencing had not yet completed
I have a large post_join_delay and sometimes my gfs2 filesystems do not mount on boot
I have a large post_join_delay and sometimes my clustered logical volumes are not active on boot

Red Hat Enterprise Linux (RHEL) 6 with the Resilient Storage Add On
lvm2-cluster (clvmd) in use on the cluster
Starting a portion of the cluster with one or more nodes not present
The amount of time it takes to fence a node is greater than than clvmd's startup timeout specified by -T (which defaults to 30 seconds). The clvmd -T timeout is either set at the default by the init script, or can be configured manually in /etc/sysconfig/cluster
- The length of time it takes to fence a node can be contributed to by:
  - post_join_delay in /etc/cluster/cluster.conf
  - A delay attribute set on a fencedevice in /etc/cluster/cluster.conf
  - Fence actions naturally taking a long time due to conditions on the device or settings that have been chosen

A Red Hat subscription provides unlimited access to our knowledgebase, tools, and much more.