Slow kdump Startup Preventing Service Failover When Using fence_kdump
Issue
- Starting of the kdump takes > 2 minutes until the cluster gets a heartbeat and marks the node as fenced. The main time at starting the initrd is spent with loading 2 modules - cciss and microcode.
- The kdump init process is moving slowly while loading the required modules for kdump to function. This causes a delay in fencing and service failover.
Environment
- Red Hat Enterprise Linux (RHEL) 6
- HP CCISS Smart Array
Subscriber exclusive content
A Red Hat subscription provides unlimited access to our knowledgebase, tools, and much more.