After a network split and recovery in a RHEL 5 or 6 cluster using fence_scsi, one node is killed and the remaining node cannot access storage devices
Issue
- In a two node cluster with
fence_scsi, if there is a a temporary network split, after recovery only one node remains in the cluster but it cannot access the storage devices - After a network issue in a two node cluster, the one remaining node is not registered to the SCSI devices any longer and gets SCSI reservation conflicts
- It seems that in a two node cluster if network connectivity is lost for a short time, the cluster ends up in an unworkable state.
Environment
- Red Hat Enterprise Linux (RHEL) 5 or 6 with the High Availability Add On
- Cluster configured to use
fence_scsi - Two-node cluster
Subscriber exclusive content
A Red Hat subscription provides unlimited access to our knowledgebase of over 48,000 articles and solutions.
Welcome! Check out the Getting Started with Red Hat page for quick tours and guides for common tasks.
