Multipath paths fail due to SCSI reservation conflicts in a RHEL High Availability cluster configured with fence_scsi before cluster services have been started
Issue
- A database stopped on a Red Hat cluster. We are trying to determine root cause. We know that there is a problem with multipathing.
- I see path failures and reservation conflicts for devices that should be managed by the cluster and
fence_scsibefore I start up the cluster services - A cluster node prints SCSI reservation conflicts in the logs while booting up
- Every time a
multipathpath-checker runs, it triggers a reservation conflict because the cluster isn't up yet and hasn't registered the devices withfence_scsi. Aren't these path-checkers read-only? Why would it trigger a reservation conflict?
Environment
- Red Hat Enterprise Linux (RHEL) 5, 6, or 7 with the High Availability Add On
- One or more nodes configured with a fence device using
fence_scsi device-mapper-multipath- Path failures due to "reservation conflict" while the node is not a member of the cluster. If this issue presents itself while the node is a member, then this specific issue is not to blame.
Subscriber exclusive content
A Red Hat subscription provides unlimited access to our knowledgebase, tools, and much more.
