I'm running a corosync cluster to be used as an ISCSI target. It runs on openfiler. I followed a tutorial from http://www.howtoforge.com/openfiler-...maker-and-drbd
. The cluster failover works when a node fully fails (tested it with by using reboot -f).
But the failover doesn't work when I put the primary node in standby (crm resource standby node1) or when I stop the corosync service on the primary one. And I can't find why. Does anyone have an idea? Did I miss something?
Some things I checked/found out:
- The drbd setup is in sync so both should be able to run as primary.
- The lvm resource ends up as "unmanaged" when it put the primary in standby (so no failover!). This causes the other resources to stop.
- It can't really stop the corosync service correctly. When I call "service corosync stop" then the service continues to log (corosync.log) a message: "still waiting for crmd to terminate".
Looking forward to your suggestions about what's going wrong. Please let me know when you need more details.