[Linux-cluster] DLM locks forever

Bob Peterson rpeterso at redhat.com
Fri Aug 31 14:30:43 UTC 2012

----- Original Message -----
| Hi,
| I am trying to create a cluster the Red Hat style. I am using cman t
| ostart
| services like fenced, dlm_controld, gfs_controld and corosync.
| At the moment I do not have fencing implemented. I start the fenced
| with the
| option "-c" because we start the cluster only manually. Basically we
| want to
| use a GFS2 file system.
| Everything works fine up to this point.
| But in a split-brain situation both nodes wait for the fencing
| forever. This
| would be OK. The admin could poweroff one machine. But how can I
| release the
| lock on the other machine because it is still waiting for the success
| of the
| fencing.
| Is there any command that tells the fenced on the surviving machine
| that
| fencing was done manually and it can continue with its operation?
| Thanks for hints.
| --
| Dr. Michael Schwartzkopff
| Guardinistr. 63
| 81375 München
| Tel: (0163) 172 50 98
| Fax: (089) 620 304 13


What version are you using? It varies depending on the software.
In the older software, you used to have to do this:
fence_ack_manual -n <fenced_node>

In the newer cluster software, there's a concept of unfencing
and stuff that I haven't been following.


Bob Peterson
Red Hat File Systems

More information about the Linux-cluster mailing list