Red Hat Bugzilla – Bug 299871
qdiskd node is erroneously considered a full cluster node
Last modified: 2009-04-16 18:51:42 EDT
From Bugzilla Helper:
User-Agent: Mozilla/5.0 (Macintosh; U; PPC Mac OS X; fr) AppleWebKit/419.3 (KHTML, like Gecko) Safari/419.3
Description of problem:
I am using a qdiskd node on a two node cluster. The qdiskd daemon run on the two nodes and the
qdiskd heuristic and master/slave process works correctly. I optain :
[root@titan1 ~]# clustat
msg_open: No such file or directory
Member Status: Quorate
Member Name ID Status
------ ---- ---- ------
titan1 1 Online, Local
titan2 2 Online
/dev/mpath/qdsk1 0 Online, Quorum Disk
The problem is that some parts of the cluster management and utilities think the Quorum Disk is a real
For example ccs_tool fails to send updated conf file to all members :
[root@titan1 ~]# ccs_tool update /etc/cluster/cluster.conf
Failed to receive COMM_UPDATE_COMMIT_ACK from /dev/mpath/qdsk1.
Hint: Check the log on /dev/mpath/qdsk1 for reason.
Failed to update config file.
lvm2 (and clvmd) fails to gain locking on lv :
[root@titan1 ~]# lvextend --size 200m /dev/san1/test
Extending logical volume test to 200,00 MB
clvmd not running on node /dev/mpath/qdsk1
Failed to suspend test
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1.configure a quorum disk partition on the cluster
2.use ccs_tool update or any lvm2 commands requiring clvmd to gain lock access
The commands fails or hang at waiting for the quorum node to respond to cman sollicitations.
The quorum node must be recognized as asuch and ignored by theses subsystems (ccs, clvmd).
I think there's a dup bug against lvm2-cluster which has already been fixed for
5.1; Patrick was working on it.
*** This bug has been marked as a duplicate of 237386 ***