Description of problem:
Edge case - creating/deleting cluster that includes luci node = ccsd caught in
loop on that node
This is probably an unsupported configuration - but can we do anythiung about
ccsd hanging? If not - then it's a release note entry.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. Create a new, multi-node cluster - include the node where luci is being run
in the cluster
2. After the cluster is (successfully) created, and all the nodes reboot, delete
the cluster from the luci/cluster tab.
3. The /etc/cluster/cluster.conf file is not deleted from the luci node
4. ccsd is caught in this loop on the luci node:
Jan 31 11:56:29 tng3-3 ccsd: Error while processing connect: Connection
Jan 31 11:56:34 tng3-3 ccsd: Cluster is not quorate. Refusing connection.
The error listed above.
This entry is written to the /var/lib/ricci queue on the luci node:
<batch batch_id="1884935388" status="2">
<module name="cluster" status="1">
<var mutable="false" name="cluster_shutdown"
<var mutable="false" name="purge_conf"
Luci can be run on one of managed nodes; some interuption of service should be
expected though, during node restarts.
Problem, described above, should not be caused by the fact that luci is running
on the node. It seems that problem lies somewhere else.
I have just tested cluster deployment/removal running luci on one of the nodes,
and it worked as expected. I used luci-0.8-30.el5.
Could you gather some more info, so we can get to the bottom of the problem.
Fixing Product Name. Cluster Suite was merged into Enterprise Linux for version
Closing this bug after being in NEEDINFO for 6 months