[Linux-cluster] cluster.conf not updated

[Sven Carstens]

Hi all,

this is a ubuntu-8.10 dual node cluster.

Starting the cluster by rebooting both nodes works ok.

But when I update the cluster.conf with ccs_tool update the
update won't be noticed by cman until the next reboot.


[node1/2]cman_tool status:
---------------------------------------
Version: 6.1.0
Config Version: 28
Cluster Name: wmh_cluster
Cluster Id: 33204
Cluster Member: Yes
Cluster Generation: 168
Membership state: Cluster-Member
Nodes: 2
Expected votes: 1
Total votes: 2
Node votes: 1
Quorum: 1
Active subsystems: 12
Flags: 2node
Ports Bound: 0 11 177
Node name: wmh01
Node ID: X
Multicast addresses: 239.192.129.54
Node addresses: 192.168.123.X
---------------------------------------

[node1/2]ccs.log
---------------------------------------
[CCS  ] Starting ccsd 2.99.09:
[CCS  ]  Built: Aug 26 2008 14:44:29
[CCS  ]  Copyright (C) Red Hat, Inc.  2004-2008  All rights reserved.
[CCS  ] Unable to add to membership: No such device
[CCS  ] Initial status:: Quorate
---------------------------------------

Then I update the cluster.conf to a new version where only the config_version
value is incremented by one.

[node1] ccs_tool update /etc/cluster/cluster.conf_edit
---------------------------------------
Config file updated from version 28 to 29

Update complete.
---------------------------------------

CCS on Node1 logs the configuration change. Node2 doesn't.
---------------------------------------
[CCS  ] /etc/cluster/cluster.conf (cluster name = wmh_cluster, version = 28) found.
[CCS  ] Update of /etc/cluster/cluster.conf complete (version 28 -> 29).
---------------------------------------
On both nodes the file /etc/cluster/cluster.conf is updated to the
new version.


[node1/2]cman_tool status still shows the old version:
---------------------------------------
Version: 6.1.0
Config Version: 28
Cluster Name: wmh_cluster
Cluster Id: 33204
Cluster Member: Yes
Cluster Generation: 168
Membership state: Cluster-Member
Nodes: 2
Expected votes: 1
Total votes: 2
Node votes: 1
Quorum: 1
Active subsystems: 12
Flags: 2node
Ports Bound: 0 11 177
Node name: wmh02
Node ID: X
Multicast addresses: 239.192.129.54
Node addresses: 192.168.123.X
---------------------------------------

Trying to force the configuration_version change with
cman_tool version -r 29 doesn't change a thing.

After such an update the cluster is inoperable and clustat shows
---------------------------------------
Timed out waiting for a response from Resource Group Manager
Cluster Status for wmh_cluster @ Thu Dec 18 15:23:29 2008
Member Status: Quorate

 Member Name                                                     ID   Status
 ------ ----                                                     ---- ------
 wmh01                                                               1 Online, Local
 wmh02                                                               2 Online
---------------------------------------

I have a similar cluster up and running on ubuntu-8.04 which is running
ccsd 2.0 (built Feb 27 2008 06:23:41)

Any hints was is going wrong here ?

CU Sven