[Linux-cluster] ccsd problems after update to RHEL 5.2/5.3

[Gordan Bobic]

Chrissie Caulfield wrote:
> Gordan Bobic wrote:
>> I have a two-node cluster and ever since I updated the kernel and
>> cluster components I cannot get more than one node running with GFS.
>>
>> Here are the package versions I have:
>> kernel-2.6.18-92.1.22.el5
>> cman-2.0.98-1
>> kmod-gfs-0.1.23-5.el5_2.4
>> gfs-utils-0.1.17-1.el5
>> gfs2-utils-0.1.53-1.1
>>
>> Node 2 starts up OK, but I see this in the syslog:
>>
>> node2 ccsd[5897]: Unable to perform sendto: Cannot assign requested address
>>
>> When I power up node2, it just gets strange and the whole thing locks up:
>> node2 openais[5941]: [CMAN ] cman killed by node 1 because we rejoined
>> the cluster without a full restart
>> node2 groupd[5953]: cman_get_nodes error -1 104
>> node2 gfs_controld[5995]: groupd_dispatch error -1 errno 11
>> node2 gfs_controld[5995]: groupd connection died
>> node2 gfs_controld[5995]: cluster is down, exiting
>>
>> So for some reason node 1's joining makes node 2 get kicked out of the
>> cluster - but worse, it doesn't seem to initiate fencing. Instead, the
>> whole cluster just locks up on GFS access.
>>
>> What am I missing? What should I be looking for in the logs? This
>> cluster worked fine before the update.
>>
>> I found this:
>> http://rhn.redhat.com/errata/RHBA-2009-0189.html
>> but updating cman to 2.0.98 as per the RHBA didn't fix the problem.
>>
> 
> it sounds like you've hit this bug:
> 
> https://bugzilla.redhat.com/show_bug.cgi?id=487397

What was the last known version of cman that works? 2.0.73?

Gordan