[Linux-cluster] Redhat 5.3 clvmd hang

[Christine Caulfield]

On 04/08/09 21:48, Curtis Collicutt wrote:
> Hi,
>
> I've been experiencing hangs with clvmd (and all associated commands: lvs, vgdsplay, ect) on a 3 node cluster.
>
> I am using cman and clvmd to create clustered lvs from a san backend.
>
> I recently upgraded from 5.2 to 5.3 because occasionally commands began to hang, and would not work until I rebooted either 1 or 2 of the nodes, and there seemed to be several fixes in 5.3.
>
> These are the versions of cman and lvm2-cluster I am using. I'm using the lvm2-cluster from centos, rather than redhat.
>
> [root at cluster6 ~]# cat /etc/redhat-release
> Red Hat Enterprise Linux Server release 5.3 (Tikanga)
> [root at cluster6 ~]# rpm -qa cman
> cman-2.0.98-1.el5_3.4
> [root at cluster6 ~]# rpm -qa lvm2-cluster
> lvm2-cluster-2.02.40-7.el5<- from centos packages
>
> and this is what nodes there are:
>
> [root at cluster6 ~]# cman_tool nodes
> Node  Sts   Inc   Joined               Name
>     1   M   2336   2009-08-04 14:39:42  cluster7
>     2   M   2336   2009-08-04 14:39:42  cluster5
>     3   M   2332   2009-08-04 14:39:19  cluster6
> [root at cluster6 ~]# cman_tool services
> type             level name     id       state
> fence            0     default  00010003 none
> [1 2 3]
> dlm              1     clvmd    00020003 none
> [1 2 3]
>
> and this is what's hanging after "service clvmd start"
>
>   7056 pts/0    S+     0:00 /usr/sbin/vgscan
>
> it just sits there forever, and I can't run any "lvm" type commands. If I kill it the clvmd init script moves onto vgchange -ayl which also hangs.
>
> Any thoughts? Is there some kind of human error misconfiguration that could be going on here?
>

If you have RHEL then I suggest you reinstall the RHEL packages and 
contact support. There is an outstanding BZ for some similar symptoms 
and I'm sure they'd be glad of some extra evidence.

Chrissie