[Linux-cluster] clvmd, can't see shared storage on node1 of 2-node cluster (long)

[Patrick Caulfeld]

linux-cluster at merctech.com wrote:
> I'm having trouble with a new two-node CentOS5 cluster (kernel 2.6.23). The clvm
> daemon doesn't run correctly on one node, and that node cannot see LVM objects,
> while it can connect to the underlying storage.
> 
> The cluster has shared SAN storage with 5 LUNs presented to the servers.
> Multiple HBAs and storage processes mean there are 4 paths to each LUN.
> DM-Multipath is used to create virtual devices for LUN access. Each server sees
> the LUNs at the SCSI level, and can see disk partitions via fdisk, and can see
> the same LUNs via multipath
> 
> The host-level storage configuration was done on node2, with text files being
> edited simultaneously on both nodes or copied from node2 to node1.
> 
> ### Error Condition ###
> The clvmd daemon does not run correctly on node1 of the two-node cluster.
> The error messages are:
> 
>   connect() failed on local socket: Permission denied

This is the important bit. LVM can't connect to clvmd because it is not
allowed to open the socket it communicates on.

Assuming this is all runs as root (and if not, there's your problem!)
it's worth checking out SELinux for reasons why the socket cannot be opened.


>   WARNING: Falling back to local file-based locking.
>   Volume Groups with the clustered attribute will be inaccessible.
> 



Patrick