[ClusterLabs] Antw: Re: SBD & Failed Peer

Jorge Fábregas jorge.fabregas at gmail.com
Mon Sep 7 13:01:18 UTC 2015


On 09/07/2015 02:20 AM, Ulrich Windl wrote:
> If a node configured to use SBD can't read the disk, it can't do SBD. To be
> safe, the node must shut down. 

Hi,

Yes but if cluster communications are still going on I find it harsh to
self-fence.  I get what you mean but at least I think this should be up
to the administrator; a configurable parameter (and that's what they've
done in SUSE with the -P option).


> However you can have redundant SBD disks; did you know?

Yes, I read about that one but I believe that if my SBD disk fails , the
rest might fail as well since they all come from the same SAN & same
storage-array.

What I'm debating in my head is whether to use a small partition within
one of the production LUNs (LUNs shared with the other node for
application data) or if I should request a separate small LUN just for SBD.

Thanks,
Jorge




More information about the Users mailing list