[Linux-cluster] qdiskd Startup Questions

Caron, Chris caronc at navcanada.ca
Wed Sep 24 21:51:42 UTC 2008

Lon H. posted a question that sort of sparked a few questions from me
about Quorum Disks...


1.	Even if there is absolutely no configuration for a quorum disk
(yet), nor is one even created (through mkqdisk)... can the Quorum Disk
Daemon still run (at startup)? You might ask why I would even want to do
this. Simply because our company preloads a working cluster through a
kickstart and a series of scripts... it would just be one less command
to have to call on every single node (at the start) after creating the
partition and updating the cluster.conf tool.  Just having the daemon
run at the start would mean you could configure the Quorum Disk from
just one node. 

	a.	(Assuming you can run the quorum disk... it would be
running on all cluster nodes)
	b.	Create the quorum disk at one of the nodes
	c.	Update the cluster.conf (+version) to support it, and
then ccs_tool it.
	d.	(am I missing a step?)

2.	When you type mkqdisk -L after creating a Quorum disk.  Does the
node that created the quorum disk have any impact on the cluster at all?
I guess I get confused with creating a 10MB iscsi mount on all cluster
nodes, and only having to run the mkqdisk once on one of the nodes...
Does this mean if that node fails, the quorum disk is gone?  Or do all
nodes access the block device that has been made available to them too
(ie, all nodes see /dev/sdc, but mkqdisk -L says it was created on
node04)... Maybe my question should be: how is knowing who created the
Qdisk useful to me?
3.	Since there is all this talk about the Quorum Disk Daemon being
the last thing to start (after fencing).... Would it be feasible to just
make a failoverdomain per node and only consisting of 1 node to
exclusively start the Quorum Disk Daemon as a cluster service (per node)
(this site explains it better:
http://sources.redhat.com/cluster/wiki/MultipleInstanceServices)... That
way leaving or joining a node into the cluster starts and stops it's
daemon as needed.





-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/linux-cluster/attachments/20080924/a1e01dbf/attachment.htm>

More information about the Linux-cluster mailing list