[Linux-HA] Is this even possible with heartbeat?

[Andrew Beekhof]

On Tue, Apr 22, 2008 at 8:38 PM, Javier Fox <javier.fox at zonarsystems.com> wrote:
> Greetings,
>
>  I am currently researching ways to implement a somewhat advanced cluster
> scheme, and I can't seem to find any evidence of anyone else attempting a
> similar setup.  I'm curious if what I'm looking to do is even possible with
> heartbeat.
>
>  The general idea is to have a cluster of 7 servers, sharing 5 sets of
> resources - so, 5 active servers and a 'pool' of 2 idle servers.

no problem so far

>  The
> cluster would be set up in such a way that if any of the 5 active servers is
> put out of commission, one of the idle servers can take up its resources,
> and, once the out-of-commission server is repaired, it joins the idle pool.

no problem
default-resource-stickiness = {some high number}

>  Conversely, no active server must ever take more than one set of resources
> - that is to say, if Server-001 and Server-002 are active, and Server-002
> dies, Server-001 should ignore this fact and let one of the idle servers
> worry about it.

yep
rsc_colocation constraints with score = -INFINITY between the resources

>
>  For the sake of completeness, the resources involved would be IP addresses
> and remotely-mounted disks (NFS/iSCSI), one each per 'set of resources'.
>
>  So, first off, does this plan make any sense?  And more importantly, is
> this something that could be accomplished with heartbeat?  Any suggestions
> or advice would be greatly appreciated.

Recommended reading:
   http://clusterlabs.org/mw/Image:Configuration_Explained.pdf