On Fri, Aug 18, 2006 at 08:10:20AM +0200, Andrew Beekhof wrote:
>
> And therein lies the problem. The user is not being consistent in
> what they're asking for.
>
> On the one hand they say: 1) A needs B and 2) make sure A is always
> running.
> And on the other they say: 3) Ignore B and anything that happens to it.
At least in my case, this does not hold. There's no A and B is a
group:
Resource Group: a3
IPaddr_10_1_1_24 (heartbeat::ocf:IPaddr2): Started sapcl02 (unmanaged)
IPaddr_192_168_1_24 (heartbeat::ocf:IPaddr2): Started sapcl02 (unmanaged)
apache_a3 (heartbeat::ocf:apache): Started sapcl02 (unmanaged)
FAILED
a3 was put into unmanaged mode even though only one resource was
being managed (apache in this case).
> Either 1) is false, tell us not to observe 2), or change 3) to A + B.
> This last option is what Lars is contemplating having the cluster do
> automatically as a feature.
What is A+B?
> In the absence of any information about B, a failure in A requires a
> crystal-ball to handle correctly.
Nobody in their right mind could expect the cluster to be able to
manage this. Since we have no crystal ball at hand we can't handle
it in a meaningful way.
Actually, _all_ resources depending on the one put into unmanaged
mode should become unmanaged as well. That seems to me the only
reasonable way to deal with it.
Cheers,
Dejan
_______________________________________________________
Linux-HA-Dev: Linux-HA-Dev@lists.linux-ha.org
http://lists.linux-ha.org/mailman/listinfo/linux-ha-dev
Home Page: http://linux-ha.org/
