On 2008-08-20T18:33:47, Robert <[EMAIL PROTECTED]> wrote:
> Well, from my perspective - which of course is the perspective of a user
> and not the developer of heartbeat - it works. By "works" I mean, when I
> "pull" the cable between the datacenters, the resources are active on one
> node only (the node on which it will be active can not be told). When I put
> the cable back, the resource may switch sides, but is active on one node
> only. So the quorum server makes sure - in this special 2 node setup - that
> the ressource is active on one node only. In my world this means "it works"
> :)
It doesn't work if combined with fencing. Try it. It will fail. The
nodes will either block indefinitely (because they can't fence the other
side), or because they cannot/do not commit suicide, and if there are
stop failures, it will not be able to recover.
>> It will only defer to the quorum server in case majority is lost. If the
>> quorum server than grants quorum, while the other side still has
>> majority, both sides will have quorum.
> Why that ? The other node sees the node failure too and a new voting
> session is started - isn't it ? So if the heartbeat is lost, both nodes
> start to check the quorum, both fail with majority and ONE is granted
> quorum via quorumd. Isn't this as it is expeted work ? Isn't this what the
> quorum plugin architecture can be used for ?
This "works" only for 2 nodes.
>> (The above line "works" for the very special case of two nodes only, but
>> that's luck, not design.)
> Does "luck" mean, that this behaviour will change in future heartbeat
> releases OR that this kind of setup is limited to 2 node setups only ?
As I don't see anyone maintaining the quorum infrastructure in
heartbeat, it's unlikely to change. If it works for you, you're lucky.
>> DO NOT USE THE QUORUM SERVER.
> One last - and sorry for the pointed remark - why is it part of the
> official heartbeat distribution then (I mean the binary package, but the
> source) ?
Because Alan thinks (thought?) the quorum server worked, and he was the
maintainer of it. I maintain he's quite wrong on that count.
In particular the interaction between quorumd & fencing is broken, and
that it does not work for >2 nodes is a severe restriction.
Everyone is of course free to use it, but I won't have anything to do
with it.
Regards,
Lars
--
Teamlead Kernel, SuSE Labs, Research and Development
SUSE LINUX Products GmbH, GF: Markus Rex, HRB 16746 (AG Nürnberg)
"Experience is the name everyone gives to their mistakes." -- Oscar Wilde
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
