[ClusterLabs] thanks for your reply, but i have a question about Corosync + Pacemaker.

Digimer lists at alteeve.ca
Sat Oct 17 09:34:53 EDT 2015


Please ask questions on the cluster labs user's list instead of emailing
me directly. I often am not the best person to ask.

On 17/10/15 08:31 AM, Shilu wrote:
> Now I want to use Corosync + Pacemaker in iscsi target
> 
>  
> 
> iSCSI Active/Passive With Virtual IPs
> 
> Server1 accesses the two targets/GWs one at a time
> 
> through one or more Virtual IPs. The server1 use oracle.
> 
> If the active target goes down, corosync/pacemaker will
> 
> activate the passive target.

I've not built this kind of a cluster, but I can't think of reason it
would be a problem. Others here might have better advice.

> IPMI fencing, one of the most common types, takes a few seconds. 

That's normal.

> I feel that this time is too long,

There are other methods, like APC switched PDUs, that might be a bit
faster. Generally though, "a few seconds" is expected.

> Can it guarantee my business is not interrupted?

No.

Anyone who tells you they can guarantee no interruption is trying to
sell you something and not being honest about it. If you need better
than a generic HA stack can provide, you will probably need to build
something yourself using carrier-grade tools like the full AIS[1] API.

This would likely require a massive investment on your end, but the
openais plugin is available, though some things like checkpoint and
whatnot are not done (last I checked, which was a long time ago). Even
then though, you can have software bugs or other unexpected problems
take you offline.

Planes and spaceships crash, too. Nothing is perfect.

digimer

1.
http://www.saforum.org/page/16627~217404/Service-Availability-Forum-Application-Interface-Specification

-- 
Digimer
Papers and Projects: https://alteeve.ca/w/
What if the cure for cancer is trapped in the mind of a person without
access to education?




More information about the Users mailing list