[Pacemaker] Configuring a cluster for asymmetric operation

Fri Jun 8 15:45:59 UTC 2012

David, 

Just so I'm clearer (when I run into this issue...) 

You have to have all of the supporting software for each resource installed (i.e. drbd, nfs, mysql, or whatever) not just the RA's? 

If you have the RA's installed you'd get "not installed" but that is not sufficient for Pacemaker to assume it's not running? I would think that not installed would be can't be running but... ;-) 

Thanks! 

Jake 

----- Original Message -----

From: "David Vossel" <dvossel at redhat.com> 
To: "The Pacemaker cluster resource manager" <pacemaker at oss.clusterlabs.org> 
Sent: Friday, June 8, 2012 11:14:18 AM 
Subject: Re: [Pacemaker] Configuring a cluster for asymmetric operation 

----- Original Message ----- 
> From: "Matthew O'Connor" <matt at ecsorl.com> 
> To: "The Pacemaker cluster resource manager" <pacemaker at oss.clusterlabs.org> 
> Sent: Friday, June 8, 2012 8:36:13 AM 
> Subject: Re: [Pacemaker] Configuring a cluster for asymmetric operation 
> 
> I ran across the same issue when putting together an asymmetric 
> cluster.  Generally, I've found that it's necessary to make sure all 
> nodes have the same resource-related backing software installed.  For 
> instance, I too have had to install DRBD on all nodes, even those not 
> running DRBD.  iSCSITarget is another one - even with the RA 
> available, 
> not having the proper service available for the monitor to query can 
> cause a monitor error - which tends to nuke the resource on all 
> nodes, 
> even the properly configured ones.  As it was explained to me, 
> Pacemaker 
> needs to be able to know that a resource isn't (or is) running on a 
> node 
> that it isn't (or is) supposed to be running on.  Lack of resource 
> agent 
> (i.e. lack of ability to monitor) is not taken as an indication that 
> such a resource isn't running there. 

Yes, this is true. There is no way around this, pacemaker will probe every known resource on every node each time it starts up.  Without having the resource agents installed on all nodes, pacemaker can not verify a resource is running or not, which will cause problems.  Even if the configuration prevents a resource from ever being started by pacemaker on a node, pacemaker will still need a way to verify that resource isn't already running when it starts up. 

-- Vossel 

> On 6/8/2012 3:36 AM, Brad Jones wrote: 
> > I have a cluster with three nodes, A B and C.  A and B can 
> > basically 
> > stand in for one another; they run a DRBD master/slave set, and 
> > failure of A where most things run normally fires up services on B. 
> > 
> > C helps arbitrate quorum and runs Stonith plugins.  I've written 
> > location rules to prohibit most of the production services from 
> > running on node C. 
> > 
> > Problem is, the DRBD RA tries to monitor on node C, even though 
> > DRBD 
> > isn't even installed.  (It fails, helpfully, "not installed.")  I'm 
> > stuck with a slowly-incrementing failcount on C. 
> > 
> > When I give all services a default location rule and set 
> > symmetric-cluster="false", some of the resources stay running on A 
> > but 
> > most of the important ones stop. 
> > 
> > This was raised back in 2010 here: 
> > <http://oss.clusterlabs.org/pipermail/pacemaker/2010-April/005873.html> 
> > but I can't find much more about this issue on the web and there's 
> > only passing mention of it in the Pacemaker documentation. 
> > 
> > Bottom line: What's required to correctly configure a cluster with 
> > symmetric-cluster set to false? 
> > 
> > FYI I'm on pacemaker 1.1.6, heartbeat 3.0.5, on Ubuntu 10.04. 
> > 
> > Thanks!  Brad 
> > -- 
> > Brad Jones 
> > Brad at Jones.name 
> > Mobile: 303-219-0795 
> > 
> > _______________________________________________ 
> > Pacemaker mailing list: Pacemaker at oss.clusterlabs.org 
> > http://oss.clusterlabs.org/mailman/listinfo/pacemaker 
> > 
> > Project Home: http://www.clusterlabs.org 
> > Getting started: 
> > http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf 
> > Bugs: http://bugs.clusterlabs.org 
> 
> -- 
> 
> Sincerely, 
>   Matthew O'Connor 
> 
> ----------------------------------------------------------------- 
> Sr. Software Engineer 
> PGP/GPG Key: 0x55F981C4 
> Fingerprint: E5DC A0F8 5A40 E4DA 2CE6 B5A2 014C 2CBF 55F9 81C4 
> 
> Engineering and Computer Simulations, Inc. 
> 11825 High Tech Ave Suite 250 
> Orlando, FL 32817 
> 
> Tel:   407-823-9991 x315 
> Fax:   407-823-8299 
> Email: matt at ecsorl.com 
> Web:   www.ecsorl.com 
> ----------------------------------------------------------------- 
> 
> CONFIDENTIAL NOTICE: The information contained in this electronic 
> message is legally privileged, confidential and exempt from 
> disclosure 
> under applicable law. It is intended only for the use of the 
> individual 
> or entity named above. If the reader of this message is not the 
> intended 
> recipient, you are hereby notified that any dissemination, 
> distribution 
> or copying of this message is strictly prohibited. If you have 
> received 
> this communication in error, please notify the sender immediately by 
> return e-mail and delete the original message and any copies of it 
> from 
> your computer system. Thank you. 
> 
> _______________________________________________ 
> Pacemaker mailing list: Pacemaker at oss.clusterlabs.org 
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker 
> 
> Project Home: http://www.clusterlabs.org 
> Getting started: 
> http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf 
> Bugs: http://bugs.clusterlabs.org 
> 

_______________________________________________ 
Pacemaker mailing list: Pacemaker at oss.clusterlabs.org 
http://oss.clusterlabs.org/mailman/listinfo/pacemaker 

Project Home: http://www.clusterlabs.org 
Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf 
Bugs: http://bugs.clusterlabs.org 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.clusterlabs.org/pipermail/pacemaker/attachments/20120608/0dbd1a1e/attachment.htm>