Thanks Andrew, the Heartbeat ldirectord RA isn't the best, I've modified it for 
my purposes and will submit a patch to Linux HA.

Luke Bigum
Systems Administrator
 (p) 1300 661 668
 (f)  1300 661 540
(e)  lbi...@iseek.com.au<mailto:lbi...@iseek.com.au>
http://www.iseek.com.au<http://www.iseek.com.au/>
Level 1, 100 Ipswich Road Woolloongabba QLD 4102

[cid:image001.jpg@01CA4033.1BF03FC0]

This e-mail and any files transmitted with it may contain confidential and 
privileged material for the sole use of the intended recipient. Any review, 
use, distribution or disclosure by others is strictly prohibited. If you are 
not the intended recipient (or authorised to receive for the recipient), please 
contact the sender by reply e-mail and delete all copies of this message.


From: Andrew Beekhof [mailto:and...@beekhof.net]
Sent: Thursday 24 September 2009 4:51 PM
To: pacemaker@oss.clusterlabs.org
Subject: Re: [Pacemaker] cluster attempting to stop resources that aren't 
installed

The RA needs to return $OCF_ERR_INSTALLED instead of any other erorr code.
Check the configuration explained doc for what the error codes mean and how 
they're handled.
On Thu, Sep 24, 2009 at 7:53 AM, Luke Bigum 
<lbi...@iseek.com.au<mailto:lbi...@iseek.com.au>> wrote:

Hi all,



I have a problem with a four node asymmetric CentOS 5 pacemaker 1.0.5 / 
heartbeat 3.0.0 cluster.



Two nodes are for ldirectord, the other two for services, currently only DNS 
while testing. My problem is that CRM / LRM thinks that an 
ocf:heartbeat:ldirectord resource is running on the dns nodes and is attempting 
to stop / cleanup the resource when there isn't any ldirectord software 
installed or configured on these nodes. This fails, putting the resource in an 
unmanaged state. Reset fail counts  and clean it up again and it repeats the 
same thing - tries to stop the resource and goes unmanaged.



At the moment there's no location constraint so the resource shouldn't try run 
anywhere (asymmetric), however I did have one before with the same result: the 
cluster would attempt to do "something" with ldirectord on the DNS nodes and 
get itself unmanaged. This is the XML I had for that constraint:

<rsc_location id="bne-ldirectord-loc" rsc="bne-ldirectord">

<rule boolean-op="and" id="bne-ldirectord-loc-rule" score="+INFINITY">

<expression attribute="anycast-server-location" 
id="bne-ldirectord-loc-rule-anycast-server-location" operation="eq" 
value="brisbane"/>

<expression attribute="anycast-server-type" 
id="bne-ldirectord-loc-rule-anycast-server-type" operation="eq" value="lb"/>

</rule>

</rsc_location>



I've experimented with lrmadmin -D on the offending nodes but no success. 
Basically the only way I can return the cluster to a nicely managed state is to 
install ldirectord on the DNS nodes and touch an empty configuration file, but 
I don't really want to do this, especially if I plan on putting many different 
kinds of services on this cluster.



I vaguely remember reading about someone with a similar problem but I can't 
find that post. Attached is a hb_report if you'd like to peruse. Any help is 
greatly appreciated.



Thanks,



-Luke



Luke Bigum

Systems Administrator

 (p) 1300 661 668

 (f)  1300 661 540

(e)  lbi...@iseek.com.au<mailto:lbi...@iseek.com.au>

http://www.iseek.com.au<http://www.iseek.com.au/>

Level 1, 100 Ipswich Road Woolloongabba QLD 4102



[cid:image001.jpg@01CA4033.1BF03FC0]



This e-mail and any files transmitted with it may contain confidential and 
privileged material for the sole use of the intended recipient. Any review, 
use, distribution or disclosure by others is strictly prohibited. If you are 
not the intended recipient (or authorised to receive for the recipient), please 
contact the sender by reply e-mail and delete all copies of this message.





_______________________________________________
Pacemaker mailing list
Pacemaker@oss.clusterlabs.org<mailto:Pacemaker@oss.clusterlabs.org>
http://oss.clusterlabs.org/mailman/listinfo/pacemaker

<<inline: image001.jpg>>

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

Reply via email to