[Pacemaker] Full API description for Fence Agent

Andrew Beekhof andrew at beekhof.net
Mon Jul 8 22:15:29 EDT 2013 

On 08/07/2013, at 5:31 PM, Andreas Mock <Andreas.Mock at web.de> wrote:

> Hi all,
> 
> after doing a bigger debugging session and reading the
> documentation more than once, I got the fence agent to
> work. In this case, the fence agent is a program which
> can be used by cman/fenced (RHEL cluster) and by
> pacemaker running in this environment as stonith device.
> 
> Only for completion:
> https://fedorahosted.org/cluster/wiki/FenceAgentAPI
> 
> My findings to get it work:
> a) Additionally (that was also said somewhere else before)
> the agent needs to implement the 'metadata' call which 
> prints a xml-document on STDOUT and was adapted from
> other scripted fence agents. I couldn't find a spec for
> that xml. I let this call also return 0.
> 
> b) Contrary to the spec above, stonith_ng seems to
> send the parameters 'nodename' and 'port'. As my
> stonith agent doesn't need that,

Do you have any parameter that is used to identify which node should be shot?
If so it should be called 'port' for consistency.  I agree that this is not 100% clear from the spec.

Also, unknown parameters should be ignored (logging an error is fine, just don't fail the operation unless a required parameter is missing)

> I've thrown an exception
> scanning these parameters which led to an error in the
> logs. => Now these parameters are valid even when not
> used.
> Someone should clarify how to react on that parameters
> correctly when not used. See c)
> 
> c) One thing I missed when configuring the stonith
> agent in pacemaker was the parameter 'pcmk_host_list'.
> Look here
> http://clusterlabs.org/doc/en-US/Pacemaker/1.1/html/Clusters_from_Scratch/_c
> onfiguring_stonith.html
> to see details. Therefor pacemaker couldn't know
> how to fence the node. This was also seen by
> issuing the command 'stonith_admin -l <nodename>',
> what I was wondering about before solving the problem.

It would be very useful to see the agent and config.
I'm glad you've got it working, but its hard to discuss whether an agent is correct/sane without knowing more about it.

> 
> Additions and corrections welcome for all
> fence agent programmers.
> 
> Best regards
> Andreas Mock
> 
> 
> 
> 
> -----Ursprüngliche Nachricht-----
> Von: Andrew Beekhof [mailto:andrew at beekhof.net] 
> Gesendet: Montag, 8. Juli 2013 05:27
> An: The Pacemaker cluster resource manager
> Betreff: Re: [Pacemaker] Full API description for Fence Agent
> 
> 
> On 04/07/2013, at 9:52 PM, Andreas Mock <Andreas.Mock at web.de> wrote:
> 
>> Hi Andrew,
>> 
>> is there some kind of agreement how to tag a message?
>> Like (DEBUG/TRACE/ERROR/WARN)?
> 
> No.  But pacemaker obeys the general convention of "errors to stderr,
> everything else to stdout".
> 
>> Is there a way message level filtering is done?
> 
> There is no filtering.
> 
>> 
>> Best regards
>> Andreas
>> 
>> 
>> -----Ursprüngliche Nachricht-----
>> Von: Andrew Beekhof [mailto:andrew at beekhof.net] 
>> Gesendet: Donnerstag, 4. Juli 2013 13:41
>> An: The Pacemaker cluster resource manager
>> Betreff: Re: [Pacemaker] Full API description for Fence Agent
>> 
>> 
>> On 04/07/2013, at 7:24 PM, Andreas Mock <Andreas.Mock at web.de> wrote:
>> 
>>> Hi digimer,
>>> 
>>> I would like to take your offer and asking the following:
>>> 
>>> The API documents says nothing about the correct way
>>> of giving messages back to the stonith daemon.
>>> So, what is the right way to write error/warn/info messages.
>>> 
>>> Looking at the scripted agents available I can find a nice
>>> mixture of using STDERR and STDOUT.
>>> What is the rule here?
>>> Can you give insights, whether STDOUT/STDERR is captured by
>>> the calling program and logged somewher (and where)?
>> 
>> In the case of pacemaker, we capture and log both.
>> 
>>> 
>>> By the way: How is it going with merging the stonith/fencing API? ;-)
>>> 
>>> Best regards
>>> Andreas
>>> 
>>> -----Ursprüngliche Nachricht-----
>>> Von: Digimer [mailto:lists at alteeve.ca] 
>>> Gesendet: Dienstag, 11. Juni 2013 15:34
>>> An: The Pacemaker cluster resource manager
>>> Cc: Andreas Mock
>>> Betreff: Re: [Pacemaker] Full API description for Fence Agent
>>> 
>>> Hi Andreas,
>>> 
>>> The metadata section of the document has not been added yet, but we 
>>> are aware of it missing and are working to add it. The rest of the 
>>> document is accurate though. If you build an agent to follow that API, 
>>> it will work with red hat's cluster and pacemaker.
>>> 
>>> In the meantime, it's not ideal, but if you call any other fence 
>>> agent and pass '-o metadata', you will see the output that the cluster 
>>> expects. It should be easy to adapt to your new agent.
>>> 
>>> If you have any trouble, please don't hesitate to ask here and we 
>>> will do our best to help.
>>> 
>>> digimer
>>> 
>>> On 06/11/2013 07:04 AM, Andreas Mock wrote:
>>>> Hi all,
>>>> 
>>>> we need to implement a fence_agent (stonith agent) for
>>>> cman/corosync/pacemaker (RHEL 6.x). I found the following documentation
>>>> https://fedorahosted.org/cluster/wiki/FenceAgentAPI
>>>> 
>>>> But in this document the required metadata action is not
>>>> described. Can anybody point me to a documentation which
>>>> is complete?
>>>> 
>>>> Where is the schema of the xml returned by 'metadata'?
>>>> 
>>>> What has to be done that a fence_agent can also be used
>>>> by pacemaker?
>>>> 
>>>> What is the right return code of action 'metadata'?
>>>> 
>>>> Is there some explanation how the stonith/fence parts
>>>> play together?
>>>> 
>>>> Best regards
>>>> Andreas Mock
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>>> _______________________________________________
>>>> 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
>>>> 
>>> 
>>> 
>>> -- 
>>> 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?
>>> 
>>> 
>>> _______________________________________________
>>> 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
>> 
>> 
>> _______________________________________________
>> 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
> 
> 
> _______________________________________________
> 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

More information about the Pacemaker mailing list