[Pacemaker] newbie question(s)

Alex Samad - Yieldbroker Alex.Samad at yieldbroker.com
Wed May 22 00:13:28 UTC 2013



> -----Original Message-----
> From: Florian Crouzat [mailto:gentoo at floriancrouzat.net]
> Sent: Tuesday, 21 May 2013 10:18 PM
> To: pacemaker at oss.clusterlabs.org
> Subject: Re: [Pacemaker] newbie question(s)
> 
> Le 21/05/2013 07:04, Alex Samad - Yieldbroker a écrit :
> > Hi
> >
> > I have setup a small 2 node cluster that we are using for HA a java app.
[snip]

> > is there anything I should be doing differently, I have seen colocation
> option and something about affinity of resources, but I used group which
> would be the best practise way of doing it ?
> >
> 
> Probably writing a better ocf script that doesn't rely on 'ps' (eww!) but on a
> real lsb 'status'.

So the test is a 2 part one, check the process is running and then make a call against the a status api.
There is a timeout period associated with the api call so I check first to see if its running

> 
[snip]
> 
> Maybe the user that runs pacemaker have the right to execute your 'ybrp'
> ocf script but not the 'ps' that you used in it ?

Don't think so as it runs fine and then occasionally fails 

> It returns 4 (<> 0), it should be quite easy to read
> /usr/lib/ocf/resource.d/yb/ybrp and search what would return 4 in the
> "status" section.

Oh silly me I was using the lib error codes an mistook 
$OCF_ERR_PERM as permanent not permissions !

But does this mean that pacemaker will not try and move the resource (IP) to the other node.

> 
[snip]

> >
> > Any help or suggestions muchly appreciated
> >
> 
> Also, fencing!

Not sure that I need it. The app is always running on both nodes, it's just the ip address that is shared

> 
> >
> > Thanks
> > Alex
> 
> 
> 
> 
> --
> Cheers,
> Florian Crouzat
> 
Thanks





More information about the Pacemaker mailing list