[ClusterLabs Developers] shutdown testing in pacemaker corosync cluster

[Ken Gaillot]

On Tue, 2020-12-01 at 09:14 +0000, Harishkumar Pathangay wrote:
> Hi,
> Thanks for your valuable reply.
> I will try setting resource stickiness and test out and mail to 
> users at clusterlabs.org mailing list.
> Very happy to receive reply from you. I am very new, to the extent
> that I do nor even know which mailing list to use.
> Please embrace me in pacemaker user team.
>  
> Thanks,
> Harish P

Hi Harish,

New users are the lifeblood of the project -- you are definitely
welcome here! Clustering is a large topic, and all of us have more to
learn.

>  
> Sent from Mail for Windows 10
>  
> From: Ken Gaillot
> Sent: 01 December 2020 03:01
> To: developers at clusterlabs.org
> Subject: Re: [ClusterLabs Developers] shutdown testing in pacemaker
> corosync cluster
>  
> On Sun, 2020-11-29 at 14:21 +0000, Harishkumar Pathangay wrote:
> > Hi,
> > I am new to pace maker and learnt lot of things from documentation
> > guide
> > Pacemaker-2.0-Pacemaker_Administration-en-US.pdf
> > Pacemaker-2.0-Pacemaker_Development-en-US
> >  
> > I was able to successfully create a db2 resource agent which will
> > start instance when cluster starts.
> > Things are not production ready, but I can reasonably say script is
> > working in a 2-node cluster.
> > DB2 will run in node dragon initially. But if one node goes down it
> > will start db2 on other node tiger.
> > Tow Nodes are Linux RHEL 7.8 – DRAGON and TIGER.
> >  
> > If I shutdown one node [say tiger] by using shut down command as
> > root, the resources are moving to the other node [say dragon],
> which
> > is good.
> > Now I power back on the node which was shut down earlier that is
> > tiger node.
> > In pcs status I can see that the node tiger is offline.
> > How to bring it online?
> 
> Hi Harishkumar,
> 
> You will get better answers from the users at clusterlabs.org mailing
> list
> for this type of question. This list is for discussing software
> development of ClusterLabs projects.
> 
> I will try to answer though:
> 
> > If I try by “pcs cluster start” on node tiger, it is getting
> online 
> 
> Correct. You can either enable Pacemaker to start at system boot
> (systemctl enable pacemaker), or leave it so that you have to
> manually
> start it after a reboot.
> 
> It's a matter of personal preference, whether you lean more to
> automatic recovery, or investigating what happened before returning a
> node back to the cluster.
> 
> > no issues there, but there is a restart of db2 in dragon node. Why
> > this is happening?
> 
> Without seeing the configuration, it's hard to say. It sounds like
> you
> want a positive resource-stickiness, though, so it would be worth
> setting that in resource defaults if you haven't already.
> 
> (Any follow-ups should go to the users list.)
> 
> > My expectation is if a node goes down and the resource is moved to
> > other node, after that original nodes comes back, I do not want
> > pacemaker to restart things or resources on existing node that is
> > dragon in my case.
> > How will I configure this?
> >  
> > Any help will be of great significance to me. I am creating lot of
> > db2 educational contents. Precisely this will be helpful for my
> next
> > video.
> > Please let me know any further information required from my side.
> >  
> > Thanks,
> > Harish P
> > YouTube.com/db2luwacademy
> >  
> > Sent from Mail for Windows 10
> _______________________________________________
> Manage your subscription:
> https://lists.clusterlabs.org/mailman/listinfo/developers
> 
> ClusterLabs home: https://www.clusterlabs.org/
-- 
Ken Gaillot <kgaillot at redhat.com>