Automation with Power

Power Business Continuity and Automation

Connect, learn, and share your experiences using the business continuity and automation technologies and practices designed to ensure uninterrupted operations and rapid recovery for workloads running on IBM Power systems. 


#Power
#TechXchangeConferenceLab

 View Only
Expand all | Collapse all

How to determine that a failover has happened

  • 1.  How to determine that a failover has happened

    Posted Tue May 19, 2009 06:48 AM

    Originally posted by: Holgervk


    Hello,

    does anybody have an idea how one could determine / be notified that a failover happened?

    Its easy to be notified of a manual rg move - there´s the rg_move event where one can add its notification.

    However, I dont see how to distinguish between a failover that happened because the other node crashed and the case where a RG was taken online because its node rebootet and cluster services have been started again.
    #PowerHA-(Formerly-known-as-HACMP)-Technical-Forum
    #PowerHAforAIX


  • 2.  Re: How to determine that a failover has happened

    Posted Tue May 19, 2009 02:43 PM

    Originally posted by: Casey_B


    Hello,

    If I understand you correctly, you are looking to distinguish between the two following cases:

    1) A resource group is brought online on a node because of hardware failure....called a fallover, or failover.

    2) The resource group moves back to it's home node...Commonly called a fallback.

    It doesn't seem you are looking for history...(Which if you are, that can be found in cluster.log, or hacmp.out)

    My first thought....and I didn't check this, yet...Is that you can add a notification for node_down, and rg_move, and you would have
    a pretty good idea of when the fallover occurs.

    And that you could add a notify for node_up and be pretty sure when a fallback occurs.

    I would think though that it would be just as important to be notified when either case occurs...
    Because either case means at least a short downtime to your applications that someone is going to want a report on....

    If you didn't want to add a notification method, one could also add a pre/post event to do something more custom if you needed that.

    Of course...there is the last option, which is a do-it-yourself option, of adding logic into your application start script to
    send off an email, or something else appropriate...
    Hope this helps,
    Casey
    #PowerHAforAIX
    #PowerHA-(Formerly-known-as-HACMP)-Technical-Forum