Forum Discussion

mflores's avatar
mflores
Level 2
15 years ago

Dependence between Service Groups in Disaster Recovery solution.

Hi,

I´m trying to implement a Disaster Recovery solution on a "two server per site" scenario but I´m having some problems.... I´ve created a service group (SG) for each server (database and middleware) and I don´t know how to set the dependence between these SG to make the entire cluster failover to the secondary site in the case of any fault on the primary. I´ve tried linking the SG´s with an "Online Remote Firm" dependence and it worked well when the fault occurs in the child-SG, but if the fault occurs in the parent-SG the cluster doesn´t failover to the entire site, in fact the child-SG never comes offline. Is there any way to do this?? I think that the "hard" rigidity dependence should work well but unfortunately it only exists in a local relationship betweet SG´s and mine are stored in separate servers.

Thanks!
business continuity
Cluster Server
Installing
Tip-How to
  • Dev_Roy's avatar
    Dev_Roy
    Level 6
    15 years ago
    Another thing to keep in mind, if you want any resource from any of the service group faults then the both the SG should failover to another node then make all the resources as Critical other wise failover may not happen if a non-critical resource fails or if no critical resource in the path of non-critical reource dependency.
  • mflores's avatar
    mflores
    Level 2
    15 years ago
    Hi and thanks for the answer!!

    I can´t make an "Online Local Hard" dependency because my SG´s are defined on different hosts and so only "Remote" and "Global" options are allowed. Is this right?

    All my resources are "Critical" and when a Child-SG resource fails everything works OK, the issue comes in when a Parent-SG resource fails becuse the "Firm" dependency doesn´t force the Child-SG to failover automatically and in the other site the Parent-SG can´t get online because of the group dependency is not met (the child stays offline).
  • Dev_Roy's avatar
    Dev_Roy
    Level 6
    15 years ago
    In case of  "online local hard" it is mandetory to have both the Service Group to be on the same node. In this case if any reources Faults in any of the Service Group it would have over both -parent and child to the other node.

    And in case if you just want if you want SGs to be running on different nodes in that case you will not be allowed to link the SGs as "online local hard" but in this case without linking  the Service Groups on the both the nodes, they should be able to failover to the other available node if any Faults occurs in any of the resource of any of the Service Group as per VCS behaviour for Critical resources.

    I think I am missing something in your question, could you please rephrase your requriement clearly once again.

    Regards,
    Dev Roy.
  • Roger_Zimmerman's avatar
    Roger_Zimmerman
    Level 4
    15 years ago
    Hi, mflores,

    first question from my point of view is, why you want to switch both application parts to the other side in case of a failure in one of them. If only one application fails and switches to the other site, the network connect should do the job like in the local situation. And if really you site is failing, then failures in both applications will toccur and both applications will switch to the other site. So maybe it is not necessary to switch both applications every times. So far so good.

    If you want to do this anyway then think about a preonline script to your parent application what checks if the child SG is running on the local site and if not then "hagrp -switch" the child group to the local site. (Take care for bringing online the parent SG on the selected node again with "hagrp -online -nopre <grp> -sys <node>" as last action in this preonline script because the online token is blown by the preonline script.) But keep in mind that every switch to the other site for the parent application then produces a downtime for the child SG!

    Hope this helps, let me know if this is going to the wanted direction.

    Best regards
    Roger