Forum Discussion

Jomy's avatar
Jomy
Level 5
11 years ago

How to prevent split brain issue in GCO

How to prevent split brain issue in GCO.

 

we have configured 2 cluster in primary and single node cluster in secondary.

It is working fine but.

when ever there is a link issue.Both the sites are going in active state and volumes are mounting on both the sides.

Is there any configuration settings to avoid this.

 

 

business continuity
Configuring
Storage Foundation for Windows
Troubleshooting
Windows Server (2003-2008)
  • mikebounds's avatar
    mikebounds
    11 years ago

    The default for Service group attribute ClusterFailOverPolicy is manual (for global service groups) and the reason for this is that if you have your issue where site is not down and it is link that is down, that the groups on the other cluster to do not take over, so have you changed this setting to Auto?

    Also for normal startup, VCS uses Service group attribute Authority, so this prevents any cluster that reboots from starting when it maybe up on the other site.

    Mike

4 Replies

Sort By
  • RiaanBadenhorst's avatar
    RiaanBadenhorst
    Level 6
    11 years ago

    Hi,

     

    If you want to run auto then you could configure a steward. See below

     

    https://sort.symantec.com/public/documents/vcs/6.1/aix/productguides/html/vcs_admin/ch16s02s07.htm

  • Mike_Magarino's avatar
    Mike_Magarino
    Level 3
    11 years ago

    Jomy,

    The configuration suggestions above are a good place to start. VCS has many preventative features that attempt to prevent split brain. Typically, in order for this to happen, all heartbeat links are being lost at the very same time. This is why we recommend more than 1 heartbeat, as it prevents a single point of failure. 

    I would suggest investigating why heartbeat links are being lost at the same time.

  • mikebounds's avatar
    mikebounds
    Level 6
    11 years ago

    The default for Service group attribute ClusterFailOverPolicy is manual (for global service groups) and the reason for this is that if you have your issue where site is not down and it is link that is down, that the groups on the other cluster to do not take over, so have you changed this setting to Auto?

    Also for normal startup, VCS uses Service group attribute Authority, so this prevents any cluster that reboots from starting when it maybe up on the other site.

    Mike

  • Yeying198201's avatar
    Yeying198201
    Level 3
    11 years ago
    Do NOT set AutoTakeover attribute on RVGPrimary resource, then even there is a link issue, remote single-node cluster will not mount volumes.