When does ServiceGroupHB do a Failover? - Veritas Cluster Server

This is a discussion on When does ServiceGroupHB do a Failover? - Veritas Cluster Server ; I have the following Scenario: SystemA & SystemB each running a NFS Service Group. Both GroupX and GroupY use ServiceGroupHB on shared disks defined as: ServiceGroupHB Xhb ( Disks = { "/dev/dsk/c1t2d9s7", 64, "/dev/dsk/c1t2d10s7", 64 } ) ... ServiceGroupHB Yhb ...

+ Reply to Thread
Results 1 to 2 of 2

Thread: When does ServiceGroupHB do a Failover?

  1. When does ServiceGroupHB do a Failover?


    I have the following Scenario:
    SystemA & SystemB each running a NFS Service Group.
    Both GroupX and GroupY use ServiceGroupHB on shared disks defined as:
    ServiceGroupHB Xhb (
    Disks = { "/dev/dsk/c1t2d9s7", 64, "/dev/dsk/c1t2d10s7",
    64 }
    )
    ...
    ServiceGroupHB Yhb (
    Disks = { "/dev/dsk/c1t2d9s7", 256, "/dev/dsk/c1t2d10s7",
    256 }
    )

    GroupX is running on SystemA and GroupY running on SystemB.


    Then both network HB are lost.
    SystemA attempts to do Failover of GroupY but ServiceGroupHB
    prevents GroupY from coming up since it's running on SystemB
    Same thing on SystemB trying to Failover GroupX.
    On each half of the Cluster, the ServiceGroupHB resource for
    the attempted failover group has an IState of W_ONLINE
    This seems to prevent a "Split Brain" situation but I have 2 questions:

    1- At this point, if SystemA crashes, shouldn't ServiceGroupHB resource
    complete it's Online and GroupX come up?
    (In my case the ServiceGroupHB resource stays in W_ONLINE)

    2- If both Systems are still running, what would be the proper procedure
    to join the two systems back as a Cluster again.
    (Only way I found was to do hagrp -flush to get out of the W_ONLINE)

    Thanks for any info

  2. Re: When does ServiceGroupHB do a Failover?


    Hi Paul,

    According to Veritas support when you lose both HB the VCS configures itself
    into 2- mini-clusters to prevent this split brain...the SG remain running
    on each mini-cluster and does not fail-over (although their GUI is misleading)...Vertas
    noted that the GUI cannot handle or does not recognize this "unknown" that
    the VCS failed into when losing both HB.

    I am unsure of how to rejoin the clusters and assumed that a shutdown and
    reboot would be required to initalize a new cluster membership..

    Cheers,

    Jim




    "Paul OBrien" wrote:
    >
    >I have the following Scenario:
    >SystemA & SystemB each running a NFS Service Group.
    >Both GroupX and GroupY use ServiceGroupHB on shared disks defined as:
    > ServiceGroupHB Xhb (
    > Disks = { "/dev/dsk/c1t2d9s7", 64, "/dev/dsk/c1t2d10s7",
    >64 }
    > )
    >...
    > ServiceGroupHB Yhb (
    > Disks = { "/dev/dsk/c1t2d9s7", 256, "/dev/dsk/c1t2d10s7",
    >256 }
    > )
    >
    >GroupX is running on SystemA and GroupY running on SystemB.
    >
    >
    >Then both network HB are lost.
    >SystemA attempts to do Failover of GroupY but ServiceGroupHB
    >prevents GroupY from coming up since it's running on SystemB
    >Same thing on SystemB trying to Failover GroupX.
    >On each half of the Cluster, the ServiceGroupHB resource for
    >the attempted failover group has an IState of W_ONLINE
    >This seems to prevent a "Split Brain" situation but I have 2 questions:
    >
    >1- At this point, if SystemA crashes, shouldn't ServiceGroupHB resource


    > complete it's Online and GroupX come up?
    > (In my case the ServiceGroupHB resource stays in W_ONLINE)
    >
    >2- If both Systems are still running, what would be the proper procedure
    > to join the two systems back as a Cluster again.
    > (Only way I found was to do hagrp -flush to get out of the W_ONLINE)
    >
    >Thanks for any info



+ Reply to Thread