Lack of failover when hme0 cable is pulled - Veritas Cluster Server

This is a discussion on Lack of failover when hme0 cable is pulled - Veritas Cluster Server ; Greetings, I have a cluster up and running using Foundation Suite HA 2000-12 with VCS 1.3.0. All aspects of the cluster work perfectly with one exception. The main network connection (hme0) can be pulled out and the console window on ...

+ Reply to Thread
Results 1 to 4 of 4

Thread: Lack of failover when hme0 cable is pulled

  1. Lack of failover when hme0 cable is pulled


    Greetings,

    I have a cluster up and running using Foundation Suite HA 2000-12
    with VCS 1.3.0. All aspects of the cluster work perfectly with one
    exception.

    The main network connection (hme0) can be pulled out and the console
    window on my Solaris 8 box sees the disconnect. ifconfig -a doesn't
    change values and the cluster will not fail over.

    I am not sure if it is a cluster set up problem or operator error.

    Any one who can tell me how to set up the IP or NIC card to
    failover when the cable is pulled, will have my warmest thanks.

  2. Re: Lack of failover when hme0 cable is pulled

    Kevin Gee wrote:
    >
    > Greetings,
    >
    > I have a cluster up and running using Foundation Suite HA 2000-12
    > with VCS 1.3.0. All aspects of the cluster work perfectly with one
    > exception.
    >
    > The main network connection (hme0) can be pulled out and the console
    > window on my Solaris 8 box sees the disconnect. ifconfig -a doesn't
    > change values and the cluster will not fail over.
    >
    > I am not sure if it is a cluster set up problem or operator error.
    >
    > Any one who can tell me how to set up the IP or NIC card to
    > failover when the cable is pulled, will have my warmest thanks.


    Vcs uses netstat -in to determine if the network interface is still
    receiving incoming packets, if they are not then VCS will fault the NIC
    and take action. It's possible that Netstat is not returning to the
    monitor script, this can been seen as monitor timeouts. Make sure all
    the patches are up to date and watch the engine_A.log when you test
    again. If you need further assistance open a case with support

    Bill Borsari

    ** Any opinions expressed are my own and not of my employer **

  3. Re: Lack of failover when hme0 cable is pulled

    Kevin,

    Is hme0 used for the heartbeat?
    If it is then the box should fall over.

    If not then you need to set up the nic and the virtual ip address assigned
    to it as a dependency in a group, otherwise VCS could care less whether the
    connection is pulled or not. If the cluster was setup by someone else they
    probably thought you waould want this...this way the interface always has
    the same address and the machine can be accessed for service through it, or
    the console. When you look at the resource group dependency tree, is the
    interface in the tree as a child of the share or app you want to have
    failover?

    Hope this helps...

    Robert Westling

    William Borsari wrote:

    > Kevin Gee wrote:
    > >
    > > Greetings,
    > >
    > > I have a cluster up and running using Foundation Suite HA 2000-12
    > > with VCS 1.3.0. All aspects of the cluster work perfectly with one
    > > exception.
    > >
    > > The main network connection (hme0) can be pulled out and the console
    > > window on my Solaris 8 box sees the disconnect. ifconfig -a doesn't
    > > change values and the cluster will not fail over.
    > >
    > > I am not sure if it is a cluster set up problem or operator error.
    > >
    > > Any one who can tell me how to set up the IP or NIC card to
    > > failover when the cable is pulled, will have my warmest thanks.

    >
    > Vcs uses netstat -in to determine if the network interface is still
    > receiving incoming packets, if they are not then VCS will fault the NIC
    > and take action. It's possible that Netstat is not returning to the
    > monitor script, this can been seen as monitor timeouts. Make sure all
    > the patches are up to date and watch the engine_A.log when you test
    > again. If you need further assistance open a case with support
    >
    > Bill Borsari
    >
    > ** Any opinions expressed are my own and not of my employer **



  4. Re: Lack of failover when hme0 cable is pulled

    kevin,
    have you set the monitoring interval to a shorter period?? you will see
    the result faster
    Thanks

    Robert Westling wrote:

    > Kevin,
    >
    > Is hme0 used for the heartbeat?
    > If it is then the box should fall over.
    >
    > If not then you need to set up the nic and the virtual ip address assigned
    > to it as a dependency in a group, otherwise VCS could care less whether the
    > connection is pulled or not. If the cluster was setup by someone else they
    > probably thought you waould want this...this way the interface always has
    > the same address and the machine can be accessed for service through it, or
    > the console. When you look at the resource group dependency tree, is the
    > interface in the tree as a child of the share or app you want to have
    > failover?
    >
    > Hope this helps...
    >
    > Robert Westling
    >
    > William Borsari wrote:
    >
    > > Kevin Gee wrote:
    > > >
    > > > Greetings,
    > > >
    > > > I have a cluster up and running using Foundation Suite HA 2000-12
    > > > with VCS 1.3.0. All aspects of the cluster work perfectly with one
    > > > exception.
    > > >
    > > > The main network connection (hme0) can be pulled out and the console
    > > > window on my Solaris 8 box sees the disconnect. ifconfig -a doesn't
    > > > change values and the cluster will not fail over.
    > > >
    > > > I am not sure if it is a cluster set up problem or operator error.
    > > >
    > > > Any one who can tell me how to set up the IP or NIC card to
    > > > failover when the cable is pulled, will have my warmest thanks.

    > >
    > > Vcs uses netstat -in to determine if the network interface is still
    > > receiving incoming packets, if they are not then VCS will fault the NIC
    > > and take action. It's possible that Netstat is not returning to the
    > > monitor script, this can been seen as monitor timeouts. Make sure all
    > > the patches are up to date and watch the engine_A.log when you test
    > > again. If you need further assistance open a case with support
    > >
    > > Bill Borsari
    > >
    > > ** Any opinions expressed are my own and not of my employer **



+ Reply to Thread