Problem about NFS in VCS - Veritas Cluster Server

This is a discussion on Problem about NFS in VCS - Veritas Cluster Server ; Hi,I have met a problem about NFS in VCS. I have configured a NFS group in two hosts system.And there is a floating IP in it.A NFS client can mount the NFS directory from floating IP. When the NFS group ...

+ Reply to Thread
Results 1 to 4 of 4

Thread: Problem about NFS in VCS

  1. Problem about NFS in VCS


    Hi,I have met a problem about NFS in VCS.
    I have configured a NFS group in two hosts system.And there is a floating
    IP in it.A NFS client can mount the NFS directory from floating IP.
    When the NFS group is online and offline in only one host,the NFS client
    can reconnect automatically,and Do not need to umount and remount again.
    But when the NFS group switch to another host,The NFS client can not reconnect
    automatically,and must umount and remount the nfs directory manully.And the
    error messages is "df:can not statvfs /mo: stale NFS file handle".
    But why? Just because that having changed the network card from one host
    to the other?

  2. Re: Problem about NFS in VCS

    Hi John,

    You have to make sure that the major and the minor numbers for the disks
    and/or VxVM volumes are the same on both nodes. There is a chapter about
    this in the VCS install guide, this will help you to sort this out.

    BR,
    Peter


    John wrote:
    > Hi,I have met a problem about NFS in VCS.
    > I have configured a NFS group in two hosts system.And there is a floating
    > IP in it.A NFS client can mount the NFS directory from floating IP.
    > When the NFS group is online and offline in only one host,the NFS client
    > can reconnect automatically,and Do not need to umount and remount again.
    > But when the NFS group switch to another host,The NFS client can not reconnect
    > automatically,and must umount and remount the nfs directory manully.And the
    > error messages is "df:can not statvfs /mo: stale NFS file handle".
    > But why? Just because that having changed the network card from one host
    > to the other?



  3. Re: Problem about NFS in VCS

    I have seen this same error but the error was the result of the
    automounted points on the server. Detailed Explanation follows

    System A has home directory volume, on /export/Home.
    There are active mount points through autofs in /home/ which
    is really /export/Home/, automount is setup as IP>:/export/Home/
    Service group fails.
    df on System A now produces error:
    is already mounted, /export/Home is busy,
    allowable number of mount points exceeded

    This results in the service group not being able to come up on System A
    again.

    The problem is that when the Service group was failed from System B back
    to System A the mounts on that side now keep it from from starting as
    well, and the only solution was to reboot one of the servers in order to
    bring the Service Group online.

    My question is how can this be avoided,. when the Service group does a
    umount -f?

    Doug Reitmeier
    douglas.reitmeier@udlp.com


    Peter Sevborn wrote:
    >
    > Hi John,
    >
    > You have to make sure that the major and the minor numbers for the disks
    > and/or VxVM volumes are the same on both nodes. There is a chapter about
    > this in the VCS install guide, this will help you to sort this out.
    >
    > BR,
    > Peter
    >
    > John wrote:
    > > Hi,I have met a problem about NFS in VCS.
    > > I have configured a NFS group in two hosts system.And there is a floating
    > > IP in it.A NFS client can mount the NFS directory from floating IP.
    > > When the NFS group is online and offline in only one host,the NFS client
    > > can reconnect automatically,and Do not need to umount and remount again.
    > > But when the NFS group switch to another host,The NFS client can not reconnect
    > > automatically,and must umount and remount the nfs directory manully.And the
    > > error messages is "df:can not statvfs /mo: stale NFS file handle".
    > > But why? Just because that having changed the network card from one host
    > > to the other?


    --
    Doug Reitmeier United Defense L.P.
    douglas_reitmeier@udlp.com 4800 East River Rd. M392
    762-572-6644 Fridley, MN 55421

  4. Re: Problem about NFS in VCS

    One thing to remember is that using autofs on the cluster nodes is not a
    good idea. If system A is sharing a file system and system A mounts it,
    it will probably not be working if the shared file system fails over to
    system B. This is due to the fact that that if you mount a file system
    shared on the local node the mount will use the loop-back address and
    not the service ip-address, and when that service fails the mount point
    will be stale. If the setup is done as stated in the VCS install guide
    it will work from all clients. I've done this many times and it works
    like a charm as long as the major and minor numbers are the same on both
    nodes. I have several customers with the NFS service running under VCS
    control and I've never seen any problems.

    /P

    Douglas J. Reitmeier wrote:
    > I have seen this same error but the error was the result of the
    > automounted points on the server. Detailed Explanation follows
    >
    > System A has home directory volume, on /export/Home.
    > There are active mount points through autofs in /home/ which
    > is really /export/Home/, automount is setup as > IP>:/export/Home/
    > Service group fails.
    > df on System A now produces error:
    > is already mounted, /export/Home is busy,
    > allowable number of mount points exceeded
    >
    > This results in the service group not being able to come up on System A
    > again.
    >
    > The problem is that when the Service group was failed from System B back
    > to System A the mounts on that side now keep it from from starting as
    > well, and the only solution was to reboot one of the servers in order to
    > bring the Service Group online.
    >
    > My question is how can this be avoided,. when the Service group does a
    > umount -f?
    >
    > Doug Reitmeier
    > douglas.reitmeier@udlp.com
    >
    >
    > Peter Sevborn wrote:
    >
    >>Hi John,
    >>
    >>You have to make sure that the major and the minor numbers for the disks
    >>and/or VxVM volumes are the same on both nodes. There is a chapter about
    >>this in the VCS install guide, this will help you to sort this out.
    >>
    >>BR,
    >>Peter
    >>
    >>John wrote:
    >>
    >>>Hi,I have met a problem about NFS in VCS.
    >>>I have configured a NFS group in two hosts system.And there is a floating
    >>>IP in it.A NFS client can mount the NFS directory from floating IP.
    >>>When the NFS group is online and offline in only one host,the NFS client
    >>>can reconnect automatically,and Do not need to umount and remount again.
    >>>But when the NFS group switch to another host,The NFS client can not reconnect
    >>>automatically,and must umount and remount the nfs directory manully.And the
    >>>error messages is "df:can not statvfs /mo: stale NFS file handle".
    >>>But why? Just because that having changed the network card from one host
    >>>to the other?

    >>

    >
    > --
    > Doug Reitmeier United Defense L.P.
    > douglas_reitmeier@udlp.com 4800 East River Rd. M392
    > 762-572-6644 Fridley, MN 55421



+ Reply to Thread