Can NFSv3 unmount failure result in failure in reboot - NFS

This is a discussion on Can NFSv3 unmount failure result in failure in reboot - NFS ; Hi, I have a scenario where a dir from server (say, /nfs_dir) is mounted on many (more than 10) clients. While halting one of the clients, I get resource busy error on mounted dir, so the dir is remounted as ...

+ Reply to Thread
Results 1 to 4 of 4

Thread: Can NFSv3 unmount failure result in failure in reboot

  1. Can NFSv3 unmount failure result in failure in reboot

    Hi,

    I have a scenario where a dir from server (say, /nfs_dir) is mounted on
    many (more than 10) clients. While halting one of the clients, I get
    resource busy error on mounted dir, so the dir is remounted as
    read-only and the client shuts down. (Other clients are still up)
    However, if the client is again booted up, it is not able to mount the
    same dir from same server. But if some delay is provided (sometimes
    more than a 2-3 minutes), the client cleanly mounts from server.
    Clients are using busybox 0.60.5 and umount is done from /. Kernel is
    2.6.5 on ppc.
    Does anyone has any ideas about this?

    Thanks!

    NT


  2. Re: Can NFSv3 unmount failure result in failure in reboot

    ntrivedi@gmail.com wrote:
    > Hi,
    >
    > I have a scenario where a dir from server (say, /nfs_dir) is mounted on
    > many (more than 10) clients. While halting one of the clients, I get
    > resource busy error on mounted dir, so the dir is remounted as
    > read-only and the client shuts down. (Other clients are still up)
    > However, if the client is again booted up, it is not able to mount the
    > same dir from same server. But if some delay is provided (sometimes
    > more than a 2-3 minutes), the client cleanly mounts from server.
    > Clients are using busybox 0.60.5 and umount is done from /. Kernel is
    > 2.6.5 on ppc.
    > Does anyone has any ideas about this?
    >
    > Thanks!
    >
    > NT


    It doesn't sound like an NFS problem. I suspect that you may have some
    network issues right after the client reboots. Could be spanning tree,
    or arp, or who knows what. Can your client ping the server, or do
    `showmount -e ', right after it boots up?

    bc


  3. Re: Can NFSv3 unmount failure result in failure in reboot

    The client has its root and some basic storage on ramdisk. The rc
    scripts point towards this remote dir on which resides the busybox too.
    So if the client fails to mount the remote dir, it gives "mount RPC:
    Unable to recieve; errno = Connection refused", and "mount: nfsmount
    failed: Bad file descriptor".
    However, if after some attempts client is able to mount the remote dir,
    while shutting down the client, that remote dir fails to unmount,
    showing resource is busy.
    -NT

    bcwalrus wrote:
    > ntrivedi@gmail.com wrote:
    > > Hi,
    > >
    > > I have a scenario where a dir from server (say, /nfs_dir) is mounted on
    > > many (more than 10) clients. While halting one of the clients, I get
    > > resource busy error on mounted dir, so the dir is remounted as
    > > read-only and the client shuts down. (Other clients are still up)
    > > However, if the client is again booted up, it is not able to mount the
    > > same dir from same server. But if some delay is provided (sometimes
    > > more than a 2-3 minutes), the client cleanly mounts from server.
    > > Clients are using busybox 0.60.5 and umount is done from /. Kernel is
    > > 2.6.5 on ppc.
    > > Does anyone has any ideas about this?
    > >
    > > Thanks!
    > >
    > > NT

    >
    > It doesn't sound like an NFS problem. I suspect that you may have some
    > network issues right after the client reboots. Could be spanning tree,
    > or arp, or who knows what. Can your client ping the server, or do
    > `showmount -e ', right after it boots up?
    >
    > bc



  4. Re: Can NFSv3 unmount failure result in failure in reboot

    ntrivedi@gmail.com wrote:
    > The client has its root and some basic storage on ramdisk. The rc
    > scripts point towards this remote dir on which resides the busybox too.
    > So if the client fails to mount the remote dir, it gives "mount RPC:
    > Unable to recieve; errno = Connection refused", and "mount: nfsmount
    > failed: Bad file descriptor".
    > However, if after some attempts client is able to mount the remote dir,
    > while shutting down the client, that remote dir fails to unmount,
    > showing resource is busy.
    > -NT


    So the mount happens automatically at boot time and it's difficult for
    you to test whether the network connectivity to the server.

    The "connection refused" message indicates that it's indeed a
    networking problem, that your client just can't talk to the server. I
    don't really know how your environment is set up. But you can try these
    to debug your problem:

    - Find out exactly at what point the mount happens.
    - Add debug statements (e.g. "ping -c 3 ", "showmount -e
    ", "ip route") to your rc script right before the mount
    happens.

    Hopefully you'll know why your client can't talk to the server.

    Cheers,
    bc


+ Reply to Thread