Going mad? NFS/TCP client make linux server go bang... - Solaris

This is a discussion on Going mad? NFS/TCP client make linux server go bang... - Solaris ; Hi, Finally have most of my hardware for my NAS box. Anyway, in preparation for jumpstating it (no CD drive in new box, have nice PXELinux setup anyway, thought: what the hell, why not?...) I tried to run the DVD ...

+ Reply to Thread
Results 1 to 3 of 3

Thread: Going mad? NFS/TCP client make linux server go bang...

  1. Going mad? NFS/TCP client make linux server go bang...

    Hi,

    Finally have most of my hardware for my NAS box.

    Anyway, in preparation for jumpstating it (no CD drive in new box, have nice
    PXELinux setup anyway, thought: what the hell, why not?...) I tried to run
    the DVD to network installer prep script an dump everything on my linux NFS
    server.

    Anyway, script not work under linux - fair enough, worth a go anyway...

    One VMWare Solaris 10 install later all going OK. Install autofs maps[1].

    Hmm, linux server (2.6.17) goes "Oops" on every attempted mount. Remember an
    old problem from 3-4 years ago where Solaris 8 NFS/TCP clients liked
    blowing up linux NFS.

    Switched to UDP (over NFS v3) and it works.

    Now, I would have thought an issue that old would be fixed by now... Is it
    still a known problem?

    Cheers

    Tim

    [1] Yeah, I was going to dump the jumpstart installer stuff straight onto
    my /vol/solaris, until the script decided it didn't like that. VMWare is in
    the process of lending it another vast "local" disk to setup onto, which
    will them get copied over to the NFS server.

  2. Re: Going mad? NFS/TCP client make linux server go bang...

    Hi Andrew,

    Andrew Gabriel coughed up some electrons that declared:

    >
    > My first suspicion would be Linux's NFSv4 implementation,
    > which I've had trouble getting to work with Solaris, and
    > even between Linux boxes.
    >


    Well, that was the first problem I had: I have enough of a setup to support
    NFSv4, but I've never configured it, though it may be advertised. Solaris
    did pick that up, and failed on auth errors as expected. So I downgraded
    the Solaris client to v3. Still kills the linux server (enough on one
    occasion to need a reboot) if using TCP. Seems to definitely be a UDP
    versus TCP issue.

    Cheers

    Tim

  3. Re: Going mad? NFS/TCP client make linux server go bang...

    On Jul 25, 4:08 pm, Tim Southerwood wrote:

    > Hmm, linux server (2.6.17) goes "Oops" on every attempted mount. Remember an
    > old problem from 3-4 years ago where Solaris 8 NFS/TCP clients liked
    > blowing up linux NFS.
    >
    > Switched to UDP (over NFS v3) and it works.
    >
    > Now, I would have thought an issue that old would be fixed by now... Is it
    > still a known problem?
    >


    Don't know about the Solaris 8 client problems but I have been
    using Linux nfs servers on kernels 2.6.17 and 2.6.12 with
    Solaris 10 clients and never had any problems using nfs ver. 3
    with tcp. I always use the default server options.
    I haven't tried version 4 yet.
    Some newer Linux distributions changed the default nfs option
    from async to sync, which could have some impact on
    performance



+ Reply to Thread