Re: NFS slow/lockups

From: Tim Kelley (tpk_at_r00tserverz.net)
Date: 10/14/04

  • Next message: Miquel van Smoorenburg: "Re: change /dev/console owner to user logged on console?" 
    Date: Wed, 13 Oct 2004 17:33:47 -0500
To: debian-user@lists.debian.org

    On Wed, Oct 13, 2004 at 09:05:21PM +0100, Alan Chandler wrote:
    > I am trying copy about 350Mb of files - each one about 2.8MB each - over an
    > nfs link. Although I primarily trying this via drag and drop inside kde,
    > command line copying is also not working
    >
    > With the sync option it is very slow - the dialog box can't report the copy
    > rate and marks it as stalled. With the async option it goes fast until about
    > 30MB are copied and then comes to a complete halt.
    >
    > I was using the nfs-user daemon, but I have also switched over to the
    > nfs-kernel one. Doesn't seem to make any difference.
    >
    > Just to show its not networking problems, I have resorted to copying the files
    > via scp - even encrypted I am getting about 0.5Mb/sec

    That's horrible! Is this a 10Mb network or dialup or what? Surely you
    mean .5 MBytes/sec (which also isn't too hot for a 10Mbit ethernet).
     
    > Any idea whats wrong and how to improve performance?

    Some basic things:

    1. If you have a switch, make sure it your interfaces are at full
    duplex

    2. try mounting with the tcp option passed to mount, you can put this
    in your auto.master or at mount time or whatever. You would need tcp
    support compiled in on the server side for this. Look for
    "CONFIG_NFSD_TCP=y" in config for your current kernel (this is only
    necessary on the linux nfs server)

    3. make sure if your using udp that you mount with
    rsize=8192,wsize=8192. If you are using tcp you can go to
    rsize=32768,wsize=32768

    4. make sure you're using nfs version 3. nfstat will tell you, in
    fact, it may tell you a bit more about what is wrong. 

    -- 
  _   _   _   _   _   _   _   _   _   _   _   _   _  
 / \ / \ / \ / \ / \ / \ / \ / \ / \ / \ / \ / \ / \ 
( t | i | m | @ | i | t | . | k | p | t | . | c | c )
 \_/ \_/ \_/ \_/ \_/ \_/ \_/ \_/ \_/ \_/ \_/ \_/ \_/ 
GPG key fingerprint = 1DEE CD9B 4808 F608 FBBF  DC21 2807 D7D3 09CA 85BF
-- 
To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org 
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
  • Next message: Miquel van Smoorenburg: "Re: change /dev/console owner to user logged on console?"

    Relevant Pages

    • nfs/ssh woes
      ... "Encrypted NFS with OpenSSH and Linux" ... 100000 2 tcp 111 portmapper ... 100000 2 udp 111 portmapper ... mount: RPC: Remote system error - Connection refused ...
      (Debian-User)
    • NFS Problems
      ... Indigo2 machine in network w/ FreeBSD, both configured for nfs server/client functions, and "nfsd" daemon running on both. ... FreeBSD machine mounts IRIX disks OK, but Indigo2 does not mount FreeBSD disks. ... 100000 2 tcp 111 portmapper ... 100000 2 udp 111 portmapper ...
      (freebsd-hackers)
    • Trouble: NFS via TCP
      ... I've got a very weird problem with NFS mounts on a RELENG_6 ... NFS directory via TCP, ... mount command, it simply hangs forever. ... Kernel config is GENERIC with a few additions for more ...
      (freebsd-stable)
    • Re: NFS inconsistent behaviour
      ... I do not want to mount unmount infinitely but was just checking ... >> of tcp connections in TIME_WAIT state. ... > and remote port so the ports stay in use for a few minutes. ... > I'd switch to NFS over udp if this is problem. ...
      (Linux-Kernel)
    • Re: [SLE] NFS
      ... >mount a nfs share. ... Did you read the error message and the man page for mount? ... supports tcp as well as udp, ...
      (SuSE)