nfs kills network ???
Hi,
I have a strange problem with nfs.
My debian Workstation is connected to a Kubuntu server via nfs. When I try to read lots of data (let's say generating thumbnails for all ma images images) the network stops after 20-60 Seconds. After 20-60 Seconds more the connection starts working again.
When it's down the server doesn't even answer to pings.
In the servers /var/log/messages a simple
Jun 23 01:54:15 Atlas kernel: [ 321.004390] r8169: eth0: link up
Jun 23 01:55:45 Atlas kernel: [ 411.004467] r8169: eth0: link up
Jun 23 01:56:09 Atlas kernel: [ 435.004491] r8169: eth0: link up
Jun 23 01:56:45 Atlas kernel: [ 471.004453] r8169: eth0: link up
Jun 23 01:56:57 Atlas kernel: [ 483.004500] r8169: eth0: link up
Jun 23 01:57:15 Atlas kernel: [ 501.004551] r8169: eth0: link up
Jun 23 01:58:33 Atlas kernel: [ 579.009313] r8169: eth0: link up
Jun 23 01:59:45 Atlas kernel: [ 651.004557] r8169: eth0: link up
appears, No error message, nothing.
The clients /var/log/messages says
Jun 23 02:03:14 Phoenix kernel: [43332.373149] nfs: server Atlas not responding, still trying
Jun 23 02:03:14 Phoenix kernel: [43332.373164] nfs: server Atlas not responding, still trying
Jun 23 02:03:14 Phoenix kernel: [43332.373173] nfs: server Atlas not responding, still trying
Jun 23 02:03:14 Phoenix kernel: [43332.373181] nfs: server Atlas not responding, still trying
Jun 23 02:03:25 Phoenix kernel: [43342.865407] nfs: server Atlas OK
Jun 23 02:03:25 Phoenix kernel: [43342.868858] nfs: server Atlas OK
Jun 23 02:03:25 Phoenix kernel: [43342.871231] nfs: server Atlas OK
the servers /etc/exports file looks like this:
/ Phoenix(async,rw,no_root_squash,no_subtree_check,crossmnt)\
192.168.178.77(async,rw,no_root_squash,no_subtree_check,crossmnt)\
Lucy(async,rw,no_root_squash,no_subtree_check,crossmnt)
The client's fstab looks like this:
Atlas:/ /atlas nfs rsize=8192,wsize=8192,user,exec,async,noatime 1 0
After switchig to sshfs the problem disappeared.
Why is that?
Would it be a good idea to file a bug to the nfs developers (and if so, how)?
Maybe it's a bad idea anyway to use nfs instead of sshfs?
btw. How can one acitvate or deactivate network devices under Jaunty? ifup and ifdown doesn't work and /etc/network/interfaces only contains the loop device. Some kind of udev magic?
UPDATE: After half an hour with sshfs the interface again stopped working (at least only once). So there seems to be a basic problem that's just extremly triggered by the use of nfs. (doesn't sound very clever though)
If anyone has a clue how to investigate further on this matter, please let men know.
Last edited by anutosho; 06-22-2009 at 09:22 PM.
|