[NFS] Performance

[BigPanda]

Hi,

NFS performance problems aren't new to FC and has been reported a couple of times, but I haven't found any solution for that.

Maybe meanwhile someone has fixed it.


My situation:
server (10.42.1.253): Linux server 2.6.8-1.521 (Fedora Core 2)
exports a directory /video via NFSv4

client (10.42.1.251): Linux client 2.4.24-ctvdr-2 (Debian - c't-vdr-Distribution)
is mounting from server the directory /video
server:/video /video nfs rw,user,wsize=8192,rsize=8192,intr 0 0
via NFSv3

Copying a file on client from client to server
cp /tmp/testdatei /video

takes extremly long and produces in client:/var/log/messages
Aug 28 17:19:47 client kernel: nfs: server server not responding, still trying
Aug 28 17:19:47 client kernel: nfs: server server OK


The LAN is not used much and via FTP I had an average speed of 24MBit/s, which is OK from my point of view. Also the CPU load is low.


Has anyone ideas how to solve the NFS performance problem?



Steffen

[crackers]

Try bumping rsize/wsize to 32768 - since the mount is named "video" I'm guessing you're looking at some very large files and the larger buffers will help quite a bit.

[BigPanda]

Thank you Crackers for your post,

you are right, I want to handle hugh files.
My VDR should put the recorded files to the server directory.

I tried your settings, but this doesn't solve the problem.

There are still tons of

Aug 29 10:22:51 client kernel: nfs: server 10.42.1.253 not responding, still trying
Aug 29 10:22:54 client kernel: nfs: server 10.42.1.253 OK
Aug 29 10:23:03 client kernel: nfs: server 10.42.1.253 not responding, still trying
Aug 29 10:23:12 client kernel: nfs: server 10.42.1.253 not responding, still trying
Aug 29 10:23:12 client kernel: nfs: server 10.42.1.253 OK
Aug 29 10:23:12 client kernel: nfs: server 10.42.1.253 OK
Aug 29 10:23:26 client kernel: nfs: server 10.42.1.253 not responding, still trying
Aug 29 10:23:26 client kernel: nfs: server 10.42.1.253 OK

in client:/var/log/messages.



Cheers,
Steffen

[jebba]

I'm seeing identical problems here.

Server: FC2, 3Ware RAID on gigabit ethernet
Clients: FC2 boxes and Debian boxes (i386 & AMD64) on gigabit ethernet

The FC2 boxes NFS mount via NFSv3 and NFSv4 and have no problem.\

The Debian boxes are mounting via NFSv3 and continually do this:
nfs: server 10.1.1.1 OK
nfs: server 10.1.1.1 OK
nfs: server 10.1.1.1 OK
nfs: server 10.1.1.1 not responding, still trying
nfs: server 10.1.1.1 not responding, still trying
nfs: server 10.1.1.1 OK
nfs: server 10.1.1.1 OK
nfs: server 10.1.1.1 not responding, still trying
....

You can still `ls` and look at the NFS mounts ok. Small mysqldumps work fine, but mysqldumps over a gigabyte or a few hundred megs always fail when dumping to NFS box. Connectivity is fine--a day of pings shows no packet loss.

Any suggestions appreciated. I'm considering mounting via TCP, but that has it's own issues.