Where is the network problem ? ( for nfs )

[Frank Li]

Hi,
When I run nfsstat -c from the client , it seems there is a network problem ( since timeout >> badxids ) according to the HP performance and tuning course . Following is the "nfsstat -c" output :
************************************************************************************
Client rpc:
Connection oriented:
N/A
Connectionless oriented:
calls badcalls retrans
227104 1 340
badxids timeouts waits
1 340 0
newcreds badverfs timers
0 0 423
toobig nomem cantsend
0 0 0
bufulocks
.....
*********************************************************************************

There is no udp overflow on the NFS server . and when I ftp the same file from the NFS server to the client , it's very faster ( maybe 6 times ) than cp this file .

Any body encounters this before ?
Note : Server is 1000M lancard and the client is 100M card !!! both are runing HPUX11.0 .

thanks

[Pedro Sousa]

Hi GDSZLUCENT!
check the following thread:
http://forums.itrc.hp.com/cm/QuestionAnswer/1,1150,0x236fdfe5920fd5118fef0090279cd0f9,00.html

good luck.

[Pedro Sousa]

Another thing...
check for these HP recommended patches:
PHKL_16074 - nfs performance fix ...
PHKL_20335 - critical (includes NFS)
PHNE_20627 - performance patch

[Vincent Stedema]

GDSZLUCENT,

When you're timeouts are rapidly increasing then there might be something wrong with your network. There will always be timeouts due to the fact that HP-UX NFS uses the UDP protocol.

You could reset your nfs statistics (nfsstat -z) and then monitor your timeouts. When the number of timeouts exceeds 5 percent of the number of calls, reduce your wsize and rsize mount parameters to 1024.

How are the NICs configured: half or full duplex?

Vincent

[Brian Hackley]

Hi,

Take a long look at the ITRC NFS Performance Assessement application note, KBAN00000261. Also the ITRC NFS Server Not Responding note NETUXKBRC00006283 is very helpful in this area.

Hope they help,

-> Brian Hackley