Take the 2-minute tour ×
Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

I've got a Fedora 10 system that's exporting some file systems via NFS.

One of the exported file systems is used very heavily on a client system when I run a search engine indexer. The file activity is 100% read (no writes).

Whenever I run the indexer the systems load average skyrockets (up to 25 at points) and rpcbind is often using 98% of the CPU capacity.

On the client system I'm mounting the remote directory with the following options: intr,nosuid,ro,rsize=8192,udp,noatime,bg,nodev

On the server, I'm exporting the directory with the following options: ro,no_subtree_check

I have the NFS server configured for 32 threads.

If it makes any difference, the NFS server is also a NIS server. And the client is bound to it as a NIS client.

Any suggestions on what I can do to reduce the load on the server? Having such a high load just doesn't seem normal to me.

Thanks!

share|improve this question
    
What is the output of nfsstat -o all on both server and client? –  Mark Wagner Mar 28 '11 at 18:56
    
Server: qtemp.net/nfsstats-server.txt Client: qtemp.net/nfsstats-client.txt –  david Mar 28 '11 at 19:15

1 Answer 1

See what ports rpcbind are using lsof -p $(pidof rpcbind)

Use tcpdump to see who is talking on those ports tcpdump -s 1600 -ni eth0 port sunrpc or port 850 you might need to look at other interfaces, including lo, in case it is locally generated traffic.

strace rpcbind to see what it is doing strace -s512 -o/tmp/tr -f -p $(pidof rpcbind)

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.