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 have 6 NFS Mounts (1GbE) on a RHEL 5.7 system with 10GbE.

Iperf shows full link saturation to all 6 interfaces simultaneously.

There is a bottleneck somewhere within the nfs client system, but I cannot find it.

It can write to 3 of the mount points simultaneously with full link saturation, but when this number is increased to 6, each point only writes at about half of link speed.

I am suspicious about the number of nfsiod threads, but cannot find this option/setting anywhere.

The problem also appears to exist on Ubuntu, but manifests differently - it completes several transfers at full speed, and some with degraded performance. On RHEL, the performance of all transfers is degraded.

How can I troubleshoot further? Thanks.

share|improve this question
    
is rhel 5 clien or server? do you use nfs 3 or 4? wh9ch wsize an r size are used. which fs? –  tigran Mar 4 '13 at 22:38
    
nfs3, rhel5 client. thanks for your help. tried various wsize/rsize with no impact. –  Brandon Mar 4 '13 at 23:05
    
looking at top, only 3 rpciod processes at a time are having much utilization, maybe a clue. –  Brandon Mar 5 '13 at 0:18
    
What are the characteristics of underlying storage? Number of disks, RPM, SATA/SAS, storage controller? How high is CPU utilization on the server? –  dusty Jun 20 '13 at 13:16

Your Answer

 
discard

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

Browse other questions tagged or ask your own question.