NFS client performance differs by mount point
 

I have an odd situation. I have a Linux server with an NFS mount over TCP to a NetApp filer. There are various application logging to the /logs mount point, which is where the filer is mounted. lsof /logs shows 250 things are open (exactly 250, if that is significant). When I mount the exact same volume to a temporary directory at the exact same time, my read performance goes from 70KB/s or so to 150+MB/s. A huge, huge difference. I'm at a loss as to why. Do any of you have any ideas?

Hey robwicks

Can you provide some more infos, e.g. the output of in both situations (mounted on /var and mounted on other directory) and the content of /etc/fstab?

mr51m0n

Hi,

When you have the volume mounted on the temporary mount point, are all of the applications logging to it? I expect with so many programs having open files when it's your /log mount point, that you simply have a limited remaining bandwidth to the disk. The question that mr51m0n asked is relevant, in that the mount options can make a difference. But to really compare the two mount points, you must also reproduce all of the traffic.

--jason