* Christopher S. Swingley ([EMAIL PROTECTED]) spake thusly: > Hi! > > I have a couple raid arrays (software raid, 2.4 kernel) that I've > been exporting to a variety of other boxen where I work. Now that > more than a few machines have these mounted (> 10, < 25), NFS has > become incredibly sluggish. It's almost impossible to edit files on > the imported filesystems because the writes hang for several seconds > every so often interrupting my typing.
FWIW it happens here with SPARC Solaris 8 NFS server. I wouldn't call it "incredibly sluggish" but every once in a while my workstation hangs for a second or two on NFS writes. >And remote logins on machines > where my home directory is imported can take 5-10 seconds before the > machine can set my home directory to the NFS mount. Doesn't happen here. > I've tried both the user space and kernel space nfs packages, and am > currently exporting with these options: > > (rw,root_squash,no_subtree_check) > > and mounting with these options: > > rw,bg,hard,intr,rsize=8192,wsize=8192 > > (although I've also tried it without the rsize and wsize options) You want to keep those [r|w]size options, they help. ... > Any suggestions / other options? Do I need to further tweak the > kernel somehow, or change some network settings? What sort of network is it? Ours is switched 100BT, but most workstations are on 10MB/s (ancient wiring). The server is on 100FDX, and when I work on another 100FDX'ed box there's no delay on NFS writes. So maybe you should look at your hub, check link modes and all that. Dima -- E-mail dmaziuk at bmrb dot wisc dot edu (@work) or at crosswinds dot net (@home) http://www.bmrb.wisc.edu/descript/gpgkey.dmaziuk.ascii -- GnuPG 1.0.4 public key We're sysadmins. Sanity happens to other people. -- Chris King in asr
