Hello,
We have a netware-NFS server which exports a filesystem to our Digital UNIX
server. Since a couple of weeks, we're having a tough time with it. It used
to work well for the last 6 months. Usage as slightly increased recently,
users are moving big files to and from UNIX to this NFS filesystem.
The netware server will be upgraded recently, we just want to know if the
problem could be created because the NFS server is not powerful enough to
handle all the requests from UNIX AND Windows clients.
We have to mount the nfs filesystem soft and using NFS V2 (This is a NetWare
requirement)
We also suspect an errant application of messing the things up but we cannot
pinpoint the exact problem. This application is a logger which repeatedly
open and close a log file in the nfs filesystem. It does not write big
amounts of data to that file.
Here are the symptoms of the problem:
Under Windows, we have problems accessing the Netware drive that is exported
to unix. Client PC freeze.
Under UNIX, the systems rapidly becomes unresponsive and we get
RPC_timeouts, I/O errors, etc...
The Netware NFS server complains that que request queue is full and
eventually crash. We rebooted the NFS server and the problem came back 15
minutes later. What we do now is shut down the netware server and unmount
the NFS file system from all UNIX boxes (This is not easy as many process
have files open on that server or many users have there CWD in the exported
directory or below, we get a DEVICE BUSY error and have to kill some
processes..... (BTW: IS there a way to force a dismount insuch situations ?))
I've included a sample of the messages we get in /var/adm/messages when this
happens. Notice that there is a lot of proc/forkdup: task creat failed
messages since a couple of weeks also, I don't know if it is related to the
problem but it tends to occur at approximatly the same time (+ - ~ 1 hr)
Maybe the NFS crash is a side effect of the procdup task creat problem, or
vice-versa.
If anyone has a fix, or hints, or ideas on how to find the culprit and fix
the problem, any help would be GREATLY APPRECIATED. Our users are
complaining and we have to ship a system in production soon.
I've attached the /var/adm/messages sample along
Thanks !
Guy Dallaire
dallaire_at_total.net
"God only knows if god exists"
Received on Fri Jan 10 1997 - 20:52:58 NZDT