About once every few days or so, our Linux NFS server (knfsd-1.4.7-7) stops servicing new NFS mount requests. The cause of this seems to be because the rpc.mountd daemon dies. This also occurred with redhat 6.0 (using rawhide's knfs-1.4.7-7). This seems to be similar to BUG #2707. To fix this, all I do is either manually restart the daemon (/usr/sbin/rpc.mountd --no-nfs-version 3) or reboot the server.
What kernel are you using? There are kernel patches that must match with knfsd version (ugh).
It occurs both with the stock redhat-6.1 kernel (2.2.12-20) and with rawhide's 2.2.12-42 kernel. Which would you recommend? (I'm sticking with the "stock" kernel for now).
This seems to be a repeat, or, at least very similar to bug #7024
I've verified the circumstances upon which rpc.mountd dies in that it dies in conjunction with ypserv/ypbind problems (which may lead to my reporting bugs for them). Anyway, shortly before each time rpc.mountd died, I also had in /var/log/messages, the errors: Dec xx xx:xx:xx server ypserv[310]: WARNING: ypserv has to many running children ! We don't fork THEN, got a few of the form: Dec xx xx:xx:xx mountd[28864]: client.domain.com has non-inet addr THEN, after a few more of both of those, rpc.mountd died. As a fix, I chose to NOT use NIS for hostname lookups by editing /etc/nsswitch.conf to include a hosts directive as: hosts: files dns Since making this change (it's been over a week), I have not experienced a single rpc.mountd crash.
assigned to johnsonm
I am happy to report now much later that I have not experienced an rpc.mountd crash since the last time I reported it. For the record, I'm now running: kernel-2.2.19-6.2.1 nfs-utils-0.3.1-0.6.x.1 glibc-2.1.3-22 So, I personally consider this bug closed. I'll mark it resolved by "errata", as I think some combiniation of the upgraded components I mentioned above probably fixed it.