Solution: downgrade to woody for nfs-utils: As I mentioned earlier:
On our file server, AMD Opteron(tm) Processor 852, running Sarge, we have something like 300 clients (PCs on Debian Sarge and Xterminals, based on Sarge servers) and ~ 1000 users. When under big load, the rpc.mountd from sarge becomes unresponsive: As top shows, its %CPU then switches between more than 95% and something like 60%, and it won't give any answer anymore until it is rebooted. (/etc/init.d/nfs-kernel-server restart won't help.) Apparently, this happens already, when the server gets a mount request every two seconds. (I am not sure how many requests it is supposed to handle.) However, we never observed a segfault of rpc.mountd. I should mention that under testing conditions with few users (up to 10 or so) or under little load, the sarge system worked quite well. After downgrading both nfs-common and nfs-kernel-server (version 1:1.0.6-3.1, from sarge) to version 1:1.0-2woody3 from woody the rpc.mountd worked well (though not very fast under big load). Downgrading just nfs-kernel-server (which contains the rpc.mountd program) seemed to work at first glance, but developed a lot of stale file handles after a couple of hours, so we also had to downgrade nfs-common (which belongs to the same software package nfs-utils) in order to obtain a stable system. Ulf Rehmann -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
