[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]
Re: -current nfs server problem
On Sun, 13 Jan 2008 00:12:19 +0200
Arto Huusko <arto.huusko%pp2.inet.fi@localhost> wrote:
> there is something broken with nfsd on sparc64 on -current.
> On 4.99.22 the nfs server worked fine. I first noticed the
> problem with 4.99.34, and it still persists with 4.99.49,
> sources updated a few hours ago.
> I don't know if this affects other platforms.
> Problem number one is that nfsd over udp does not work at
> all. I just did a clean reboot of the server, without starting
> nfsd in rc. After that, I started nfsd with
> nfsd -u -n 4
> No messages were printed, and exit code from the command is 0.
> However, after this there are no nfsd processes in the system.
> Despite this, is see
> udp 0 0 *.2049 *.*
> in netstat -an output. But nfs volumes can't be mounted (well,
> no nfsd processes, so I'm not surprised).
> nfsd -t -n 4
> does work, and I can also mount nfs volumes from clients.
> However, things don't work at all well. I'm trying to build
> a new system, with pkgsrc mounted from this server, and
> everything hangs up very badlym at least when writing to the
> nfs volume. Also large copies from the nfs server to client
> stall. I have used NetBSD 4.99.49 and Mac OS X 10.4 as
> clients, and both clients fail in similar ways.
NFS on -current sparc64 works for me with udp mounts, except that I feel
that the CPU usage of nfsd has been unusually high for some time
(probably since around 4.99.34.).
I also get the impression that the slower the client and/or network is,
the higher the load on the server becomes. Maybe it's busy-waiting
on some lock, I don't know.
Does it make a difference if you start a single nfsd (-n 1)?
Main Index |
Thread Index |