Re: NFS server slow when more clients access it
Lukas Kubin said:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> I'm having troubles with the speed of nfs-user-server, when the number of
> computers booting root system from it exceeds the approximate number of
> 5--7. The lower number is quite fast. The boot procedure of Debian "base"
> then takes about 15 to 20 minutes (tested for 17 computers booting
> together). Is it a feature of NFS or is something going wrong?
>
> Details:
> 1. I boot kernel from floppy
> 2. this is the lilo's append:
> "rw nfsroot=CORRECT_IP_ADDRESS:/home/nfsroot,rsize=8192,wsize=8192
> ip=::::::'dhcp'"
> 3. The system init executes just the items in rcS of Debian base system
> 4. I can't use the nfs-kernel-server since I couldn't access the
> server's mounted drives when tried it
any particular reason your using nfs-user-server over nfs-kernel-server?
just curious is all.
What kernel(on both client & server)?
I use nfs-kernel-server so it may not apply but how many nfs processes
are you running? my /etc/init.d/nfs-kernel-server has this
line in it:
RPCNFSDCOUNT=16
# Number of servers to be started up by default
(I think I changed it from the default)
I have noticed in my recent casual tests that the nfs server on
redhat 7.3 is much more solid then the one on debian 2.2. I don't
know if its the kernel(since the kernel runs the nfs server) or if
its the userspace stuff(since rpc.statd is what dies on me in debian,
it's a userspace program). But I've had rpc.statd die half dozen times
if not more in the past week, and the redhat box(7.3) hasn't so much
as hiccuped once. both servers only serve 2 clients, and only 1 client
is used at a time, both systems have over a gig of ram so load is nil.
linux isn't alone in having a buggy nfs. On solaris I came accross a
NFS bug that was documented in sun's bug database more then 5 years
ago and Sun still hasn't managed to fix it yet. It's a pretty bad
bug that takes down the entire NFS system(though not the OS).
nate
Reply to: