lkml.org 
[lkml]   [1999]   [Feb]   [21]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
Subjectknfsd : some questions
Date

Hello,

Since NFS (server & client sides) would be for us the main reason (not
to say the only one) to switch from the 2.0.36 kernel to the 2.2.x serie,
I am playing with it and have some questions.

We have a medium sized network with around 500/600 NFS machines beeing
able to mount each NFS server (even if, by now, for the Linux server, we
have managed to have no more than 50 machines accessing the NFS partitions
at a given time).

1) using knfsd, is it necessary to run kexportfs -a in the startup
script. Here is what I am currently using :

/etc/rc.d/init.d/nfs :
[...]
case "$1" in
start)
# Start daemons.
echo -n "Starting NFS services: "
daemon rpc.kmountd
daemon rpc.knfsd
/usr/sbin/kexportfs -a <- needed ?
echo
touch /var/lock/subsys/nfs
;;
[...]

My worry is that, when rebooting, knfsd keeps the current
estimated state of the remote mountages in its files
/var/lib/nfs/xtab (and rmtab). When running kexportfs -a,
the command blocks for a very long time, depending on how
many entries there were in theses files (typically more
than 110 entries in xtab, 60 in rmtab). Starting from
fresh (empty files) is almost instantaneous.

My /etc/exports file is :

/u0 @urd(rw)
/u1 @urd(rw)

If I do not run kexportfs -a, I can not mount partitions
via NFS, so I really need to run it. But 10 to 15 minutes
spent on the command is too long.

Has someone else seen the problem.

2) The servers may have to reboot (we are doing that perdiodically,
once every week sometimes). I have observed bad behavior from
the 2.2.x Linux clients when one of their server is rebooting
(the load gets very high, and the machine is almost unusable).

Feb 2 14:42:52 scar kernel: nfs_revalidate_inode: /// getattr failed, ino=2, error=-13

Feb 20 12:44:32 willy kernel: nfs_revalidate_inode: /// getattr failed, ino=1879050241, error=-5

Feb 12 14:09:23 dgps6 kernel: nfs: task 13833 can't get a request slot
Feb 12 14:09:23 dgps6 kernel: nfs: task 13978 can't get a request slot
Feb 12 14:09:25 dgps6 kernel: nfs: task 13928 can't get a request slot
Feb 12 14:09:31 dgps6 kernel: nfs: task 14012 can't get a request slot
Feb 12 14:09:31 dgps6 kernel: nfs: task 14047 can't get a request slot

3) We plan to replace an old Sun IPC which serves as DNS server,
NIS master, sendmail router (and deliver), samba, xntp, UUCP, plus
other things, (and which has a propension to CPU Panic up and
there these latests weeks) by a Linux box.
We need NFS performance for mail delivery (via procmail), so
for me the only reasonnable option is to use a 2.2.x kernel.
I guess we will have to wait a few more weeks for things to
stabilize, but what is your opinion ?

A+
--
Thierry Danis

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:50    [W:0.634 / U:0.096 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site