Problems with sync NFSv3
Hello
I guess some of you have NFS mounted shares so I like to hear about your
experiences. I'd like to use nfs-kernel-server with "sync" in
/etc/exports but on my setup this is unbelievable slow.
Trying to use sqwebmail on a 200 mail mailbox needed 30 instead of 3s
and even a short
time perl -e 'for (my($i)=0; $i<1000; $i++) {open(F,">t.$i");close(F);}'
took me 5s on syncNFS instead of 0.5s local FS or at least 1s on
asyncNFS.
The network transfer itself only needs ~2MB but I see with vmstat, that
the server writes about 40MB buffer-out. Why? Any ideas how to improve
it?
Does everybody else use async NFS? I thought as the latest NFS package
even defaults to sync it is more recommended. Does anybody really
understand how reliable async NFSv3 (not v2 there are told to be big
differneces) in cases of server or client crashes is? I've heard about
a client side cache that stores every write until it receives a
commit-ok...
bye,
-christian-
P.S.: The vmstat of the above perl skript:
procs memory swap io system cpu
r b w swpd free buff cache si so bi bo in cs us sy id
0 0 0 26804 9684 187304 316504 0 0 0 0 104 17 0 0 100
0 0 2 26804 8596 187304 316504 0 0 0 4484 1679 3704 6 6 88
0 0 2 26804 9656 187304 316504 0 0 0 8556 3213 7404 1 7 92
1 0 2 26804 9656 187312 316516 0 0 4 8160 3135 7096 0 8 92
0 0 2 26804 9596 187312 316516 0 0 0 8660 3244 7449 0 9 91
0 0 2 26804 7752 187312 316516 0 0 0 8216 3090 7093 9 10 81
0 0 0 26804 9376 187312 316516 0 0 0 6780 2579 5907 0 9 91
0 0 0 26804 9400 187312 316516 0 0 0 0 114 34 0 0 100
--
Christian Hammers WESTEND GmbH - Aachen und Dueren Tel 0241/701333-0
ch@westend.com Internet & Security for Professionals Fax 0241/911879
WESTEND ist CISCO Systems Partner - Authorized Reseller
Reply to: