Christian Hammers wrote:
On Tue, Oct 22, 2002 at 12:25:26PM +0000, Doug Clements wrote:
I mount with rw,-b,-i,-T,-3,-r=8192,-w=8192
I export with  -maproot=0 -alldirs
You are not using Linux right? I cannot find these options in mount(8), nfs(5), exports(5).
I am using FreeBSD.

rw is read/write
b is "fork to background if initial mount fails"
i is "make mount interruptable"
T is "use TCP instead of UDP"
3 is "force NFSv3"
w and r are the same for linux

I also have jumbo frames enabled on interfaces which use NFS at 9000 bytes.
how?
in my rc.conf (FreeBSD-style interface configuration), I have this:
ifconfig_em0="inet 192.168.5.2 netmask 255.255.255.252 media 1000baseSX mediaopt full-duplex mtu 9000"

rc basically sticks the arguments in quotes in an ifconfig command, and the option that does the mtu is "mtu 9000". I'm also using the intel gigabit nic drivers, but I'm hoping this problem doesn't go that far.

I just time that it's at least an NFS problem and independend of
webmail. Took me 0.081s on the harddisk.

# time perl -e 'for (my($i)=0; $i<1000; $i++) { open(F,">t.$i"); close(F); }'
real    0m11.907s
Over NFS:
real    0m3.198s
user    0m0.092s
sys     0m0.289s

Locally:
real    0m2.069s
user    0m0.115s
sys     0m0.200s

Looks maybe like it's writing the cache file?
Yes sqwebmail-curcache.gdbm with entries like this:
  FILE1034549358.10387_3.mail3b192REC193FILENAME=1034549574.10647_3.mail3b:2,
  FROM=Richard Stallman
  SUBJECT=Bitkeeper outragem, old and new
  SIZES=3K
  DATE=1034549302
  SIZEN=3090
  TIME=1034873350
  INODE=192395
Maybe investigating nfsstat to examine errors and retries would be a good idea?

--Doug


Reply via email to