>
> On Tue, 26 Sep 2000, TTSG wrote:
>
> > Maybe a CPU % of 8-12, NFSops of 400-600. (I haven't gotten
>
> That's nothing. I don't think the filer is to blame for the
> load. Cache ages?
>
Varies. I've seen >60, 7, etc.
>
> > I tried it on Sif. Sif went nuts with file locking. CGI's that need
> > to do file lock seem to take FOREVER (7-10 seconds). If I change the CGI to
> > point to a local filestore, the load goes from maybe 2.X to 6-7.X . The
> > CPU goes 100% with an increased user and kernel.
>
> I don't know how much time you have to spend on this, but you
> might try mounting a set of disk from one of your U1's to the other, to
> see if the problem is with NFS in general or with the filer's
> implementation of NFS.
>
Don't I just introduce another unknown, the 2nd U1s ability to serve
the content? I don't have time to futz, since this is a production system,
but I need to do whats necessary. I'm afraid also that making the 2nd U1 a
NFS server will impact ITS performance. I'm already getting a 15-20% I/O
wait outta it, which is why we are looking towards the Filer.
>
> > Thats what I figured. I need to throw 2 of them, serving 7-8Mb/s
> > content out the front end.
>
> Hmmm...I don't know how much cache a F720 has, but maybe you
> need something with more cache to serve out that much data.
>
I've got 4 BSDI's, each pushing 7-8Mb/s out the front, to another
F720 with no hassles. If it wasn't for the fact that the BSDI dropped in
and went well, I wouldn't even CONSIDER this as a solution.
>
> > > What network pieces are between Thor and the F720?
> > >
> > Thor -> 4 ft network cable -> Cisco 2924 switch -> 8 ft network
> > cable -> Cisco 5505 -> 6 inch network cable -> RJ45/Punchdown block ->
> > 200ft network cable -> RJ45/Punchdown block -> 6 inch network cable ->
> > RJ45/Punchdown block -> 24 ft network cable -> RJ45/Punchdown block ->
> > 8 ft network -> Cisco 2924 -> 8 ft network cable -> F720
>
> 200 ft of network cable? Is that legit, i.e., within the CAT5
> specification?
>
Yea, and there aren't any collisions or symptoms of problems between
the Cat5505 and remote 2924.
>
> Also, are all the speed negotiations fixed? Many people post of
> bad performance which is due to either the filer or the switches being
> set to auto-negoatiate. Force both to be fixed.
>
I don't let ANYTHING in our network auto-negotiate. To me, thats
a non-truth. ESPECIALLY with Sun equipment around. 8-)
>
> > > NFS v2 or v3? If v3, over TCP or UDP?
>
> You say NFS v3 over UDP. You might also try to force this to be
> NFS v2, just to see what happens with the locking, which seems to be the
> source of the bottleneck.
>
Maybe on the Sol 7, not 5.5.1 . I'll try V2 at some point.
>
> You can disable v3 complete on the filer with
> the nfs.v3.enable option, or force a vers=2 mount with the mount
> command.
>
> Until next time...
>
Thanks, Tuc/TTSG