Hal Stern said that our nfsd threads are set way too low, we'll bump
it up and watch to see if things improve. I still don't know what
the udpNoPorts numbers mean.
My original posting:
> Our nfs server (solaris 2.3) machines occasionally get really slow so we start
> watching graphs of various network statistics reported by netstat
> to see if they'll give us some tuning clues. We noticed some interesting
> patterns for udpNoPorts and udpInOverflows. I remember some messages
> in this list saying that udpInOverflows should be zero, if not, the number
> of nfsd threads should be increased. We up'ed the nfsd from 32 to 48
> and watched it for a few days and so far the change didn't seem to make
> much difference. These servers support > 700 pc's, not all at once :-),
> what would be a reasonable number to set nfsd to?
> The udpNoPorts seems very consistent across machines so we suspect
> it's just certain kind of traffic on the net that affect
> them all but we have no ideas what it means.
> From stern@sunrise.East.Sun.COM Tue Oct 8 09:12:25 1996
> looks like you're running out of nfsd threads. on a big
> server, you'll need many more than 48; under solaris there's
> no penalty for having too many. try using at least 1,000
> or so, especially with 700 clients
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:11:11 CDT