It'd help if you read my posts before you presume I don't know what I am talking about. I never said all threads were busy, only that the second number tells you that's what happened. The big number in OP's post (third number and first in the histogram) means exactly what we "both" have said, that a client IO was waiting to be serviced. The delay could be caused by anything in the way of doing IO. In my case for instance, I have lot's of clients doing lots of IO so more threads clears the problem up. In his case , it might that the IO to the disks is not fast enough or needs to be tuned.
If the OP ever plans on having more than one client use this server, then he will very likely want to bump the threads at some point, as well as tune the backend in order to prevent unnecessary queueing. Done... -C On Mon, Oct 10, 2011 at 9:21 PM, Joshua Baker-LePain <jl...@duke.edu> wrote: > On Mon, 10 Oct 2011 at 8:38pm, Corey Kovacs wrote > > > Well, I dind't miss, just didn't mention it. The big number is not the >> 1-10% >> of the threads being busy, it's how much time was spent waiting for a >> thread >> to free up to complete an I/O. You could interpret that as busy of course, >> but "why" it's busy is more important. >> > > Sorry, but that's incorrect. From the NFS FAQ at < > http://nfs.sourceforge.net/>: > "Review the contents of /proc/net/rpc/nfsd, especially the line that begins > with 'th'. The first number on that line is the total number of NFS server > threads that are started and waiting for NFS requests. The second number > indicates whether at any time all of the threads were running at once. The > remaining numbers are a thread count time histogram." > > And from section 5.6 of the NFS HOWTO at <http://nfs.sourceforge.net/** > nfs-howto/ar01s05.html <http://nfs.sourceforge.net/nfs-howto/ar01s05.html> > >: > > "The last ten numbers on the th line in that file indicate the number of > seconds that the thread usage was at that percentage of the maximum > allowable. If you have a large number in the top three deciles, you may wish > to increase the number of nfsd instances." > > So, again, the big number in the OP's "th" line was the number of seconds > in the first decile of thread usage (i.e., not a big deal). And given the > OP's second number (the one right after the thread count) was 0, at no point > were all the threads busy. > > > -- > Joshua Baker-LePain > QB3 Shared Cluster Sysadmin > UCSF > > ______________________________**_________________ > rhelv5-list mailing list > rhelv5-list@redhat.com > https://www.redhat.com/**mailman/listinfo/rhelv5-list<https://www.redhat.com/mailman/listinfo/rhelv5-list> >
_______________________________________________ rhelv5-list mailing list rhelv5-list@redhat.com https://www.redhat.com/mailman/listinfo/rhelv5-list
