[EMAIL PROTECTED] wrote on Tue, 29 Aug 2006 17:05 +0530:
> I think what I saw yesterday was that after I had run some benchmarks on
> pvfs2-IB, the pvfs2-client-core was showing 100% cpu activity for a
> couple of hours after that. Things are fine today.
>
> Im running pvfs2-1.5.1.
If you continue to have high-load problems on either the client or
server, please take a look at the CVS head. Just after the 1.5.1
release, I added code to enable blocking mode on VAPI, so that both
client and server will fall back to an in-kernel blocking poll()
when there has been no network activity for over 10 ms. Then,
prompted by your query, I added a fix yesterday to reduce the load
further.
I took a look at the diff from 1.5.1 to HEAD and there's too many
changes already to be able to send you a small patch against 1.5.1
with any confidence. Or it shouldn't be too long until 1.6.0 if you
can struggle along until then.
-- Pete
_______________________________________________
Pvfs2-users mailing list
[email protected]
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users
