Subject: Re: nfsds using all CPU, STATE = "vnlock"
To: Johan A. van Zanten <firstname.lastname@example.org>
From: Bill Stouder-Studenmund <email@example.com>
Date: 10/14/2007 18:26:26
Content-Type: text/plain; charset=us-ascii
On Sun, Oct 14, 2007 at 06:39:35PM -0500, Johan A. van Zanten wrote:
> Bill Stouder-Studenmund <firstname.lastname@example.org> wrote:
> > Pay attention to the ones not in "vnlock". Processes in "vnlock" are=20
> > sleeping, waiting on a vnode lock. They thus aren't directly consuming=
> > CPU. You really want to know what the vnode lock holder is doing while=
> > this is happening.
> When i've noticed it in top, 3-4 of the total 4 nfsds were in vnlock.
> Would they appear to be using a lot of CPU if it's a spin lock?
No. "vnlock" means they're sleeping.
It could be that one of them is grabbing the lock & doing lots of work,=20
with the others piling up on the lock. And it just rotates which one is=20
doing all the work.
> > I'm going to blame softdeps. Can you try things w/o them?
> Will do, as time permits.
> > > The system seems to recover, but i have noticed some of the recorded
> > > programs have lost substantial chunks of data, though that could
> > > also be due to many other factors, such as antenna reception.
> > Some sort of cleanup is happening and gobbling lots of CPU.
> Yeah, the notion i had is that the recording software buffers some, but
> then just starts dropping data as the writes queue.
It could be that we're not doing a good job of pushing data out.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (NetBSD)
-----END PGP SIGNATURE-----