[mythtv-users] [mythtv] mythcommflag taking all the cpu 0.19 svn revision 9806
Osma Ahvenlampi
oa at iki.fi
Mon May 8 03:10:19 EDT 2006
On su, 2006-05-07 at 08:07 -0400, Khanh Tran wrote:
> On 5/6/06, Chris Pinkham <cpinkham at bc2va.org> wrote:
> > * On Sat May 06, 2006 at 10:38:28PM -0400, Khanh Tran wrote:
> > > Now I don't know what the problem is. I removed those entries and I
> > > completely disabled mythcommflag jobs on the frontend/backend in
> > > question. Xine still pauses, however not nearly as bad as when
> > > mythcommflag is running...
> > Are you sure you're using Xv output? If you're having to do
> > video scaling in software and not hardware, that would slow down
> I'm not sure what's happening. Xine has never been an issue for me
> until I went to the SVN head branch. I'm not so sure it's
> mythcommflag now, due to my previous update. What I am noticing now
> is that the whole box seems to freeze for a couple seconds every so
> often.
You never actually stated the machine's real state during this time that
I noticed. The loadav of 3.2 is mostly meaningless in this, as you're
running different kinds of loads simultaneously, and as such the problem
may be something totally different than cpu overload. Look at CPU
user/sys/nice/idle/wait and memory + swap allocation. My guess is that
you might be seeing mythcommflag hog memory and cause the machine to go
to swap - I regularly see this happening with both commflag and
transcode.
Try running 'vmstat 10' in a terminal in the backgroun. It'll show you
system state every ten seconds like this:
procs -----------memory---------- ---swap-- -----io---- --system-- -----cpu------
r b swpd free buff cache si so bi bo in cs us sy id wa st
2 0 0 112076 68796 486592 0 0 222 78 381 771 9 2 84 4 0
0 0 0 112076 68808 486596 0 0 16 0 342 494 1 0 98 1 0
^ little swap use ^ should have memory left over for disk cache
^ ^ should not have swap activity
^ ^ should have disk i/o
^ the average of this column is 'loadav' ^ ^ ^
most cpu used should be us(er)
should have id(le) cpu
should not have a lot of wa(iting for io to finish) cpu
If I'm correct, you'll be seeing activity in the swap in/out and cpu
wait columns. A couple of percent of cpu wait is fine, but under a swap
storm you'd see anything up to high-80's.
--
Osma Ahvenlampi <oa at iki.fi> http://www.fishpool.org
More information about the mythtv-users
mailing list