Why is my load ave so high now?

Aaron Konstam akonstam at sbcglobal.net
Fri Jul 24 20:41:41 UTC 2009


On Fri, 2009-07-24 at 11:08 -0400, Kevin J. Cummings wrote:
> On 07/24/2009 10:15 AM, Aaron Konstam wrote:
> > On Thu, 2009-07-23 at 22:51 -0400, Kevin J. Cummings wrote:
> >> When I was running F8, my server averaged a load ave oof around 4.
> >>
> >> Now that I'm running F10, and bittorrent is no longer running, in
> fact,
> >> not much of anything besides Seti at Home  (BOINC client running
> >> astro_pulse), my load average is up around 11 and frequently
> exceeds 12
> >> (and of course when it exceeds 12, it stops receiving emails).
> Here's a
> >> 5 second snapshot from top:
> >>
> >>> top - 22:48:06 up 2 days, 18:51,  5 users,  load average: 11.15,
> 11.33, 11.63
> >>> Tasks: 250 total,   2 running, 247 sleeping,   0 stopped,   1
> zombie
> >>> Cpu(s):  5.9%us,  1.8%sy, 92.0%ni,  0.0%id,  0.0%wa,  0.3%hi,
> 0.0%si,  0.0%st
> >>> Mem:   2074172k total,  1932680k used,   141492k free,   108872k
> buffers
> >>> Swap:  3911816k total,      552k used,  3911264k free,   977568k
> cached
> >>>
> >>>    PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+
> COMMAND
> >>>   3234 root      39  19 50292  46m 2072 R 91.7  2.3   3520:10
> astropulse_5.06
> > What is astropulse that is using 91.7% of your cpu? You also seem to
> > have a large number of tasks running and sleeping which might up
> your
> > load average.
> 
> astropulse is the SETI at Home BOINC client that I run (NICEd to 19).
> It only uses "excess cycles" and in the past my load average has
> never 
> exceed the 3-5 range, except when I was doing "real" work on the
> system 
> (like running firefox, thunderbird, and other real programs),
> something 
> I almost never do anymore since I bought myself a laptop.
> 
> Yeup, but not consuming any real CPU resources.
> 
> I guess what I'm asking is if only 1 job is grabbing most of the CPU, 
> then what's causing the system to thrash?  (Is a load average of 12 
> considered a thrashing system?  sendmail thinks it is.)
Two suggestions:
1. run vmstat 2 30  
to see how many context switches are occurring and the wait time for
processes etc. A load time of 11 means there are a large number of
processes waiting for cpu time. I think it is inaccurate to say no cpu
resources are being used.

2. If you can get a hold of the article on the Real Time Scheduler found
in the August 2009 issue of Linux Journal. To me astropulse is running
away with your CPU time and noting else can get cpu adequate cpu time.
--
=======================================================================
About the only thing we have left that actually discriminates in favor
of the plain people is the stork.
=======================================================================
Aaron Konstam telephone: (210) 656-0355 e-mail: akonstam at sbcglobal.net




More information about the fedora-list mailing list