High load by HAL on Sun X4200

[Brodie, Kent]

Weird- my ORIGINAL email never made it to the list.   Here it is---  a
very very weird HIGH LOAD problem caused by hal---   resolved a day or
so later with .fdi file(s) that exclude the virtual devices.

 

I think the email problem was my inclusion of "lshal" output, which was
pretty big...    not included this time.    Message edited a little
since I found what was happening-  but important to include this info
for the archives.   -kcb

 

The big question is:   Why is hal freaking out with the virtual
devices??

 

 

Hi all!  I'm new to the list, and seeking help (aren't we all?) 

 

Here's a summary of my problem:

 

OS: RedHat Enterprise Linux AS 4 (kernel 2.6.9-42.0.3.ELsmp)

HAL:  hal-0.4.2-4.EL4 (standard redhat distribution- their latest
version)

System: SUN Sunfire X4200

 

The issue is that HAL is causing load spikes.    Over time, (and on a
completely IDLE system), the system load starts near zero, spikes up to
a load of about 1 for several minutes (as in, between 10 and 20
minutes), then drops back down slowly to zero.   Lather, rinse, repeat,
about 2 to 3 times an hour.   Not good.

 

After spending DAYS on this and trying to figure what was making my load
go up (yet, no "top" processes consuming actual cycles that I could
see), I finally determined that HALDAEMON is most definitely the cause.
Here's a visual performance graph ---
http://cochran.hmgc.mcw.edu/hobbitgraph.png    The load drop to zero
near the end reflects shutting down haldaemon.

 

Further, I suspect it MIGHT be some kind of interaction between hal's
device probes, and possibly the Virtual devices that the X4200 serves up
(i..e, virtual cdrom, virtual floppy) courtesy of the ILOM/console
environment.    Either that, or confusion about the mptsas-served
devices.   I have other (new) Sun boxes, configured exactly the same way
(RedHat AS 4, etc..), but they do not exhibit this symptom-it's *only*
my X4200's that do this.     [Editor-CONFIRMED - it was the virtual
devices...]

 

While "disabling" haldaemon certainly makes the problem go away, that's
obviously not an acceptable solution to the problem--  but it does at
least help verify what's going on........

 

My *guess* is that I can probably solve this by disabling the device
probing for whatever device(s) are causing me grief.    I've already
tried disabling or tweaking tons of BIOS options (ACPI, etc)-to no
avail.     [Editor-resolved temporarily with .fdi file that excludes the
virtual floppy and cdrom]

 

I'd really appreciate any insight on this (as would probably RedHat
and/or Sun......)  -- help!  :-)

 

 

---------------------------------------------------------

Kent C. Brodie - brodie at phys.mcw.edu

Department of Physiology

Medical College of Wisconsin

(414) 456-8590

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.freedesktop.org/archives/hal/attachments/20061220/292d0fbe/attachment.htm