Hi,
I've got an issue with a machine that runs hot once [a Mailman Python
process] appears to go into a tight loop and consumes 99% of the CPU for a
longer amount of time.
Output from sensorsd shows the temperature over a 15 minute period:
Nov 11 21:44:02 2006 | temp,25.00 | temp,67.00 | temp,27.00 |
Nov 11 21:45:01 2006 | temp,25.00 | temp,68.00 | temp,27.00 |
Nov 11 21:46:01 2006 | temp,25.00 | temp,74.00 | temp,27.00 |
Nov 11 21:47:01 2006 | temp,25.00 | temp,77.00 | temp,27.00 |
Nov 11 21:48:01 2006 | temp,25.00 | temp,78.50 | temp,27.00 |
Nov 11 21:49:01 2006 | temp,25.00 | temp,80.50 | temp,27.00 |
Nov 11 21:50:01 2006 | temp,25.00 | temp,81.00 | temp,27.00 |
Nov 11 21:51:01 2006 | temp,25.00 | temp,87.00 | temp,27.00 |
Nov 11 21:52:01 2006 | temp,25.00 | temp,87.50 | temp,27.00 |
Nov 11 21:53:01 2006 | temp,25.00 | temp,88.50 | temp,27.00 |
Nov 11 21:54:01 2006 | temp,25.00 | temp,89.00 | temp,27.00 |
Nov 11 21:55:01 2006 | temp,25.00 | temp,90.50 | temp,27.00 |
Nov 11 21:57:01 2006 | temp,25.00 | temp,91.00 | temp,27.00 |
Nov 11 21:58:01 2006 | temp,25.00 | temp,92.00 | temp,27.00 |
Nov 11 21:59:01 2006 | temp,25.00 | temp,92.00 | temp,27.00 |
Nov 11 22:00:01 2006 | temp,25.00 | temp,93.50 | temp,27.00 |
At this point the CPU halts. I've tried using CPU throttling in the BIOS (if
CPU hot, 75% idle) to no avail.
I will be removing Mailman from the process list in the short term, but in
order to safeguard against this sort of thing in other code (a friend of
mine noticed that his socket polling code did almost the same thing to his
Mac before he rewrote it, for example), is there anything at the OS level I
can do? Or would I need to write a "cpu hogwatcher" which just takes
"necessary action"?
This particular machine is running 3.6/x86.
Kind regards,
Tor
