Hi All,
I'm investigating a problem where our Amanda gets a "data timeout" on
the root partition of our fileserver.
First: while estimating I see:
/tmp/amanda# psg dump
amanda 15587 0.0 0.7 2052 976 pts/7 S 12:52 0:00 sh /usr/bin/amdump bu
amanda 15596 0.0 0.6 2240 868 pts/7 S 12:52 0:00 dumper
amanda 15597 0.0 0.6 2240 868 pts/7 S 12:52 0:00 dumper
amanda 15944 0.5 1.0 2020 1340 ? S 13:28 0:00 dump 1sf 1048576 -
/dev/hdc1
amanda 15945 0.2 0.9 1984 1268 ? S 13:28 0:00 dump 1sf 1048576 -
/dev/hdc1
amanda 15946 0.2 0.9 1984 1268 ? S 13:28 0:00 dump 1sf 1048576 -
/dev/hdc1
amanda 15947 0.2 0.9 1984 1268 ? S 13:28 0:00 dump 1sf 1048576 -
/dev/hdc1
amanda 15949 1.4 0.6 1912 820 ? D 13:28 0:01 dump 2sf 1048576 -
/dev/hdc1
root 15983 0.0 0.3 1236 452 pts/10 S 13:30 0:00 grep dump
/tmp/amanda# kill -HUP 15944 15945 15946 15947
/tmp/amanda# psg dump
amanda 15587 0.0 0.7 2052 976 pts/7 S 12:52 0:00 sh /usr/bin/amdump bu
amanda 15596 0.0 0.6 2240 868 pts/7 S 12:52 0:00 dumper
amanda 15597 0.0 0.6 2240 868 pts/7 S 12:52 0:00 dumper
amanda 15949 3.3 0.8 1912 1084 ? D 13:28 0:04 dump 2sf 1048576 -
/dev/hdc1
root 15985 0.0 0.3 1232 424 pts/10 S 13:30 0:00 grep dump
/tmp/amanda#
What I THINK happens is that Dump says "estimaged XXXX tape
blocks". Then Amanda tries to kill dump, but before that is done, dump
forks the worker processes. I see the log saying something about
killpgrp, so I agree that they should all get killed. However it's not
happening!
This tremendously slows down the estimate process: there are a bunch
of dumpers still dumping to /dev/null while the next estimate (usually
on the same disk, or even the same partition) runs.
This possibly also slows down the actual dump enough for us to get
timeouts.
Possible solution: Open a pipe instead of connecting the output of the
dumper to /dev/null. Closing the pipe should pretty reliably kill off
the dumpers.
Roger.
--
** [EMAIL PROTECTED] ** http://www.BitWizard.nl/ ** +31-15-2137555 **
*-- BitWizard writes Linux device drivers for any device you may have! --*
* There are old pilots, and there are bold pilots.
* There are also old, bald pilots.
