As some of you might noticed, Trac is sometimes not available or has
pretty bad response times, without any visible reason. (No notices or
hints about this in the Trac-, Apache- or Syslog)

The only thing I can see, that the Trac process hogs ~100% CPU time
quite often and sometimes, it seems to be stuck. However, this night the
Server died due to a out of memory, as you can see on this [1] picture.

I'm very tempted to blame Trac for this as well, I have no prove since
there wasn't enything in the logs at all, but this problem seems to be
known, as you can see on this [2] picture.

It was everytime Trac that hogged CPU time and memory, but in the past I
noticed it just in time and could fix it by rebooting the Apache.

Also, here [3] is a small picture that demonstrates the impact of Trac,
before the out-of-memory trac was enabled and after the OOM I disabeled
Trac to see what changes. Please don't reply with "You have < 20% CPU
usage and already start crying?" posts, I don't and thats just fyi ;)

To make a long story short, I want to hear your opinions about this. I'm
somewhat unhappy about Trac right now, and looking for solutions or
workarounds. To be precise, I don't think that it is such a great idea
to give Trac such a central role besides the Bugtracker.

[1] http://files.wz2100.net/temporary/trac_memory_day.png
[2] http://files.wz2100.net/temporary/trac_memory_month.png
[3] http://files.wz2100.net/temporary/trac_cpu.png

-
Best regards,
Kamaze

_______________________________________________
Warzone-dev mailing list
[email protected]
https://mail.gna.org/listinfo/warzone-dev

Reply via email to