[tor-bugs] #22255 [Core Tor/Tor]: Frequent OOM kills of tor process

Tor Bug Tracker & Wiki blackhole at torproject.org
Wed Jun 7 19:01:00 UTC 2017


#22255: Frequent OOM kills of tor process
--------------------------+------------------------------------
 Reporter:  DeS           |          Owner:
     Type:  defect        |         Status:  new
 Priority:  High          |      Milestone:  Tor: 0.3.0.x-final
Component:  Core Tor/Tor  |        Version:
 Severity:  Major         |     Resolution:
 Keywords:                |  Actual Points:
Parent ID:                |         Points:
 Reviewer:                |        Sponsor:
--------------------------+------------------------------------

Comment (by DeS):

 O.k. Good News I added all the parameters to my skript and today - we had
 two OOMs.
 Now bad news. If you will look at the relevant time you will see the logs
 start to be strange.

 Very easy explaination. The skript which collects the stats needs more
 than a minute to run. And cron fires the next one. So we have several
 outputs in the files at the same time.

 Stupid me - I did not put an lock file in - but wait. If the machine is
 that slow - what good would it be to have disk access - so no real
 solution as well. As well not running it via cron would not really help.
 The server has not even reseources to execute date and put it into an
 file.

 The fact is i can try some things but to have clean logs about a
 completely dying machine is like  sending a probe into a black hole and
 expect pictures back.


 But now. Lets step back a little and think.

 a) We see the tor process use around 250-400 MB normally.
 b) When they get killed they are bigger 550 - 600 MB
 c) We have a machine with 3877580 MB (3.8~ GB)

 We can not try to have logfiles from everything but if you ask me this
 will not help. All you would see is the memory usage of tor going up to
 600 MB. So what.
 The Question is - where are around 1-2 Gig of Memory gone?
 Even in the messed up logs you will no monster process consuming this
 memory. The biggest user and growing process is tor. And the question is
 why does it behave like this.
 I do not belive a 25% rise of tcp connection will swallow up all the
 memory so the problem must be somewhere else - if you ask me - in tor

 any chance tor has debug features (memory allocation ...)

--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/22255#comment:56>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online


More information about the tor-bugs mailing list