Opened 3 years ago

Closed 3 years ago

#19819 closed defect (fixed)

100% CPU usage and restarts after upgrade from 0.2.8.5-rc to 0.2.8.6

Reported by: hacktic Owned by:
Priority: High Milestone: Tor: 0.2.8.x-final
Component: Core Tor/Tor Version: Tor: 0.2.8.6
Severity: Major Keywords: review-group-7
Cc: Actual Points:
Parent ID: Points:
Reviewer: Sponsor:

Description

Upgraded from Tor 0.2.8.5-rc to 0.2.8.6 on a pi2 using http://deb.torproject.org/torproject.org/dists/jessie/

on start-up it goes to 100% CPU usage and after about 3 minutes default service kills and restarts tor, this goes into a loop, last message in tor log: Bootstrapped 0%: Starting

Child Tickets

Change History (7)

comment:1 Changed 3 years ago by hacktic

debug log, last line:[debug] router_parse_list_from_string(): Read router etc.

as this router parse list is pretty long, could it be that parsing of this list takes longer then the default service timeout and it kills the process via "systemd[1]: tor@… start operation timed out. Terminating." ??

Last edited 3 years ago by hacktic (previous) (diff)

comment:2 Changed 3 years ago by hacktic

Update: couldnt get systemd timeout increased somehow (TimeoutSec=infinity etc.)

ended up deleting cached-descriptors cached-descriptors.new cached-microdescs cached-microdescs.new, restarted and now its running again.

My theory is that once my box runs for a couple of days (it was running for about 1 week earlier) it builds up enough of a cache that on restart it will hit the systemd service timeout, is there a way to limit the cache created ?

comment:3 Changed 3 years ago by hacktic

Component: Core TorCore Tor/Tor

Update 2: dug a bit deeper, checked /lib/systemd/system/tor@… which has TimeoutStartSec=120

I assume if this gets bumped to TimeoutStartSec=500 we might not run into the above issue any longer.

comment:4 Changed 3 years ago by hacktic

Status: newneeds_review

Update 3: when upgrading to 0.2.8.6-2 and even just running a relay for 2 days it will hit this limit.

please increase /lib/systemd/system/tor default.service TimeoutStartSec= from 120 to 300 to fix this issue

thanks

comment:6 Changed 3 years ago by nickm

Keywords: review-group-7 added

comment:7 Changed 3 years ago by dgoulet

Resolution: fixed
Status: needs_reviewclosed

Nothing to do for "tor" here and next 028 package will have the fix.

Note: See TracTickets for help on using tickets.