[tor-bugs] #33211 [Circumvention/Snowflake]: proxy-go sometimes gets into a 100+% CPU state

[Tor Bug Tracker & Wiki]

#33211: proxy-go sometimes gets into a 100+% CPU state
-------------------------------------+-----------------------------------
 Reporter:  dcf                      |          Owner:  (none)
     Type:  defect                   |         Status:  needs_information
 Priority:  Medium                   |      Milestone:
Component:  Circumvention/Snowflake  |        Version:
 Severity:  Normal                   |     Resolution:
 Keywords:                           |  Actual Points:
Parent ID:                           |         Points:
 Reviewer:                           |        Sponsor:
-------------------------------------+-----------------------------------

Comment (by dcf):

 Replying to [comment:7 dcf]:
 > I also had it happen again today, with the quic snowflake client from
 #33336. I left the browser idle for about a half an hour, and when I came
 back the fans were spinning and it was using 160–180% CPU. But just as I
 started to investigate, it resolved itself spontaneously and returned to
 normal levels.

 I got this again just now, and this time I happened to be watching the
 log. It happened right after my wifi dropped out, which caused the broker
 HTTP request to stall for 5 minutes before timing out. (Side note, we
 should probably reduce that timeout.) The high CPU happened immediately
 after the "connection" timed out" log message. It lasted for less than a
 minute, then went back to normal.
 {{{
 2020/02/19 23:25:49 WebRTC: Set local description
 2020/02/19 23:25:49 WebRTC: Got ICE candidate: host [scrubbed]
 2020/02/19 23:25:50 WebRTC: Got ICE candidate: srflx [scrubbed] related
 [scrubbed]
 2020/02/19 23:25:50 WebRTC: Done gathering candidates
 2020/02/19 23:25:50 WebRTC: Got ICE candidate: srflx [scrubbed] related
 [scrubbed]7
 2020/02/19 23:25:50 WebRTC: ICEGatheringStateComplete
 2020/02/19 23:25:50 Negotiating via BrokerChannel...
 Target URL:  snowflake-broker.azureedge.net
 Front URL:   ajax.aspnetcdn.com

 2020/02/19 23:30:57 BrokerChannel Error: read tcp [scrubbed]->[scrubbed]:
 read: connection timed out
 2020/02/19 23:30:57 Failed to retrieve answer. Retrying in 10s
 }}}
 Possibly there is a pent-up timer or other recurrent event that creates a
 lot of work while the client is blocking in the HTTP request.

--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/33211#comment:9>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online