#26121: BUILDTIMEOUT_SET totals are still off
------------------------------+--------------------------------
Reporter: mikeperry | Owner: mikeperry
Type: defect | Status: assigned
Priority: Medium | Milestone: Tor: 0.3.4.x-final
Component: Core Tor/Tor | Version:
Severity: Normal | Keywords:
Actual Points: | Parent ID:
Points: | Reviewer:
Sponsor: |
------------------------------+--------------------------------
While testing and examining onion service timeout rates and trying to
verify them from a controller script, I realized that our timeout rates
are still off in the BUILDTIMEOUT_SET event.
The reason for the discrepancy is that we're double-counting circuits that
transition into MEASUREMENT_EXPIRED. Every circuit that transitions into
MEASUREMENT_EXPIRED is first counted as a timeout in
circuit_build_times_mark_circ_as_measurement_only(). Of those that do not
complete within the measurement window, we again count as a "closed"
circuit in circuit_build_times_count_close(). If a measurement circuit
succeeds, we do *not* count it as a success (see
circuit_build_times_handle_completed_hop()).
This means that the total_circuits value in
cbt_control_event_buildtimeout_set() should be timeouts+succeeded, and not
timeouts+succeeded+closed. (Again, counting closed in the total there
double-counts "closed" MEASUREMENT circuits, which were also counted as
timeout circuits earlier).
Since this is just a control port stats change, it would be nice to get it
into 0.3.4 (and maybe 0.3.3) so it is easier to get more accurate data
about how the circuit build timeout is interacting with vanguards there.
--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/26121>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online
_______________________________________________
tor-bugs mailing list
tor-bugs@lists.torproject.org
https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-bugs
