15:03:31 <karsten> #startmeeting metrics team
15:03:31 <MeetBot> Meeting started Thu Dec 10 15:03:31 2020 UTC.  The chair is karsten. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:03:31 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic.
15:03:35 <karsten> hi!
15:03:43 <karsten> first meeting after an extended break!
15:04:09 <karsten> I added all issues to the pad that I'm currently (supposed to be) working on.
15:04:44 <gaba> nice to have metrics meetings back!
15:04:47 <karsten> if we want to talk about more topics, let's add them to the agenda, too.
15:04:52 <karsten> yes. :)
15:05:14 <gaba> it may be a short one today
15:05:57 <karsten> gaba: did you see my message on ipv6 stats options?
15:06:06 <gaba> karsten: i couldnt read it yet
15:06:30 <karsten> okay. can you take a look later and let me know if you need anything else?
15:06:33 <gaba> i will check it later. but you can talk tomorrow if we still need to check if there is anything else missing?
15:06:36 <gaba> ok
15:07:11 <karsten> I can also be around after this meeting. easier than come in for another meeting tomorrow.
15:08:09 <karsten> dennis_jackson: one thing you might care about is tpo/metrics/analysis#33076.
15:08:45 <dennis_jackson> Yes, I was just catching up on this
15:08:46 <karsten> I haven't looked in detail yet, but I think the question is why those measurements with clear 10-second delays did not show up in shadow simulations.
15:09:18 <gaba> karsten: not sure if you have time to reply to nick here so we can add this issue to the january roadmap https://gitlab.torproject.org/tpo/core/tor/-/issues/40224
15:09:27 <karsten> gaba: will do.
15:10:00 <dennis_jackson> As far as I remember, Shadow takes measurements with the exact same onionperf binary? So presumably not a configuration or measurement problem.
15:10:32 <karsten> not exactly onionperf, but same tgen binary, AFAIK.
15:11:05 <karsten> one theory is that these delays are caused by DNS lookups going wrong or exits being overloaded because of too many DNS requests.
15:11:39 <acute> hi! sorry for being late, another meeting overlapped
15:11:40 <karsten> which is something that shadow simulates, so it doesn't have the same characteristics.
15:11:48 <karsten> hi acute!
15:12:11 <karsten> dennis_jackson: but this is based on hearing this just once in a voice meeting and not having time to look at the (long) issue again.
15:12:12 <dennis_jackson> I wonder if we/I could use Arthur's historical exit DNS checks to confirm or refute that.
15:12:39 <karsten> where would we get those?
15:12:57 <karsten> (I know there's tickets for adding that data to metrics, but that didn't happen yet.)
15:13:29 <dennis_jackson> https://arthuredelstein.net/exits/
15:13:40 <dennis_jackson> There's a link to historical data as a JSON
15:13:47 <karsten> acute: http://kfahv6wfkbezjyg4r6mlhpmieydbebr5vkok5r34ya464gqz6c44bnyd.onion/p/tor-metricsteam-2020.1-keep <- if anything stands out as something we should discuss, let's do that.
15:13:51 <karsten> nice!
15:14:25 <dennis_jackson> I think it tries visiting example.com and then the IP for example.com in succession. So DNS behaviour should only impact half the results
15:15:49 <karsten> okay. worth looking into.
15:16:28 <karsten> gaba: regarding ipv6 stats,
15:17:03 <karsten> relays need to run recent enough versions, support IPv6, and report these statistics which are off by default.
15:17:28 <gaba> im  not totally sure if they are off by default from reading the proposal and dgoulet's reply
15:17:39 <gaba> that is why we should talk with them to be sure this is not a bug
15:18:28 <karsten> ConnDirectionStatistics are 0 by default.
15:19:00 <karsten> anyway, they're on the thread, too.
15:19:35 <gaba> ok. I will add ConnDirectionStatistics to the blogpost but I will check with them tomorrow to be sure there is no misunderstanding.
15:19:47 <karsten> sounds good.
15:20:10 <karsten> acute: tpo/metrics/onionperf#40012
15:20:26 <karsten> do you mind taking a look at that?
15:21:15 <acute> no problem
15:21:40 <karsten> great! thanks. :)
15:21:48 <karsten> it even comes with a patch.
15:21:59 <acute> have reassigned it just now :)
15:22:08 <karsten> cool!
15:22:41 <karsten> what else should we talk about?
15:23:00 <gaba> nothing else from me
15:23:17 <gaba> we will have only 1 more meeting this year and then back the second week of january
15:23:35 <dennis_jackson> I was wondering if there was a ticket for the postgresql database?
15:23:48 <karsten> which one?
15:24:11 <karsten> tpo/metrics/onionperf#33259
15:24:12 <karsten> that one?
15:24:45 <dennis_jackson> Ah I think I misread, it's just a funding proposal I think
15:24:57 <dennis_jackson> I was reading back through the previous meeting agendas
15:25:28 <dennis_jackson> "O1.2 Evaluate measurement data using a big PostgreSQL database"
15:25:57 <karsten> right. no concrete plans for that right now.
15:26:03 <karsten> it would be fun, of course. ;)
15:26:09 <karsten> but that's not enough.
15:26:47 <karsten> alright, next meeting next week?
15:27:26 <gaba> yes
15:27:46 <karsten> awesome! have a good week, everyone! bye! o/
15:27:49 <dennis_jackson> o/
15:27:52 <acute> o/
15:27:59 <karsten> #endmeeting