16:01:27 <karsten> #startmeeting metrics team
16:01:27 <MeetBot> Meeting started Thu Mar 14 16:01:27 2019 UTC.  The chair is karsten. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:01:27 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic.
16:01:35 <karsten> - Sponsor 13 report (irl)
16:01:42 <karsten> can you give an update how this is going?
16:02:01 <karsten> was the exonerator/rserve stuff of any use?
16:02:06 <irl> report is in progress, it is tiring work, it is about 65% done
16:02:15 <irl> the exonerator/rserve stuff was great
16:02:18 <karsten> can I help?
16:02:33 <karsten> by doing more braindumps? by reviewing?
16:02:42 <karsten> by doing code metrics?
16:02:42 <irl> i think that right now you cannot help, but you may be able to review sections as they are completed later
16:02:49 <karsten> okay.
16:03:01 <karsten> when do you think you'll have something to review for me?
16:03:06 <irl> tomorrow evening
16:03:11 <karsten> sounds good!
16:03:45 <karsten> sorry that it's tiring work.
16:03:57 <irl> there's just so much information to load into my brain all at once
16:04:04 <karsten> that's what happens when the proposal process becomes very tiring.
16:04:20 <irl> i should get tor to buy me a bigger brain
16:04:28 <acute> hello
16:04:39 <irl> hi acute!
16:04:41 <karsten> hi acute!
16:04:50 <acute> sorry I'm a wee bit late, coming from another appointment
16:04:53 <karsten> bigger brains sounds great. sign me up, too.
16:05:03 <karsten> I'll have a question for you later, acute.
16:05:09 <irl> heh, that is probably all i can say on the report for now.
16:05:16 <karsten> okay, cool!
16:05:21 <karsten> - TorDNSEL spec #29624 (irl)
16:05:27 <karsten> I assume this answers that, too.
16:05:32 <karsten> sponsor 13 first.
16:05:49 <irl> i have not looked at this since we last discussed it
16:05:54 <karsten> but did you see my question about keeping tordnsel alive?
16:05:59 <karsten> makes sense.
16:06:14 <irl> i did see that, i think it is harder than just a few quick changes
16:06:31 <irl> i think that it is likely that we would have to read changelogs and migration guides
16:06:32 <karsten> I just wonder how we'll get something running in two weeks.
16:07:09 <irl> we just need to get the data into collector, so we can have a complete hack of a solution to do that temporarily
16:07:18 <irl> we can keep an eye on it and restart it if it falls over etc.
16:07:41 <karsten> hmm?
16:08:00 <karsten> I'm lost.
16:08:15 <karsten> by the end of this month, the current tordnsel will be killed, right?
16:08:18 <karsten> at least that's the plan.
16:08:34 <irl> oh wait, yes, i guess that also includes check and the dnsel server
16:08:37 <irl> not just the scanner
16:08:49 <karsten> yes, I think that's all part of tordnsel.
16:08:59 <karsten> it's all very confusing, but I think it does many things.
16:09:09 <karsten> and we may have a replacement for part of it, which is check.
16:09:12 <irl> i don't remember any haskell at all
16:09:31 <irl> i only had two lectures on it as part of my undergraduate degree
16:09:52 <irl> and that was on computation theory more than applied programming
16:09:57 <irl> how is your haskell?
16:10:06 <karsten> it doesn't exist, but that doesn't stop me yet.
16:10:21 <irl> if you want to have a go, i think it is probably worth trying
16:10:32 <karsten> I was mainly wondering if you have any facts that should stop me from even trying.
16:10:43 <irl> arlo did say that it would be a lot of work
16:10:50 <karsten> ah. hmmm.
16:10:57 <karsten> I should ask arlo in that case.
16:11:16 <karsten> okay, moving on. I'll ask him.
16:11:21 <irl> ok
16:11:44 <karsten> - OnionPerf worst-case performance (karsten)
16:12:03 <karsten> so, I resumed this work after all.
16:12:17 <karsten> we have 4 questions to answer.
16:12:39 <karsten> worst case latency, worst case bandwidth, failures/timeouts, total exit capacity.
16:12:46 <karsten> I made tickets for the first two:
16:12:56 <karsten> #29772 and #29773.
16:13:16 <karsten> if you have a moment, please take a look. can be after the next sponsor 13 draft.
16:13:33 <irl> ok
16:13:43 <karsten> I also made two tickets for the failure cases and really slow cases I found earlier: #29743 and #29744.
16:13:48 <karsten> one question on those:
16:13:55 <karsten> which tor version does op-ab use?
16:14:03 <irl> i am not sure but can find out
16:14:11 <karsten> is it recent?
16:14:15 <karsten> or as old as op-nl et al.?
16:14:22 <irl> yes, it will have been a current version at the time it was set up
16:14:30 <karsten> sounds great. can I have the logs?
16:14:37 <irl> yes, i can get those tar'd up
16:14:44 <karsten> that would be great!
16:14:45 <irl> i might even be able to just directly expose them on the webserver
16:14:48 <karsten> or that.
16:15:16 <karsten> regarding the others, can we update them to recent tor versions?
16:15:26 <karsten> or is there a reason not to do that?
16:15:49 <irl> we can, there's no reason to not do it other than that it takes time
16:15:50 <karsten> also not more urgent than sponsor 13.
16:15:53 <karsten> okay.
16:16:02 <irl> i want to also stop using our own tor and instead use tor from deb.tpo
16:16:15 <karsten> works for me.
16:16:26 <karsten> ideally, the .tpf files would say which version was used.
16:16:36 <karsten> but, that's the future.
16:16:41 <irl> i think we had a ticket for this
16:16:51 <karsten> probably. :)
16:17:02 <karsten> acute: I have a question to you:
16:17:11 <acute> yes
16:17:26 <karsten> we were asked whether we can say more about failures and timeouts.
16:17:40 <karsten> I think so far our model for distinguishing the two is imperfect.
16:18:01 <karsten> I found a couple cases so far where things went wrong.
16:18:27 <karsten> I was wondering if you'd be able to help me with classifying these error cases and somehow including an error code of some kind in .tpf files?
16:18:45 <karsten> assuming you still have time for such a thing.
16:18:50 <acute> ok, sure
16:18:57 <karsten> I have graphs showing these errors, and I have stream IDs.
16:19:06 <karsten> but I don't really know the onionperf code that well.
16:19:24 <acute> so in the json output the errors are differentiated well
16:19:30 <karsten> ideally, we'd be able to determine from tor controller logs which error case we just observed and include that in the .tpf output.
16:19:39 <karsten> ah, I didn't look at the json output.
16:19:55 <acute> also, side comment about 29744
16:20:00 <acute> #29744
16:20:05 <karsten> can I send you what I found, and we try to match these errors with what's in the json output?
16:20:10 <karsten> yes?
16:20:37 <acute> so the timeouts and stallouts can be easily controlled from the tgen download graph
16:20:55 <acute> yep happy to do the matching
16:21:15 <karsten> controlled how?
16:21:22 <karsten> okay, cool, will send you what I have.
16:21:23 <acute> for 5m files, the stallout and timeout are set to 3600 seconds
16:21:34 <karsten> you mean to avoid those cases?
16:21:45 <karsten> I think ideally they shouldn't happen in tor.
16:22:00 <karsten> if we avoid them in onionperf, they'll still exist in other applications.
16:22:29 <acute> got you - the bug is about finding why the transfers time out
16:22:57 <karsten> they just stall. nothing times out, except for onionperf after those 3600 seconds.
16:23:12 <karsten> but it seems not very helpful to have a download stall for 3500 seconds and then rush to completion.
16:23:26 <karsten> maybe there's an easy explanation. in tor.
16:23:46 <acute> ok
16:24:01 <karsten> great! feel free to comment on #29744, too!
16:25:23 <karsten> alright, moving on!
16:25:24 <karsten> - NLnet proposal (gaba)
16:25:28 <acute> cool, will do - and if you give me a bunch of torperf files I can match them up to OP error codes
16:25:53 <karsten> great, thank you!
16:26:06 <karsten> so, we don't have a gaba.
16:26:13 <irl> on the first question: everything we do enhances privacy and trust
16:26:18 <karsten> :)
16:26:23 <irl> on the second question: if we can't do it, then it won't happen
16:26:50 <karsten> how about we meet with gaba early next week?
16:27:08 <karsten> that would be after the next sponsor 13 draft and still soon enough to do something for the proposal.
16:27:34 <karsten> I'm a bit worried that we'll run out of march sooner than we'd like to.
16:27:51 <irl> monday i can do any time, tuesday i am busy from 1500
16:28:08 <karsten> let's try monday then. I'll ask gaba.
16:28:12 <karsten> monday afternoon probably.
16:28:19 <karsten> morning might be hard for gaba.
16:28:24 <irl> friday next week i fly to prague for IETF and return wed 27th
16:29:02 <karsten> okay.
16:29:21 <karsten> let's move this to email with gaba. or next monday.
16:29:29 <irl> ok cool
16:29:47 <karsten> 15 seconds left!
16:29:50 <karsten> anything else? :)
16:30:03 <irl> nothing more from me
16:30:10 <karsten> perfect. thanks, irl and acute!
16:30:13 <karsten> bye! :)
16:30:14 <irl> bye!
16:30:21 <karsten> #endmeeting