17:11:37 <nickm> #startmeeting network team meeting 26 July
17:11:37 <MeetBot> Meeting started Mon Jul 26 17:11:37 2021 UTC.  The chair is nickm. Information about MeetBot at http://wiki.debian.org/MeetBot.
17:11:37 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic.
17:11:40 <nickm> boom
17:11:40 <asn> ok here we go
17:11:42 <nickm> it begins
17:11:50 <nickm> pad at https://pad.riseup.net/p/tor-netteam-2021.1-keep
17:12:22 <nickm> how are we all doing with assignments and in-progress things on the kanban?
17:13:20 <dgoulet> not going through the tickets fast enough but nothing falling over either, so lose-win :)
17:13:31 <nickm> i hear you
17:13:35 <asn> im closing lots of stuff that have been open for months.
17:13:43 <asn> i think vg-lite is also ready for merge
17:13:47 <asn> at least according to the comments
17:14:04 <nickm> asn: mike has acked and so have I, I think. Is the spec 100% right now?
17:14:11 <asn> yep i think so
17:14:23 <asn> there is a spec branch lnked from the MR comments wit hthe right spec
17:15:23 <nickm> ok.  You alright with merging it , or would you like me to do that?
17:15:41 <asn> i can merge it
17:15:49 <nickm> great!
17:16:01 <asn> excellent
17:16:05 <asn> there is that
17:18:04 <nickm> next is release status: I don't think we got any critical tickets for stable,  did we?
17:18:44 <dgoulet> nope
17:19:20 <nickm> ok. new issues from other teams?
17:20:38 <dgoulet> bunch of tickets for anti-cens. team but no idea if any of them is critical :(
17:21:02 <gaba> we have some important tickets but we can wait for ahf
17:21:13 <gaba> it is something that is on his plate and can wait for him
17:21:22 <gaba> some pt sponsor 30 tickets taht are important
17:23:47 <nickm> ok
17:23:54 <nickm> do we have any discussion things for this week?
17:24:03 <asn> not from me
17:25:02 <nickm> ok. on to s61?
17:25:20 <mikeperry> ok
17:25:27 <gaba> ok
17:25:52 <gaba> mikeperry: it would be good to see what is the rest of things that acute is finishing wrt onionperf
17:25:55 <mikeperry> I am still working on fixups of the alg branch. mostly refactoring to make the api clean at this point. almost done tho
17:26:32 <mikeperry> I spoke with jnewsome, dgoulet, and hiro about how to compare shadow sim results with the live onionperf results
17:26:42 <mikeperry> which is related to this graphing thing
17:27:35 <mikeperry> in my view, the remaning piece for acute is adjusting the graphs on metrics-web to exclude the measurements from the guard-based 'a' onionperf instances from before a buildtimeout is learned
17:27:51 <mikeperry> iirc, she had a patch for that but it was not deployed yet?
17:28:26 <mikeperry> I thought this was ticket https://gitlab.torproject.org/tpo/network-health/metrics/website/-/issues/40005, but it does not have said patch
17:28:38 <acute> ah,  I had a patch for adjusting throughput calculation
17:28:47 <mikeperry> oh yes, also that
17:29:02 <hiro> yeah that ticket I had in mind to work on these first two weeks that I am back
17:29:19 <hiro> but if there is a patch ready we shouldn't do things twice
17:29:29 <acute> I can look at writing a patch for this also
17:29:35 <mikeperry> maybe I am confusing the throughput patch with the filtering patch
17:30:14 <acute> but as I mentioned on the throughput ticket, I have no experience deploying metrics-web
17:30:25 <hiro> yes I can take taht over no worries
17:30:28 <mikeperry> what is the throughput ticket again?
17:30:29 <acute> perhaps irl or hiro could do this
17:30:40 <acute> hiro: that sounds great!
17:30:56 <mikeperry> is it https://gitlab.torproject.org/tpo/network-health/metrics/statistics/-/issues/40005 ?
17:31:06 <mikeperry> another 40005 heh
17:31:29 <mikeperry> maybe that is why I got confused
17:32:53 <acute> yes that's the one!
17:33:03 <mikeperry> acute: do you know where the tgen models that onionperf uses live? we likely want to add them to shadow sims so we can pull those exact results from shadow too
17:33:31 <acute> yes, they are part of onionperf
17:33:48 <acute> they live https://gitlab.torproject.org/tpo/network-health/metrics/onionperf/-/blob/develop/onionperf/model.py
17:33:55 <mikeperry> are they easy to extract into a tgen model? I am not sure how that stuff connects together
17:34:26 <acute> yes, onionperf dumps an xml file with the model at runtime
17:34:26 <mikeperry> ah so that file outputs a tgen model description
17:35:17 <mikeperry> can we give onionperf a commandline option just to output these model files? like a dryrun or sth like that?
17:35:43 <acute> yes, it already outputs them as xml in the onionperf-data directory
17:37:20 <mikeperry> the code that manages guards and the filtering is onionoperf python yeah? I wonder if that can actually run under shadow
17:37:27 <mikeperry> unfortunately jnewsome is on vacation this week
17:38:49 <acute> also, the tgen model is always the same, minus the server address/port
17:39:16 <acute> yes, all of it is python
17:39:25 <mikeperry> so anyway, that throughput and the filtering for metrics-web is the short term priority. but at some point we need to figure out the best way to compare shadow results to onionperf on live
17:40:25 <acute> what is the best thing to progress at this point?
17:40:34 <hiro> what would need to be added to onionperf?
17:40:43 <hiro> I can take the metrics-web stuff
17:41:00 <hiro> I think there was a ticket left with an open question https://gitlab.torproject.org/tpo/network-health/metrics/onionperf/-/issues/40024 maybe this
17:41:18 <mikeperry> I don't think onionperf actually needs anything more. it outputs what we need. the trouble is that shadow does not
17:43:00 <hiro> uhm wouldn't it be possible to reuse part of the code from onionperf so that data is plotted the same way?
17:43:15 <hiro> but maybe this is more a long term thing
17:43:35 <mikeperry> yeah, possibly the onionperf code can just run under shadow, but I do not know for sure
17:43:56 <hiro> how soon do we need thsi functionality?
17:43:56 <mikeperry> jnewsome would know, but vacation
17:44:17 <hiro> like we are about to make a roadmap with geko when he will be back
17:44:29 <hiro> so maybe we can find time for this
17:45:02 <mikeperry> yeah, within 2-3 weeks or so. basically as soon as I have a flow control implementation
17:45:13 <acute> so we can definitely use the same parser and also the same plotting code if the output is tgen 1.0.0 logs
17:45:37 <hiro> yes thought so
17:46:10 <mikeperry> the answer to the metrics-web question on https://gitlab.torproject.org/tpo/network-health/metrics/onionperf/-/issues/40024 is again https://gitlab.torproject.org/tpo/network-health/metrics/statistics/-/issues/40005 and https://gitlab.torproject.org/tpo/network-health/metrics/website/-/issues/40005
17:46:15 <mikeperry> I will update that ticket
17:48:20 <mikeperry> done
17:49:15 <mikeperry> does that make sense?
17:49:45 <acute> yes, thank you - I will look more at https://gitlab.torproject.org/tpo/network-health/metrics/website/-/issues/40005
17:49:53 <mikeperry> we're playing a lot of musical chairs with people in and out, of course. that's probably adding to the confusion :/
17:50:43 <hiro> ok thanks mikeperry
17:52:07 <mikeperry> asn also reviewed and assisted some of the onionperf backend guard work with acute, so he may have some knowledge worth capturing there. but acute did the heavy lifting
17:52:45 <hiro> I have to sync with asn about v3 metrics too
17:53:27 <mikeperry> dgoulet: I am hoping to have a really nice and clean congestion control branch and spec updated for you rsn. I am working on making it pretty. little bit of ocd rabbithole over it, but every fixup makes it a bit cleaner, so I keep going :)
17:53:38 <dgoulet> mikeperry: I'm excited!
17:54:45 <mikeperry> I'll let you know as soon as it is squashed and the spec is updated
17:54:51 <mikeperry> I think that is it for s61
17:55:41 <nickm> ok, then unless there's something else, let's call the meeting done?
17:55:47 <nickm> going once.....
17:56:05 <nickm> done!
17:56:07 <nickm> Thanks, everybody!
17:56:10 <gaba> o/
17:56:10 <nickm> #endmeeting