16:58:53 #startmeeting Network team meeting, 14th november 2022 16:58:53 Meeting started Mon Nov 14 16:58:53 2022 UTC. The chair is ahf. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:58:53 Useful Commands: #action #agreed #help #info #idea #link #topic. 16:58:56 hello hello welcome 16:59:18 our pad is at https://pad.riseup.net/p/tor-netteam-2022.1-keep 16:59:21 giving folks a few to join 16:59:37 hi all! 17:00:13 o/ 17:00:21 o/ 17:00:39 o/ 17:01:29 hello hello 17:01:53 how are folks doing with their boards: https://gitlab.torproject.org/groups/tpo/core/-/boards ? 17:02:33 GeKo: you, david and i should probably chat about https://gitlab.torproject.org/tpo/core/tor/-/issues/40661 at some point 17:02:37 Still buried in arti bridge/PT. I did look at the outstanding tickets marked with that milestone... 17:02:39 it's in Next but has no assignee right now 17:02:56 Diziet: nice 17:03:02 But last week so I can't remember any conclusions other than plucked some onto my Doing/Next 17:04:07 I'm currently without a specific assigned-to-me arti 1.1 item, so I am: 17:04:10 1) attacking TODO pt-client stuff 17:04:17 2) looking at all the other tickets on the milestone 17:04:27 3) looking at the rest of my backlog for low-hanging fruit for 1.1 17:04:35 ahf: yep, will do once he is back 17:04:42 4) offering to offload anything that eta / diziet are busy with :) 17:05:27 [end of status] 17:05:50 GeKo: sounds good 👍 17:06:02 nickm: nice, ya, i'd say that is part of the grant arti#69 plan 17:06:49 ok! 17:06:55 very good, david is out, so let's skip the release stuff 17:07:09 Diziet: when it wouldn't disrupt you, please have another look at arti!847; I tried to answer your questions there 17:07:11 david got the 0.4.7.11 out last week, so there is room to celebrate there 17:07:17 (woo!) 17:07:38 nickm: Yes, I saw, thanks... 17:07:51 I thought you would like to get the plumbing ASAP so we can break it all together, so I have been doing that. 17:07:58 Should have a thing RSN 17:08:26 yeah, makes sense 17:08:54 i see nothing from other teams 17:08:59 Do we have a test case for running with PT or bridge ? 17:09:23 arti gang, so, nickm caught me in our 1:1 before i was about to make a mistake for the announcements today. we have NO meeting tomorrow on IRC at 14:00, BUT, we do have our BBB sync tomorrow at 17 UTC 17:09:24 nickm: If you're feeling at a loose end you might want to write a test that fails... 17:09:38 so please remember that. i added it to the network team-specific calendar in nextcloud too 17:10:04 hmm. There is a chutney case for that, but our pure-arti integration tests are a bit absent. 17:10:10 Wait, what, BBB at 1700 ? 17:10:19 tomorrow 17:10:30 the arti public sync meeting that we had a thread about the other week 17:10:32 I thought we were doing the BBB *and* the IRC at the same time, just alternating weekx 17:10:44 we have no IRC meeting tomorrow at all 17:10:55 OK, fine. 17:10:56 the IRC meetings have just been a weird arti 1.1.0 thing 17:11:00 1700 then. 17:11:03 excellent 17:11:13 Diziet: hm, trinity-1686a's work on arti!15 will be helpful here 17:11:25 ahf: can you please also remind eta when she's online? 17:11:32 err, chutney!15 17:11:36 nickm: ye 17:11:40 ty! 17:11:57 hello 17:12:04 apologies, my train was in a tunnel 17:12:04 oh she is here - now i just spammed her in pm 17:12:15 (hi eta!) 17:12:38 hey nickm 17:12:59 i see no announcements or anything like that, so unless people have more they need to sync on, i'd pass the mic to mikeperry 17:13:06 ack re the meeting 17:13:13 o/ 17:13:18 My phone calendar app is maulfunctioning 17:14:04 so during the 0.4.7.11 release process, I tuned the congestion control params some more. I have a set of param updates to deploy: https://gitlab.torproject.org/tpo/network-health/analysis/-/issues/44 17:14:19 nickm: re chutney!15, yay 17:14:30 some of them will benefit fro the metricsport monitorihng in 0.4.7.11, some of them can be done now-ish 17:14:48 hiro: the second set of params will benefit from https://gitlab.torproject.org/tpo/network-health/analysis/-/issues/43 17:16:47 I also ran a sim looking at CircuitPriorityHalflifeMsec=1.. strangely, it did not seem to change circewma behavior, which makes me wonder if circewma might be causing some problems with congestion control throughput, or may have other bugs... 17:16:57 I am not going to rabbithole on that, but it is noteworthy 17:17:36 juga began discussing scoping obfs4 load balancing with meskio: https://gitlab.torproject.org/tpo/network-health/onbasca/-/issues/130 17:18:09 juga: it looks like meskio answered most of your questions there? lmk if you have additional questions for me 17:18:51 mikeperry: i just never tried to create a circuit with a bridge, so thats what i still wanted to look at 17:19:33 mikeperry: did you create a ticket for a future look into the CircuitPriorityHalflifeMsec unexpected behavior or? 17:20:37 ahf: not yet.. geko is also investigating guard+fast bottlenecks here: https://gitlab.torproject.org/tpo/network-health/analysis/-/issues/41 17:21:16 basically, the trio of kist, circewma, and guard+fast bottlenecks are our main suspicions for why congestion control is not consuming more bandwidth on the network 17:21:48 so the priority is finding out which, more so than changing every last quirk of circewma 17:21:49 mikeperry: sounds good! 17:22:25 but it is noteworthy that tuning the circewma param had zero effect 17:22:38 that could mean it is broken, or it could mean that for congestion control it is mostly irrelevant 17:23:12 ok mikelerry I will finish that this week 17:23:20 *mikeperry 17:23:33 hiro: ok cool, thanks 17:24:04 geko: lmk if you want to discuss those param updates. I think we can do some of them before dgoulet gets back and before hiro does that work, if you have time 17:24:51 I will also be out next week. taking the three days before US thanksgiving off 17:25:25 mikeperry: i can ping you tomorrow? 17:25:29 sure 17:25:36 i am almost out for today 17:25:39 okay 17:25:42 np 17:25:45 nice 17:26:40 I think that is it unless people have other questions 17:26:56 * ahf is good 17:27:03 * GeKo too 17:27:14 * juga too 17:28:01 let's call it then! 17:28:02 thanks all 17:28:04 #endmeeting