16:59:23 <ahf> #startmeeting Network team meeting, 28th March 2022
16:59:23 <MeetBot> Meeting started Mon Mar 28 16:59:23 2022 UTC.  The chair is ahf. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:59:23 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic.
16:59:25 <ahf> yo yo
16:59:31 <mikeperry> o/
16:59:37 <ahf> first day of me being very confused about summer time since it changed in the weekend
16:59:42 <eta> o/
16:59:48 <dgoulet> yello
16:59:48 <GeKo> hi!
16:59:53 <ahf> pad is at https://pad.riseup.net/p/tor-netteam-2022.1-keep
17:00:06 <Diziet> o/
17:00:10 <ahf> o/
17:00:14 <ahf> giving folks a few to join
17:00:52 <nickm> hihi
17:01:42 <ahf> how are you all doing with your boards: https://gitlab.torproject.org/groups/tpo/core/-/boards ?
17:01:54 <juga> o/
17:02:23 <nickm> no trouble; kinda blocked on a few things, but not blocked on a few others
17:03:31 <ahf> does that mean that you are blocked on less-than-a-few-others or ? :o
17:03:47 <ahf> like we need to coordinate some unblocking or is that already happening? XD
17:04:31 <nickm> it's happening; I just have to do a lot of context switching
17:05:20 <ahf> oki, let me know if i can help there with some coordination
17:05:24 <ahf> ok i don't spot anything off here
17:05:40 <Diziet> nickm: Sorry about my contribution (delay on 433)
17:05:43 <ahf> dgoulet: you have Good News Everyone on release stuff 8)
17:06:12 <dgoulet> 0.4.7.5-alpha was released on friday! so far so good
17:06:20 <ahf> very nice
17:06:20 <dgoulet> and we hope it is the last alpha before the road to stable!
17:06:27 <nickm> Diziet: s'ok, it'll be fine
17:06:33 <dgoulet> that is about it... as for ETAs on that stable, unclear but I suspect "faster than usual"
17:06:46 <dgoulet> pretty much it
17:07:41 <ahf> very good
17:08:15 <ahf> i don't see anything incoming from other teams. we have the DoS discussion that is still ongoing on the private ticket, which is good, but otherwise nothing seems like it needs our attention here and now
17:08:49 <ahf> no announcements or anything. this is the last meeting in march, so next week we have the s61 standalone meeting
17:08:54 <ahf> mikeperry: you wanna talk s61 now?
17:09:02 <mikeperry> ok yeah
17:09:45 <mikeperry> so I updated the sim plan with all the stuff that has been one for the alpha, the stuff to do pre-stable, and the stuff post-stable
17:09:49 <mikeperry> https://gitlab.torproject.org/mikeperry/tor/-/blob/cc_shadow_experiments_v2/SHADOW_EXPERIMENTS.txt#L1183
17:10:25 <mikeperry> the last thing for the stable that I want to check is large downloads, to make sure that throughput remains the same long-term
17:10:37 <mikeperry> hiro is working on that in https://gitlab.torproject.org/jnewsome/sponsor-61-sims/-/issues/15
17:11:40 <mikeperry> I relocated the utilization CDF graphs for post-stable, since the large-download test was more important, and after other sim runs, we only need the utilization stuff to check our calibration with the lvie network
17:11:52 <mikeperry> and to support further tuning post-congestion control
17:12:22 <mikeperry> GeKo: this is partly why I dropped your question about the analysis graphs; I am thinking we can discuss that next week
17:13:10 <mikeperry> it is looking to be fairly involved to parse and extract all that data, and the large-download test is much simpler and more immediately useful
17:13:25 <GeKo> okay, sounds good
17:13:49 <mikeperry> juga: I did some commenting on the sbws congestion control MR and ticket, as you saw. it is looking good so far
17:14:20 <juga> mikeperry: ok, i didn't work much on that last week
17:14:40 <juga> (just cause other stuff)
17:14:48 <mikeperry> yeah np
17:15:24 <mikeperry> I think we're in good shape tho
17:15:35 <juga> ok
17:15:51 <mikeperry> this week I hope to finally write those unit tests for flow control and the CC alg test vectors, heh
17:16:23 <mikeperry> there is a small risk that the large-bandwidth result might require some tweaking, but hopefully not
17:17:07 <mikeperry> does anyone have any questions? I don't have much else
17:17:18 * juga is good
17:17:30 <GeKo> just a quick note
17:17:33 <mikeperry> the meeting will be Apr 4th, 15utc
17:17:53 <GeKo> i've been crunching overload data we accumulated over the months, see: https://gitlab.torproject.org/tpo/network-health/analysis/-/issues/24
17:17:57 <GeKo> and https://gitlab.torproject.org/tpo/network-health/analysis/-/issues/27
17:18:08 <GeKo> i think things are not looking unresaonable
17:18:30 <GeKo> in particular with the ntor onionskin fix being the likely culprit for the things we saw in analysis#27
17:18:36 <mikeperry> oh are we gonna backport the ntor onionskin fix to 0.4.6?
17:18:57 <GeKo> i do think given all of that we should use the 18h window for nethealth purposes
17:19:02 <GeKo> and on relay-search
17:19:12 <GeKo> it gives us less noise and better data overall
17:19:23 <GeKo> and is thus less confusing to relay operators
17:19:41 <GeKo> if that sounds good then i can file tickets and we can get that going
17:19:52 <mikeperry> this might be a ticket for hiro for the relaysearch portal? hiro did the first version of that?
17:20:00 <GeKo> yes
17:20:12 <GeKo> and i need to adapt the overload monitoring tools
17:20:34 <GeKo> other than that i am good
17:21:04 <GeKo> dgoulet: for the backport above?
17:21:33 <dgoulet> I'm not entirely sure... 046 is EOL "soon" so unclear if this would worth the work. Maybe if this just comes up all the time on our 046 relays?
17:22:17 <GeKo> we can cut down the noise with the 18h overload window in our netork health tools
17:22:41 <GeKo> so, not too urgent from that side
17:23:36 <dgoulet> ack
17:23:43 <ahf> maybe flag it for backporting and then we can evaluate if we need to roll a release anyway?
17:23:48 <mikeperry> dgoulet: fwiw, even if the large-bandwidth test shows a bad result, that would be a change that won't impact anything until cc_alg=2 is set
17:23:48 <ahf> the change wasn't huge iirc
17:24:07 <GeKo> ahf: +1
17:24:12 <mikeperry> so in theory we can already start planning rc and stable fairly quickly, just to get it out and start the switch from 046
17:24:48 <dgoulet> mikeperry: yeah that is basically what I was hoping for
17:26:06 <ahf> ah, you want to roll people over to 0.4.7 here as part of this
17:26:23 <dgoulet> yeah...
17:26:44 <mikeperry> well, perhaps instead of bothering with backports, yeah. and to open the merge window for 048 stuff
17:27:46 <nickm> FWIW, if we do not change our support policy, the EOL date for 046 will need to change:
17:28:05 <ahf> yes, that is gonna be a bit grim (048 merge window)
17:28:05 <nickm> our policy has been that we support releases for 9 months, or until 3 months after the next release is stable, whichever is later
17:28:13 <mikeperry> ah
17:28:16 <dgoulet> yeah 3 months after stable is what I had in mind
17:28:19 <nickm> great
17:28:35 <mikeperry> that would argue for a backport of the ntor overload thing maybe? idk
17:29:03 <dgoulet> plausible
17:29:43 <mikeperry> we should have good overload stats going at the point of cc_alg=2
17:29:53 <mikeperry> to make sure that it's not making things worse
17:30:17 <mikeperry> so if 046 is still in the picture, we should backport the overload thing before cc_alg=2
17:30:54 <dgoulet> ok
17:31:02 <mikeperry> anyway we have a bit of time to think about that and let it bake in 047
17:32:18 <mikeperry> I think that's all for s61 then?
17:32:33 <ahf> nice
17:32:41 <GeKo> <-- is still good
17:32:51 <ahf> anything else we need to talk about whilst we are all here?
17:33:08 <ahf> if i am off by a bit in meetings this week, please prod me. it takes me a few days to switch over
17:34:17 <Diziet> My timezone tip again: if using muggle-style calendaring tools rather than Unix-TZ-style ones, set the "location" of your meetings to Reykjavik.
17:34:53 <Diziet> Iceland use UTC all year round.  That avoids the thing where the muggle tools say "UTC" but mean "London time"
17:35:06 <ahf> i have my google calendar copies of our meetings set to UTC so my phone is mostly right, but sometimes i don't listen to my phone's *beep* 10 min before a meeting
17:35:22 <ahf> but yes, good point with iceland
17:35:38 <ahf> ok folks, i am gonna call it for now. see you all at all the other meetings during the week and on thurday for our sync
17:35:39 <ahf> o/
17:35:40 <ahf> #endmeeting