17:00:31 <ahf> #startmeeting Network team meeting, 22 February 2022
17:00:31 <MeetBot> Meeting started Tue Feb 22 17:00:31 2022 UTC.  The chair is ahf. Information about MeetBot at http://wiki.debian.org/MeetBot.
17:00:31 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic.
17:00:37 <Diziet> We could have just had the two meetings concurrently in one channel.  Wouldn't have been confusing at all.
17:01:00 <ahf> hello hello
17:01:02 <ahf> Diziet: lol, yes
17:01:09 <nickm> hihi
17:01:18 <ahf> pad is at https://pad.riseup.net/p/tor-netteam-2022.1-keep
17:01:34 <jnewsome> o/
17:02:06 <ahf> how are people doing with their boards: https://gitlab.torproject.org/groups/tpo/core/-/boards ?
17:02:44 <dgoulet> yes
17:02:53 <nickm> think so.
17:03:18 <ahf> i don't see anything off either
17:03:32 <ahf> dgoulet: anyting on releases?
17:03:50 <dgoulet> nope
17:04:20 <ahf> don't see anything incoming from other teams
17:04:39 <ahf> i have a question for the arti gang
17:04:57 <ahf> for phase 1.1.0 we have the bridge support goals and some PT support (maybe)
17:05:27 <ahf> would it make sense for us to invite the anti-censorship team to one of our public arti meetings to sync with them on what arti means to them too?
17:06:08 <ahf> we have a lot of tickets where we want to do things with PT's that we can't easily do today, but maybe with some thinking it would be easier to do in arti
17:07:12 <nickm> makes sense; right now we have some discussion with meskio on one of the tickets, but there's no reason we can't expand it
17:07:27 <nickm> doesn't seem like we're in a huge hurry imo since we don't start that milestone for a number of months...
17:07:31 <ahf> sounds good. people OK if i just invite them and mention a bit what the topic could be there?
17:07:33 <ahf> ya, agreed
17:07:34 <nickm> ... so we could also wait till cecylia is back
17:07:36 <ahf> it's more to get them thinking
17:07:45 <ahf> hm, that might be a good idea
17:08:03 <ahf> i'll add it on my todo for picking up in ~2 weeks then
17:08:34 <ahf> very good
17:08:52 <ahf> mikeperry: you can do s61 now
17:09:10 <mikeperry> kk
17:10:01 <mikeperry> last week I finished up the last fixups for the negotiation MR (https://gitlab.torproject.org/tpo/core/tor/-/merge_requests/525/). a couple threads need to be resolved
17:10:23 <mikeperry> but otherwise I think it is ready. I ran some sims with it, and it still works as before
17:11:06 <mikeperry> I believe I have also found the reason why the negotiation branch had more relay cell queue logs than pre-negotiation. it was basically a typo when migrating log patches. running a sim to confirm that.
17:11:30 <ahf> nice
17:11:37 <mikeperry> I think it's pretty much ready for merge
17:12:05 <mikeperry> jnewsome has been working on making shadow support onion services in Shadow
17:12:46 <mikeperry> there are some timeout issues that are visible from 0.4.6, which make it a bit difficult for tgen to let enough xfers complete, but these are not major blockers afaict
17:13:15 <ahf> hm, interesting. wonder if that is something we should look into outside of s61 too?
17:13:19 <mikeperry> jnewsome: did you want to talk about that, and/or hiro's utilization CDF branch? (hiro is out)
17:13:46 <jnewsome> yeah, i filed an issue in the core tor component
17:13:55 <mikeperry> yeah, it is basically a UX issue. it might be possible to fix fairly easily, but might also require some digging
17:14:01 <jnewsome> pulling it up; i'll add to the pad
17:14:49 <mikeperry> the tl;dr is that 20% of circuits decide to start using the old-school 60 second timeouts in some cases
17:14:57 <jnewsome> but yeah have been chasing down tgen transfer error rates in onion services under shadow, which turns out to be due to timeouts while building the circuits, which turns out to be tor having long timeouts for individual extensions (i'm fuzzier on that last part)
17:15:15 <ahf> hm, funky
17:15:17 <mikeperry> possibly because of retry and additional-hop cases
17:15:39 <jnewsome> I also stumbled across a tgen prng but last week that seems like it could affect results, but in small test sims doesn't seem to amke much difference
17:15:52 <ahf> it was my impression we generally get occasional complaints about "odd timeout issues" with v3 onions, but it was always my impression they happen in specific periods (like under network attack and what not)
17:16:21 <jnewsome> https://gitlab.torproject.org/tpo/core/tor/-/issues/40570
17:16:30 <mikeperry> yeah, we perhaps did not have enough onion services running, as compared to the number of clients accessing them. that might have exacerbated it
17:17:11 <ahf> thx jnewsome
17:17:17 <mikeperry> jnewsome increased the number of onion services, and is also increasing the number of onion perfclients to get more data points
17:18:22 <ahf> cool
17:18:35 <mikeperry> dgoulet: the tl;dr for you is that we may be ready for an 0.4.7-alpha soon. probably as soon as we get some onion perf data
17:18:44 <jnewsome> yeah. as for the utilization CDF branch I think it's nearly ready. some things are not as i'd prefer for maintainability, but i'd be ok with merging for expedience if we're blocked on it
17:18:47 <dgoulet> I'm ready for this
17:18:52 <jnewsome> do you know when hiro is back?
17:19:00 <ahf> what's next steps for the merge of CC work? :-p a push? :-p
17:19:12 <mikeperry> jnewsome: hiro is back tomorrow
17:19:29 <mikeperry> we dont *need* the utilization CDFs asap. it is OK to get it a bit cleaner if you prefer
17:19:49 <mikeperry> testing onion services for any obvious regressions with the new negotiation+congestion control is a bit more important
17:20:35 <jnewsome> mikeperry: ok. i just merged the change I needed in tornettools. i'll launch a new baseline sim with onion services after this meeting
17:21:04 <mikeperry> ahf: my read is that there is a unit test that nickm requested that he still needs to look over, and then there's some tickets I filed for the other threads. everything just needs to have "resolved" clicked afaict
17:21:29 <ahf> cool!
17:21:32 <ahf> very close
17:22:10 <nickm> mikeperry: are you blocked on that at all?  I wasn't likely to get to it soon but I can reshuffle my priorities if i need to
17:22:36 <ahf> the earlier we can get this in the earlier we can get the alpha out
17:23:05 <mikeperry> nickm: yeah, I think I did the stuff you asked. it would be good to get it wrapped up so we can start looking at an alpha
17:23:11 <dgoulet> yeah we also have an important fix in the pipe for health team for this next alpha
17:24:09 <mikeperry> in the next day or two we should have the sim results from onion services with congestion control to make that call
17:24:11 <nickm> ok, i'll be on it today
17:24:51 <ahf> sweet
17:24:57 <ahf> do we have anything else for today's meeting?
17:25:16 <mikeperry> juga,geko: anything from network-health and sbws?
17:25:23 <GeKo> nope
17:26:04 <mikeperry> ah juga might still be on holiday
17:26:07 <mikeperry> ok
17:26:10 <ahf> bueno
17:26:13 <ahf> let's call it then
17:26:27 <ahf> thanks all for joining. tomorrow there is both big vpn meeting and all hands
17:26:33 <ahf> see you all in #tor-dev or wherever we talk next
17:26:36 <ahf> #endmeeting