16:58:43 <ahf> #startmeeting Network team meeting, 15th August 2022
16:58:43 <MeetBot> Meeting started Mon Aug 15 16:58:43 2022 UTC.  The chair is ahf. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:58:43 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic.
16:58:47 <ahf> yoyoyoyoyoy0y0y0y
16:58:50 <dgoulet> o/
16:58:54 <nickm> hihi
16:58:58 <ahf> our pad is at https://pad.riseup.net/p/tor-netteam-2022.1-keep
16:59:18 <mikeperry> o/
16:59:23 <mikeperry> ahf back!
16:59:32 <ahf> let's start by taking a look at our gitlab, which i haven't had much chance to look at yet, but our board is at https://gitlab.torproject.org/groups/tpo/core/-/boards
16:59:49 <ahf> it's some pretty exciting stuff in ~Doing right now!
16:59:52 <ahf> mikeperry: ya!
17:01:05 <ahf> i don't see anything "off" with the board state - i do have some questions, but i can follow up with individuals on those things - nothing urgent and blocking there, just probably to satisfy my curious mind
17:01:09 <ahf> excellent!
17:01:24 <ahf> dgoulet: you've been rollong some releases i have seen! anything in the upcoming pipeline there for c tor?
17:01:47 <dgoulet> yes and no. basically onion service PoW is probably next there
17:01:58 <dgoulet> and I suspect Conflux once mikeperry is out of grantlife hell
17:02:13 <dgoulet> but nothing "short term" as this release had its fair share of fixes
17:02:25 <ahf> sweet
17:02:27 <ahf> very nice
17:02:30 <GeKo> o/
17:02:37 <ahf> o/ GeKo
17:02:38 <GeKo> hi ahf! wb
17:02:56 <ahf> thank you!
17:03:15 <ahf> some stuff have been added to the agenda that i don't know if something i need to dive into in this meeting or if it was something y'all discussed earlier
17:03:22 <ahf> Check-in about arti1.0.0, post scoping, talk about arti#88 to see if its realistic to meet the deadline, or if we can share some of the work.
17:03:24 <ahf> this thing
17:03:40 <jnewsome> o/
17:03:49 <ahf> the overall arti thing is something i just wanted to chat with nick, eta and ian about
17:03:52 <ahf> o/ jnewsome!
17:03:52 <nickm> We talked about it some, I think that another check-in wouldn't be out-of-line.
17:03:57 <eta> yeah, sgtm
17:04:00 <ahf> ya, but that is another meeting, right?
17:04:07 <nickm> okay by me
17:04:24 <ahf> ok, sounds good
17:04:55 <ahf> this becomes a bit of a hybrid meeting right now .. also looking a bit at some of the things we usually do for first-monday-of-the-month
17:05:18 <ahf> dgoulet: we have a few items assigned to triage-bot, but we can look at that at some point. nothing urgent there
17:05:28 <dgoulet> ok
17:05:53 <ahf> i do see some stuff on ACT, but it looks like it's all related to pt/bridges in arti - exciting!
17:06:08 <ahf> nothing from any other teams
17:06:10 <nickm> ("ACT"?)
17:06:34 <ahf> anti-censorship team
17:06:39 <nickm> ah
17:06:43 <nickm> ty
17:06:45 <ahf> we have two announcements:
17:06:46 <ahf> [2022 08 14] [ahf] ahf will return, and micah will be afk
17:06:46 <ahf> [2022 08 15] [nickm] Reminder: nickm will be afk Aug 18 (Thu) through Aug 22 (Mon).
17:06:50 <ahf> very good
17:06:55 <dgoulet> oh also
17:07:03 <dgoulet> my afk status is quite unclear but very soon
17:07:19 <ahf> no discussion items - had there been no discussion items in the meeting in the last few weeks or have they just happened without an entry there?
17:07:24 <ahf> dgoulet: ya! hoping it all goes well <3
17:07:28 * juga says hi
17:07:32 <ahf> o/ juga
17:07:39 <ahf> mikeperry: you wanna do s61 stuff?
17:07:50 <mikeperry> ok
17:08:40 <mikeperry> so we got 0.4.7.10 out, which has some minor changes to RTT calc, and changes to slow start behavior
17:09:19 <mikeperry> I updated prop#324 with these changes, but still need to update the RTT test vectors and make congestion control test vectors for arti
17:09:49 <mikeperry> that ticket is here: https://gitlab.torproject.org/tpo/core/tor/-/issues/40642
17:10:22 <mikeperry> it is still confidential because we found adversaries doing strange shit while investigating DoS attacks
17:11:13 <mikeperry> once enough of the network has upgraded to 0.4.7.10, we have some param tuning to do to deal with them
17:11:28 <ahf> hm. interesting
17:12:51 <mikeperry> dgoulet and ahf will need to help me with looking at those tuning values impact on test relays
17:13:05 <ahf> yeah
17:13:12 <ahf> i am skimming the ticket right now
17:14:12 <mikeperry> there is a summary at the bottom of it. it does not explain the attack we saw, tho
17:14:54 <mikeperry> the end result is that we should see less memory use at guard relays while people upgrade, and hopefully further improvements as we lower some params
17:15:39 <ahf> sweet. memory usage at the guards came up at both of the relay ops meetings i attended in my vacation
17:16:27 <mikeperry> jnewsome,hiro: I still have link-sharing and comparison to live, and the guard cutoff stuff on the TODO, but I also want the network to stablize a bit anyway still, after said tuning
17:17:02 <jnewsome> mikeperry: ack. lmk if you'll need more runners this week
17:17:19 <GeKo> (hiro is out, so this might need to wait a bit)
17:17:26 <mikeperry> jnewsome: ok. not immediately. still getting my bearings and watching this upgrade happen
17:17:42 <mikeperry> GeKo: ah, yeah ok. that's fine
17:18:53 <mikeperry> juga: are you back in action? how is the bwscanner_cc=2 stuff going? do you need any feedback/review there yet?
17:19:31 <juga> yes about back
17:19:32 <juga> sbws#40130 (upload, with fixed sizes) is "donish", GeKo reviewing.
17:19:38 <jnewsome> at some point we'll need to think about how sad we'll be if and when our access to chameleon cloud goes away, and what to do about it
17:19:53 <GeKo> yeah, that review is on my plate for this week
17:20:02 <mikeperry> jnewsome: uh oh. renew might not be so easy this time?
17:20:38 <mikeperry> GeKo,juga: ok I can do a pass after that, just lmk
17:20:47 <juga> mikeperry: ok, thx
17:20:50 <GeKo> sounds good
17:21:07 <GeKo> i might have questions for you during my review as well
17:21:07 <jnewsome> there's no problem that I know of, but I think in general this has been some opportunistic access to a currently-underutilized resource
17:21:09 <GeKo> we'll see
17:21:45 <mikeperry> jnewsome: so if they get more customers, they may decide to boot us basically?
17:21:46 <jnewsome> maybe it's not worth worrying about until it's actually an issue; just thinking a bit about how disruptive it'd be if we lost that access with shortish notice
17:21:54 <jnewsome> but i guess we always have at least our 1 big runner
17:22:24 <jnewsome> mikeperry: more that we're piggy-backing on georgetown-micah's grant to use it
17:23:03 <mikeperry> and they may use it all before october? what is the timeline we're talking about here?
17:23:22 <jnewsome> i don't know of any such immediate issue
17:24:16 <jnewsome> it's just a bit precarious both in that micah needs to keep reusing it and at some point might want to use it himeself
17:24:24 <jnewsome> *needs to keep renewing
17:24:53 * juga needs to go afk, back online in ~1h
17:24:56 <mikeperry> it is still useful to be able to run concurrent sims quickly, esp when 4 different adversaries show up trying to melt the network at once
17:26:01 <mikeperry> I don't expect conflux to need quite as much tuning as congestion control, but it would be useful to confirm assumptions quickly...
17:26:16 <jnewsome> yup, definitely. maybe it's not worth worrying about yet. i'm just noticing that we're becoming a bit dependent on it, but we'll probably have some heads up before if and when it goes away
17:26:17 <mikeperry> at any rate, we can start begging elsewhere
17:28:10 <mikeperry> dgoulet,ahf,jnewsome: speaking of which, we should see about stress testing the pow stuff at some point, since that adversary is not going away and is still messing with perf stats and overload. we can perhaps sync about that later this week, maybe we an armadev
17:28:28 <ahf> yeah, sounds like a good idea to get roger in too
17:28:35 <dgoulet> yah
17:28:43 <ahf> i have no clue if he is at defcon or what right now - i have seen that it is going on right now, but odn't know when it ends
17:28:43 <dgoulet> I will get onto that code this week
17:30:11 <ahf> is the most recent branch still in your gitlab homedir, dgoulet ?
17:30:20 <ahf> dgoulet/tor.git that is
17:30:21 <dgoulet> it is upstream somewhere
17:30:28 <dgoulet> errr remote*
17:30:38 <dgoulet> I think actually most recent is in arma's
17:30:46 <mikeperry> roger and I did some updates to it. it is on the ticket
17:30:52 <GeKo> mikeperry: i still have tpo/network-health/analysis#38 on my plate. is it smart to get back to that one now given the dos?
17:30:56 <ahf> sweet
17:31:17 <mikeperry> https://gitlab.torproject.org/tpo/core/tor/-/issues/40634#note_2821140 and down
17:32:06 <mikeperry> GeKo: I think after we get enough relays to upgrade to 0.4.7.10, and see where we are
17:33:06 <GeKo> okay
17:33:23 <mikeperry> I am hoping we can knock out one of the 4 said adversaries with 0.4.7.10 and some tuning, which shoud make that analysis a little less noisy
17:33:32 <GeKo> and with "see where we are" you mean where we are in our cc debugging efforts?
17:33:51 <GeKo> because the current graph looks a lot like pre-cc
17:33:56 <GeKo> https://metrics.torproject.org/bandwidth.html?start=2021-10-01&end=2022-08-15
17:34:06 <GeKo> at least the bw history
17:34:31 <ahf> https://metrics.torproject.org/bandwidth.html?start=2021-10-01&end=2022-08-15
17:34:32 <GeKo> (the 1000 gbit/s on the left is from rob's flooding experiment back then)
17:35:33 <mikeperry> we have other datapoints that indicate fuckery is still going on. it is just being mitigated. in a few days, we should be able to mitigate it further
17:35:43 <mikeperry> and get more Guard relays back
17:36:04 <GeKo> crossing fingers
17:36:12 <GeKo> (that's all from my side)
17:37:40 <mikeperry> yeah I think I am done too.. I am thinking about follow-up meetings here tho
17:38:01 <mikeperry> we have one that could be dgoulet,ahf,geko, and another that could be dgoulet,ahf,armadev,jnewsome
17:38:37 <ahf> yeah
17:38:38 <GeKo> sounds good to me
17:38:46 <mikeperry> I suppose we can email for that maybe?
17:38:47 <ahf> i would be good if they can be scheduled >= wednesday this week
17:38:55 <ahf> as in wed, thu, or fri (or later)
17:39:00 <mikeperry> ok
17:39:01 <GeKo> wfm
17:39:12 <GeKo> i might be out most of my day tomorrow anyway
17:39:13 <jnewsome> sg
17:39:21 <ahf> cool
17:39:26 <dgoulet> wfm
17:39:33 <ahf> anything else for today's meeting?
17:39:37 <jnewsome> one small thing from my side
17:39:41 <ahf> go
17:39:43 <jnewsome> https://gitlab.torproject.org/tpo/tpa/team/-/issues/40833
17:39:53 <jnewsome> we now have some generally available gitlab runners that can run shadow
17:40:05 <ahf> sweet
17:40:14 <jnewsome> so, this unblocks me from getting an arti shadow sim test in
17:40:28 <dgoulet> !
17:40:32 <jnewsome> but more generally if anyone has reason to use shadow in CI, it should be feasible now
17:40:52 <ahf> that is pretty damn nice
17:41:09 <ahf> will you work with the arti gang directly when this needs to enter their .gitlab-ci configs directly?
17:41:23 <jnewsome> ahf: yup, i have a PR in progress for that
17:41:34 <ahf> sweet^2
17:41:37 <ahf> very cool
17:42:15 <ahf> anything else for today? i will reschedule the 1:1's that we missed today, but i still have a queue of info i need to pass over first
17:42:39 <ahf> let's call the meeting and return to whatever people needs to do next! \o/
17:42:40 <ahf> thanks all!
17:42:42 <ahf> #endmeeting