16:59:37 <ahf> #startmeeting Network team meeting, 7th September 2021
16:59:37 <MeetBot> Meeting started Tue Sep  7 16:59:37 2021 UTC.  The chair is ahf. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:59:37 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic.
16:59:46 <ahf> hello hello, welcome, our pad is at: https://pad.riseup.net/p/tor-netteam-2021.1-keep
16:59:51 <dgoulet> greetings
17:00:15 <ahf> let's give folks a little bit to join in
17:00:21 <jnewsome> o/
17:00:46 <nickm> hihi
17:01:06 <ahf> let's get going
17:01:12 <mikeperry> o/
17:01:36 <GeKo> o/
17:01:48 <ahf> i don't see anything new to our release status lists
17:02:05 <ahf> but i think dgoulet and i need to put our heads together and talk a bit about what we learned from the last release series?
17:02:34 <dgoulet> we did that part
17:02:40 <dgoulet> I think we need to do part 2 :)
17:02:43 <dgoulet> activate()
17:02:44 <ahf> ya
17:02:57 <ahf> i think the things we want to talk a bit about was changelog being the big one, no?
17:03:27 <ahf> can we take that during this week maybe ? i want to prioritize sponsor stuff first this week before i dive into anything else
17:03:45 <dgoulet> sure np
17:04:20 <ahf> ok, i don't see this meeting as a "first meeting of the month" so skipping a few items
17:04:32 <ahf> otherwise i'm gonna be so confused next week with the s61 sync earlier
17:04:43 <ahf> ugh, i skipped an item!
17:04:49 <ahf> how is everybody's boards looking?
17:05:23 <nickm> not sure; I'm still getting back to mine :/
17:05:30 <ahf> hehe, is fine on day #1 after vacation XD
17:05:33 <ahf> no rush
17:06:11 <nickm> dgoulet, mikeperry: Is there a ticket for specifically the part of the s61 work I'm doing?
17:07:08 <dgoulet> not to my knowledge but the main flow control branch is under early review and so I would think that is at least the starting point for hooking ntorv3 in there... mikeperry would know
17:07:16 <mikeperry> nickm: there is one for all of the negotiation logic, but the main thing that would be most helpful that you can do there is just hook up the protover and stub the calls to the new ntorv3
17:07:36 <nickm> what is the ticket id?  I'll start a subticket for the part I'm doing.
17:07:51 <nickm> (I would like to have something assigned to me here so I can track it better)
17:07:52 <mikeperry> if dgoulet can handle the equivalent for hs descs and the hs intro extension fields, then I can do the more complicated logic and use of the fields
17:07:55 <mikeperry> https://gitlab.torproject.org/tpo/core/tor/-/issues/40444
17:07:59 <nickm> thanks, mikeperry !
17:08:08 <dgoulet> mikeperry: yes HS stuff is on my side, spec + code
17:09:28 <mikeperry> if either/both of you prefer to break off your pieces into your own tickets, you (or me) can do that
17:09:45 <mikeperry> but I am fine just taking a branch for each of you and working on it from there
17:10:27 <ahf> very good
17:11:14 <ahf> there is no discussion items. on the thursday meeting, we need to talk a bit about team goals and objectives for us to prepare for all hands next week
17:11:25 <ahf> i'll talk a bit about what i have noted down and then we can do some feedback on it
17:11:32 <ahf> otherwise i think we can move to the s61 part of the meeting
17:12:21 <nickm> What is the all-hands next week for?
17:12:45 <ahf> it's the thing where team leads talk about OKR's for their teams
17:13:17 <nickm> ah!
17:13:26 <GeKo> the next round after matt, duncan, and cecylia started
17:13:33 <ahf> ye
17:13:40 <ahf> mikeperry: wanna run the s61 part?
17:13:50 <mikeperry> ok
17:14:44 <mikeperry> I have not made a ton of progresss; still distracted, but much less so. I saw dgoulet's prelim review of flow control. thanks! I will make fixups this week
17:14:52 <dgoulet> \o/
17:16:09 <mikeperry> dgoulet: the other checklist items on https://gitlab.torproject.org/tpo/core/tor/-/issues/40450 for you are slightly higher prio than the negotion stuff. do you have a preference for how we coordinate on those things?
17:16:30 <mikeperry> I can do fixups and squash, and you can take it for a bit, or discuss them first, or ?
17:16:53 <dgoulet> mikeperry: hmmm my prio list here was 1) Flow control review, 2) Ratelimit edge conn, 3) HS Specs for congestion control
17:16:56 <dgoulet> mikeperry: should I reconsider?
17:17:17 <mikeperry> yeah that sounds about right, though I would like to spotcheck with your knowledge of the oomkiller abnd kist queues too
17:17:26 <mikeperry> to decide if we want to tweak or parameterize those
17:17:28 <dgoulet> mikeperry: ok! lets meet up then when you are there?
17:17:38 <mikeperry> shadow sim might hit those, idk
17:17:39 <dgoulet> I left a comment about KIST in the review
17:17:43 <dgoulet> and we can definitely discuss that
17:17:59 <dgoulet> maybe my understanding there is off so voice would be desirable
17:18:23 <mikeperry> ok yeah I have stable internet now so BBB is possible
17:18:51 <dgoulet> anytime
17:19:30 <mikeperry> jnewsome,hiro: how is the shadow sim data and baseline sime size scaling shaping up?
17:20:03 <jnewsome> i'm generating the right tgen and onionperf logs now
17:20:18 <jnewsome> i've enabled guards, but there's still a client scaling factor, so the results may be wonky
17:21:04 <hiro> I have to work on the baseline part and the metrics you requested. hopefully this week will finish it
17:21:11 <jnewsome> i'm working on getting rid of that (by scaling up the # of clients), but currently running into an unknown failure. semi-blocked on storage issues in the CI for debugging e.g. https://gitlab.torproject.org/tpo/tpa/team/-/issues/40375
17:22:54 <jnewsome> speaking of which I should clarify on that issue that it's (kind of) blocking for me now. I can maybe work around by logging less in the meantime but it's a fiddly slow feedback loop
17:23:20 <mikeperry> yeah, bleh
17:24:02 <mikeperry> ahf: do you have enough gitlab knowledge to help improve that, or offer guidance?
17:24:09 <mikeperry> or hiro, idk
17:24:12 <ahf> hm
17:24:43 <mikeperry> idk if space is actually scare or if this is just another conservative limit
17:24:50 <ahf> i don't think i can set settings there
17:24:53 <ahf> it's just the default i think
17:25:18 <jnewsome> yeah I don't get the pushback for increasing the 4 MB console limit, especially since they can do it jsut for the shadow runner
17:26:16 <ahf> yeah, i think that value should just be bumped to something larger than 4 MB's for sure. i don't even know what a webconsole is here, but i assume it's some kind of shell via the runner
17:26:42 <jnewsome> yeah - that'd let me interactively debug while it's still running
17:27:04 <jnewsome> but if the job fails and doesn't save logs i'd still be stuck
17:27:06 <ahf> ah, it's on the runner's settings, then i cannot do anything magical :-/
17:27:11 <ahf> i think i can only touch the gitlab config itself
17:28:15 <ahf> let's see what the sysadmins says, they have also had long weekend and just returned today
17:29:14 <nickm> also see tpa/gitlab#95
17:29:23 <nickm> I think there's some good discussion there about dpace issues
17:29:27 <nickm> space
17:29:30 <nickm> not sure if it's relevant
17:29:51 <ahf> i think that is for artifacts, which can be GC'ed - i don't think log output gets GC'ed in the same way
17:30:41 <nickm> ah ok
17:31:07 <jnewsome> yeah, I could definitely see hesitating to increase this much if there's not some GC in place
17:32:02 <ahf> ya, but just doubling it or 4x should be OK at least i think :-S but don't know how the space situation looks for GL
17:33:16 <jnewsome> anyway we'll see. i'm not 100% stuck; just need to log less to the console and try again
17:33:50 <ahf> cool!
17:34:22 <mikeperry> ok
17:34:34 <mikeperry> GeKo: gaba asked me if we still need to keep https://gitlab.torproject.org/tpo/network-health/metrics/analysis/-/issues/33077 open
17:34:55 <GeKo> hrm
17:34:56 <mikeperry> GeKo: you and juga are comparing results a diff way, right? so those graphs not necessarily needed?
17:35:00 <jnewsome> mikeperry: here's a run with guards enabled, but client scaling still in place, if you care to sanity-check results. https://gitlab.torproject.org/jnewsome/sponsor-61-sims/-/jobs/35340/artifacts/browse/jobs/
17:35:22 <mikeperry> jnewsome: thanks. is this still a 1% sim?
17:35:31 <jnewsome> yup
17:35:44 <GeKo> mikeperry: right
17:35:56 <GeKo> we *could* do something like that
17:36:10 <GeKo> but it is not really necessary i think
17:36:13 <mikeperry> ok
17:36:16 <GeKo> so, closing is fine
17:37:18 <GeKo> i wonder where the code for that is, though
17:37:22 <GeKo> might be handy
17:37:27 <GeKo> at some point
17:37:52 <GeKo> but i can dig for that later
17:38:28 <mikeperry> so I the last item I'm aware of is that juga still needs a decision on what to do about pinned exits vs onions for https://gitlab.torproject.org/tpo/network-health/sbws/-/issues/40022. I think we are now leaning towards onions, since both approaches require hacks, and maybe onions less so
17:39:11 <mikeperry> but other things are currently higher priority than getting that done (which is fine for now)
17:39:15 <nickm> I think the exit approach would require less hacking, but I'll defer to whoever is doing the hacking
17:39:46 <GeKo> who is it? :)
17:40:19 <mikeperry> well if it ends up being onions, probably dgoulet. dgoulet also not happy with .exit, so he is in theory motivated ;)
17:40:34 <mikeperry> but maybe there is a way to do exits without .exit
17:40:41 <dgoulet> onion side np... the .exit notation is very MEH for me
17:40:42 <mikeperry> I have not had a chance to put more brain into it though
17:40:51 <dgoulet> as in with the past of that thing :P
17:40:53 <nickm> Did you try mapaddress + .exit ?
17:41:13 <nickm> That should work.
17:42:27 <mikeperry> ah is that part of arma2's hack already? looks like it may be
17:42:41 <mikeperry> so that .exit can only be used if we're doing an addrmap from the control port?
17:43:03 <nickm> i think .exit is alread supported on addrmap from controlport, no hack needed
17:44:57 <ahf> ok
17:45:05 <ahf> are we at the end? :-)
17:46:19 <dgoulet> maybeeeee
17:46:53 <ahf> one can never know if the meeting is truly at its end 8)
17:47:00 <ahf> does anybody have more we need to chat about?
17:47:19 <mikeperry> sorry I lost internet again
17:47:23 * jnewsome is good
17:47:27 <mikeperry> maybe not so stable after all
17:47:27 <ahf> mikeperry: ah
17:47:44 <mikeperry> anyway I am good I think
17:47:47 <ahf> excellent
17:48:00 <GeKo> i am good, to
17:48:02 <GeKo> o
17:48:14 <ahf> thanks everybody o/ next meeting is on monday, and remember we do the first-of-the-month meetings on next monday, which means a long network team meeting AND we have the s61 bbb sync earlier
17:48:24 <ahf> update your calendars if needed 8)
17:48:25 <ahf> thx o/
17:48:27 <ahf> #endmeeting