16:59:28 <ahf> #startmeeting Network team meeting, 29th November 2021
16:59:28 <MeetBot> Meeting started Mon Nov 29 16:59:28 2021 UTC.  The chair is ahf. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:59:28 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic.
16:59:31 <ahf> hello hello
16:59:37 <ahf> our pad is at https://pad.riseup.net/p/tor-netteam-2021.1-keep
16:59:40 <mikeperry> o/
16:59:41 <juga> o/
16:59:52 <ahf> please wrap up your notes on the pad *this week* before our december forum status update post :-)
17:00:06 <GeKo> o/
17:00:08 <ahf> you have until thursday late evening to add things there
17:00:23 <eta> o/
17:00:42 <ahf> how are people doing with their respective boards at https://gitlab.torproject.org/groups/tpo/core/-/boards ?
17:01:54 <nickm> oh hi all!
17:02:05 <nickm> ahf: doing okay.
17:02:08 * eta stealth adds notes to the pad
17:02:12 <ahf> hm, think some of the stuff in global Doing probably needs some help. roger have some things there
17:02:24 <ahf> gonna talk with david when he is back on that
17:02:31 <ahf> otherwise also looks OK here
17:03:03 <ahf> david isn't here, so i am not gonna ask about release status, but i don't see anything terrible new having arrived from the side
17:03:32 <ahf> we also look OK on tickets coming from other teams
17:03:52 <ahf> i see no discussion items or announcements
17:04:12 <ahf> we are pretty close to the holiday though, everybody should probably remember that the org starts being very sleep around 16-17-18 of december usually
17:04:29 <ahf> so if there is stuff where the entire team needs to be in on it, then it should probably happen before
17:04:51 <ahf> ian, our next rust person, starts in january and nickm and i talked today about running his entrance pretty much like we did with eta as that has gone so well
17:05:18 <ahf> unless we have anything else, i'll let mikeperry take over for performance sponsor work
17:05:24 <ahf> (that i have forgotten the sponsor number of for some reason)
17:05:31 <mikeperry> ok. my net connection is a little iffy
17:05:34 <mikeperry> s61
17:05:38 <ahf> s61!
17:05:39 <ahf> thx
17:06:01 <mikeperry> So main thing is I added tests for negotiation to the MR
17:06:17 <mikeperry> see https://gitlab.torproject.org/tpo/core/tor/-/merge_requests/495#note_2762873
17:07:01 <mikeperry> I found a memory leak in the ntorv3 server side, and a minor lpgic bug client side, so I pushed fixups and then squashed down
17:07:06 <mikeperry> yay testing
17:07:13 <ahf> ahh, very nice
17:07:49 <mikeperry> while I wait for review from dgoulet and nickm, I likely will start on  unit tests for flow control
17:07:59 <mikeperry> and then congestion control algs themselves
17:08:09 <ahf> mikeperry: i see david is reviewer on this one, but he is out right now as you can see in the team signal group. do we need to reallocate a reviewer on this thing?
17:08:09 <mikeperry> unfortunately our shadow gitlab disk issues continue
17:08:17 <ahf> maybe nickm is already on this as well?
17:08:33 <mikeperry> ahf: well nickm could start early if he is so inclined
17:08:49 <mikeperry> but dgoulet will need to read the code anyway for the hidden service piece
17:08:53 <nickm> in that case assign me as reviewer  with a note to give it back to dgoulet once done?
17:08:58 <mikeperry> ok
17:09:13 <ahf> mikeperry: ya, i think you need two sets of eyes on this anyway, but it could make sense to have some early feedback there
17:09:27 <mikeperry> yeah
17:09:30 <ahf> the vanguard reviewer order is changed
17:09:58 <ahf> nickm: ok, sounds good. mike can do that then
17:10:38 <mikeperry> done
17:11:20 <mikeperry> so yeah, it was a short week last week, so that MR and the next batch of shadow sims was all I did
17:11:44 <mikeperry> I suppose I will have to follow up with jnewsome and anarcat to figure out this disk issue
17:12:05 <ahf> cool
17:12:29 <jnewsome> yup. I think anarcat is adding disk, but I should probably take another look to see if there's any low hanging fruit to reduce our usage
17:13:48 <mikeperry> ok
17:14:06 <mikeperry> geko,juga: how was last week for sbws and s61 related network health?
17:14:35 <GeKo> i looked at gabelmoo's issue
17:14:40 <mikeperry> I see that the network continued to run over the holiday, so always a good sign ;)
17:14:48 <GeKo> heh
17:14:49 <GeKo> :)
17:15:00 <juga> from my side no much things
17:15:12 <GeKo> i *think* we might have found the problem n that gabelmoo's web server did not enable ipv6
17:15:18 <GeKo> and some relays did not like that
17:15:24 <GeKo> that shuold be fixed now
17:15:36 <GeKo> and i'll look at logs later this week to double-check
17:15:46 <GeKo> so, fingers crossed :)
17:16:03 <mikeperry> oh I replied on the sbws graphing ticket last week
17:16:10 <GeKo> there was not much done on other fronts
17:16:25 <GeKo> i think i need a dgoulet to make progress on the dns issue
17:16:25 <juga> mikeperry: i saw, thanks, there're 2 sbws and 1 torflow in the usa
17:16:35 <juga> i might overlap the graphs of those
17:16:52 <GeKo> and then there was the usual ongoing net-health work in s61 land
17:16:57 <mikeperry> ok yeah that may be good
17:17:50 <mikeperry> juga: such CDF graphs will also be useful later with congestion control, too. so might be good to keep them as a network-health tool
17:18:03 <mikeperry> we can derive load balancing metrics from them
17:18:12 <juga> mikeperry: ok, it's already a tool
17:18:19 <juga> though not very fast
17:18:31 <juga> (the tool)
17:19:00 <mikeperry> I think geko already had a ticket about load balancing investigation too?
17:19:45 <mikeperry> I remember pointing geko at https://gitlab.torproject.org/legacy/trac/-/wikis/org/roadmaps/CoreTor/PerformanceMetrics#balancing-metrics for that ticket, but I forget the ticket
17:19:53 <GeKo> hrm, aha
17:20:35 <mikeperry> and for reference, hiro will be working on https://gitlab.torproject.org/tpo/network-health/metrics/analysis/-/issues/40008, for sim calibration
17:20:40 <mikeperry> which has some overlap
17:21:18 <juga> ic, the sbws graphs are still missing the separation into guards, exits, ...
17:21:20 <mikeperry> but that ticket won't use svws
17:21:26 <mikeperry> err sbws
17:21:26 <GeKo> https://gitlab.torproject.org/tpo/network-health/team/-/issues/120
17:21:33 <GeKo> i think it was that one
17:21:54 <mikeperry> aha yes thats it. roger added my comment
17:22:23 <mikeperry> ok I turned notifications on for that one. hopefully that means I see updates on it now
17:22:26 <GeKo> i actually linked the graphing ticket, great
17:22:39 <juga> what is being done at tpo/network-health/metrics/analysis#33077
17:22:54 <juga> the graphs you commented, mikeperry
17:23:47 <mikeperry> so we can graph three things in CDFs like this: utilization, normalized sbws multiplier, and raw sbws measured values
17:24:16 <mikeperry> utilization doesn't much involve sbws, but that is what hiro is doing in https://gitlab.torproject.org/tpo/network-health/metrics/analysis/-/issues/40008
17:25:03 <mikeperry> normalized stream capacity is the CDF-Relay-Stream-Capacity metric
17:25:54 <mikeperry> and I think that is also what you are doing in https://gitlab.torproject.org/tpo/network-health/metrics/analysis/-/issues/33077 right?
17:26:04 <mikeperry> those CDFs look normalized
17:26:19 <juga> mikeperry: nope, i i'm doing the CDF-Relay-Stream-Capacity
17:26:24 <mikeperry> yeah
17:26:34 <juga> ah, yes
17:26:40 <juga> sorry, missread
17:27:00 <mikeperry> ok great, yeah so that tool can be used for https://gitlab.torproject.org/tpo/network-health/team/-/issues/120 too
17:27:14 <GeKo> yeah, that's what i thought at least
17:27:59 <juga> do we need graphs for sbws raw measurements?
17:28:14 <mikeperry> I think normalized is better
17:28:20 <juga> ah, ok
17:29:57 <mikeperry> so with negotiation MR, I added a hidden torrc to always negotiate CC, regardless of consensus. that should help with https://gitlab.torproject.org/tpo/network-health/sbws/-/issues/40022
17:30:11 <GeKo> oh, right
17:30:20 <GeKo> so when do you think we need that?
17:30:29 <GeKo> or would ideally have this available?
17:30:58 <mikeperry> I want to get through some more sims to pick reasonable defaults based on sim results, and merge that in the same release as negotiation
17:31:24 <mikeperry> our two blockers there are gitlab disk, and dgoulet availability for the onion service piece
17:31:38 <mikeperry> I may also be able to do the onion service piece picking up from where he left off
17:31:43 <mikeperry> but hopefully he will be back soon
17:32:27 <mikeperry> at any rate, I still think it is possible before Dec 16th or so
17:33:01 <juga> oh, quite soon
17:33:01 <ahf> cool
17:33:12 <GeKo> so, q1/2022 for that ticket?
17:33:16 <mikeperry> at least for a merged branch
17:33:18 <mikeperry> yeah
17:33:33 <GeKo> k
17:34:57 <mikeperry> ok any other comments, questions, concerns?
17:34:57 <ahf> ok, nothing else?
17:35:02 * ahf has none
17:35:10 <GeKo> i am fine, thanks
17:35:14 <juga> me too
17:35:33 <ahf> awesome. thanks folks. remember to update pad before thursday evening utc!
17:35:37 <ahf> #endmeeting