17:01:53 <ahf> #startmeeting Network team meeting, 13 december 2021
17:01:53 <MeetBot> Meeting started Mon Dec 13 17:01:53 2021 UTC.  The chair is ahf. Information about MeetBot at http://wiki.debian.org/MeetBot.
17:01:53 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic.
17:01:55 <ahf> yeah, sorry
17:02:01 <dgoulet> o/
17:02:05 <GeKo> hi
17:02:07 <ahf> hello hello
17:02:10 <ahf> https://pad.riseup.net/p/tor-netteam-2021.1-keep is our pad
17:02:11 <nickm> hi everyone
17:02:24 <juga> o/
17:02:31 <mikeperry> o/
17:02:35 <jnewsome> o/
17:02:38 <ahf> i failed at doing the report last week, so just continue adding content to the december part of the pad and i'll get a draft out for the november report today or tomorrow
17:02:39 <ahf> ok!
17:02:44 <ahf> is people's boards looking alright?
17:03:01 <ahf> https://gitlab.torproject.org/groups/tpo/core/-/boards
17:03:08 <dgoulet> good board
17:03:38 <nickm> ok w me
17:03:53 <ahf> heh, i moved some android things that i thought GP wanted into doing last week but they don't seem that interested in it now, so think i'll move those back into ... not-doing
17:04:35 <ahf> ok, i don't see anything there
17:04:52 <ahf> dgoulet: wanna talk a bit about release stuff here?
17:05:18 <dgoulet> sure, not much to say apart that ahf and I need to talk but we'll likely have a 046+ release this week in order to address the DNS exit problem
17:05:37 <ahf> so 0.4.6 and 0.4.7 new alpha without the CC stuff in it?
17:05:40 <dgoulet> once ahf and I talk, I'll make an announcement with regards to timeline
17:05:47 <ahf> and then we stop taking in other stuff for 0.4.7 until the CC things land?
17:05:49 <ahf> ok
17:05:49 <dgoulet> ahf: correct, well 046 stable and 047 alpha
17:05:50 <ahf> sounds good
17:05:57 <dgoulet> yes
17:06:16 <ahf> i have a boring discussion item:
17:06:20 <ahf> [2021-12-13] Holiday is getting closer. Last Monday meeting the 20th and last Thursday meeting the 16th? Let's skip Thursday meeting the 6th of January and have our first 2022 meeting the 10th of January, 2022?
17:06:40 <dgoulet> fine by me
17:06:42 <ahf> this is a question, but are people OK with this? i don't think we need to have a meeting the 6th when we all return and i assume some people take those two days off
17:07:00 <nickm> No objection here
17:07:05 <ahf> this means that our upcoming thursday meeting will be the last one here in 2021
17:07:17 <GeKo> sounds good, fwiw
17:07:20 <ahf> mikeperry: is that OK for your CC plans? we can spawn the team if needed for anything of course
17:08:25 <ahf> :o
17:08:54 <mikeperry> yeah, should be fine. There is some discussion I want to have re what to prepare for sim runs before the break vs after
17:09:13 <jnewsome> sgtm
17:09:13 <ahf> ok, but that is with the simulator gang mostly, right?
17:09:35 <gaba> the s61 meeting of january will be on the 10th I assume
17:09:49 <ahf> yeah, i'd say we let the 10th be the "first meeting of the month" :-)
17:09:59 <mikeperry> yeah. we could potentially exercize negotiation during this time, if we make a new branch based on the current draft
17:10:01 <GeKo> heh
17:10:11 <ahf> sounds good
17:10:20 <ahf> mikeperry: the next item here is just s61 things, so you wanna take over that?
17:10:31 <mikeperry> ok
17:10:52 <mikeperry> yeah, so jnewsome got us 5 ephemeral cloud runners, so we have been churning through the sim plan at a fast pace now
17:11:03 <mikeperry> too fast for me to keep up on weekends :)
17:11:30 <ahf> :o
17:11:38 <ahf> very nice
17:11:48 <ahf> how on earth did you get that? :o
17:11:50 <dgoulet> nice
17:11:59 <mikeperry> jnewsome and hiro also worked to re-run the baseline using Rob's September flooding period and matching consensus params
17:12:08 <ahf> cool
17:12:16 <jnewsome> chameleon cloud - micah has a substantial quota grant for tor-related activities
17:12:30 <ahf> awesome!
17:12:36 <ahf> micah the georgetown person?
17:12:40 <jnewsome> right
17:12:55 <ahf> awesome, thank him for that \o/
17:13:25 <jnewsome> will do :). yeah it's been a big help
17:13:44 <mikeperry> we have several options for what to do next, and what we do depends on what we can get ready
17:14:13 <ahf> cool
17:14:24 <mikeperry> for example, we could plan to take the draft of negotiation, and set up the ability to run mixed network sims
17:15:09 <mikeperry> I am thinking this might be my preferred approach, but that might have too many ingredients to get set up basically this week
17:15:56 <hiro> what would you need on my side mikeperry ?
17:16:11 <hiro> if we go for this approach?
17:17:11 <mikeperry> hiro: I think for you, not much would change. Ultimately we'd still want to keep an eye on the utilization CDFs, so I think that would remain next for you
17:17:27 <hiro> ok sounds good
17:17:35 <mikeperry> but jnewsome would have to add a way to specify alternate branches for some percentages of background clients in Shadow
17:18:04 <jnewsome> i think we could do it relatively simply as a post-processing step for now
17:18:11 <mikeperry> and I would have to add some additional loglines, and scraping, and rebase the experiment branch on top of the negotiation brnach
17:18:21 <jnewsome> is it just the background traffic that we need to split? or also e.g. the perf nodes?
17:18:56 <mikeperry> jnewsome: well, the experiments I am thinking about would be where some of the background markov clients are on 0.4.6-stabke, and some are on 0.4.7
17:19:08 <mikeperry> so maybe this is too complicated
17:19:15 <mikeperry> for this week
17:19:41 <jnewsome> i think that wouldn't be too hard. i was thinking it gets more complicated if you want to split the perf nodes too and then split the results
17:19:57 <mikeperry> ah no, we do not need to split the perf nodes for this
17:20:39 <mikeperry> basically this gives us the ability to see if some/much 0.4.6 background activity affects the 0.4.7 perfclients and 0.4.7 background clients
17:21:21 <mikeperry> and we could add in some cheaters
17:21:41 <mikeperry> if dgoulet has an updated negotiation branch, we could use that. or we could just use mine
17:21:58 <jnewsome> if we're adding cheaters, do we also need to enable DoS mitigations?
17:22:06 <jnewsome> tornettools currently disables them in the torrc files
17:22:20 <mikeperry> no. we need to look at the oomkiller logs tho
17:22:34 <mikeperry> and maybe some other loglines
17:24:26 <mikeperry> so if you think this is potentially doable just to add in multiple tor versions for markovclients jnewsome, perhaps we talk more fater the meeting
17:24:28 <ahf> hopefully no jndi ldap loglines
17:24:48 <mikeperry> lol yeah no java here
17:24:54 <jnewsome> mikeperry sounds good
17:25:35 <mikeperry> GeKo,juga: any news from sbws and network-health?
17:26:05 <juga> mikeperry: only main new is that we figured out an issue with gabelmoo
17:26:28 <juga> #40112
17:26:44 <juga> it was the web server keep-alive timeout
17:26:44 <mikeperry> it slightly concerns me to be potentially down a geko in January and after.. keeping an eye on the network when CC is turned on will be helpful.. but perhaps juga and hiro can help cover that
17:26:44 <GeKo> no network-health news this week apart from dgoulet and i making a plan for dealing with dns overload warning fatigue
17:27:16 <GeKo> we think we need to remove dns as an overload indicator
17:27:40 <mikeperry> yeah, I saw. unfortunate but seems like the best plan
17:27:41 <GeKo> it's not usable mostly due to external factor we can't control
17:27:53 <mikeperry> I had not anticipated that DNS servers would blackhole us
17:27:59 <mikeperry> but now we know that is a thing :/
17:27:59 <GeKo> yeah
17:28:30 <GeKo> mikeperry: re: jan, we'll see i am optimistic to get back to network-health then
17:28:42 <GeKo> if not, i'll come up with a plan b
17:29:20 <mikeperry> GeKo: ok. yeah most likely we won't flip the switch until more like Feb, depending on the upgrade rates, etc. these negotiation sims will help us learn more when it is good to do that
17:29:40 <GeKo> sounds good. no worries either way
17:30:50 <mikeperry> the break may bring up new surprises for us, and some distracttion when we all come back to our exploded inboxes, heh
17:31:09 <GeKo> (i gotta go now, ttyl everyone o/)
17:31:12 <ahf> o/ GeKo
17:31:42 <ahf> anything else for today? otherwise let's call it
17:32:11 <mikeperry> I am good
17:32:19 * dgoulet good
17:32:22 <ahf> anybody else who have anything?
17:32:56 <ahf> okay then
17:32:58 <ahf> #endmeeting