16:59:48 <ahf> #startmeeting Network team meeting, 30rd August 2021
16:59:48 <MeetBot> Meeting started Mon Aug 30 16:59:48 2021 UTC.  The chair is ahf. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:59:48 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic.
16:59:57 <ahf> hello hello
17:00:05 <ahf> meeting pad at https://pad.riseup.net/p/tor-netteam-2021.1-keep
17:00:10 <dgoulet> o/
17:00:18 <ahf> o/
17:00:23 <ahf> nick is out this week
17:00:45 <ahf> how are folks doing with their tpo/core boards?
17:01:29 <dgoulet> stalled a bit I would say
17:01:50 <dgoulet> but that was expected last week
17:01:51 <ahf> ya
17:02:17 <ahf> is good. i have some s30 stuff still and then i'll also dive into arti land i think
17:02:25 <dgoulet> exciting
17:02:30 <ahf> i don't think we have anything to talk about release wise
17:02:43 <ahf> haven't seen anything security related or something that could warrent activity there
17:02:56 <ahf> i guess you and i have some follow up to do from our lessons learned from doing the last release series
17:03:02 <dgoulet> no and we are likely going to release an alpha on Sept 15th but keep the merge window open
17:03:04 <dgoulet> to accomodate s61
17:03:08 <ahf> ya
17:03:10 <ahf> very good
17:03:27 <dgoulet> that is about "the" news of last week about releasees eheh
17:04:00 <ahf> is the s61 release thing something new that happened while i was in the woods?
17:04:06 <dgoulet> yes
17:04:14 <dgoulet> informal decision on our Thu. sync
17:04:18 <ahf> what caused it?
17:04:25 <ahf> just out of curiosity
17:04:37 <dgoulet> simply that we could be tight to get congestion control in by that date
17:04:53 <dgoulet> so we simply opened the door to expand the merge window while still doing an alpha
17:05:00 <ahf> sweet, good call
17:05:26 <ahf> i met Toke that mike perry have worked on congestion control with. promised him a beer but then there was some emergency and i missed him
17:05:36 <ahf> he had found it to be very interesting to help out in those threads
17:05:37 <ahf> ok!
17:05:52 <ahf> don't see any reminders or discussion topics
17:06:00 <ahf> mikeperry: do you wanna run the s61 part?
17:06:11 <ahf> next week is the bbb version of the s61 meeting
17:06:12 <mikeperry> sure
17:06:32 <GeKo> ahf: you could think abuot what to do with the hackerone tor tickets
17:06:45 <GeKo> i think you wanted to at least at some point
17:06:46 <dgoulet> wait, there are some open stil?
17:06:51 <GeKo> yes
17:06:53 <ahf> oh, i thought that had been handled
17:06:57 <GeKo> no
17:07:08 <dgoulet> I'm monitoring the new ones but didn't know we had some open before asn left :S
17:07:11 <dgoulet> ok damn, I'll get on that
17:07:16 <GeKo> thx
17:07:20 <ahf> dgoulet: awesome, thanks
17:07:33 <ahf> i have noted i need to figure out how to enter those too - i have the user there
17:08:48 <mikeperry> so the immediate thing for s61 is getting the shadow sims running full scale, for baseline
17:09:20 <mikeperry> jnewsome and hiro are working on that. you folks here?
17:09:25 <jnewsome> yup
17:09:49 <jnewsome> i made some progress last week saving the raw logs and doing some extra compression
17:10:33 <jnewsome> unfortunately the simulation breaks after ~15m sim time in the latest shadow. spent some time tracking that down
17:11:13 <jnewsome> I have it pinned to a slightly older shadow version now and it's working again. should be able to make some more progress this week
17:11:43 <jnewsome> still debugging why it's broken on new-shadow but it's a slow debugging process; don't want to further block the other tasks on it
17:12:28 <ahf> you think it's gitlab ci runner related?
17:12:56 <jnewsome> hmm, no particular reason to think so, no
17:13:06 <mikeperry> were you able to add the oniontrace logs for hiro? maybe that ended up causing more resources to be used?
17:13:36 <jnewsome> the logs are there now; I still need to add the extra event types that were requested. I should be able to do that today
17:14:06 <jnewsome> it's definitely a shadow issue - narrowed it down to working at one commit and not working on the next
17:14:13 <jnewsome> but it's a tricky commit hence still debugging
17:14:23 <ahf> ok
17:15:19 <mikeperry> aight
17:15:54 <mikeperry> were you able to scale up the sim size with the old code, or is this new fix needed fto help that?
17:16:40 <jnewsome> i haven't tried scaling up again yet; that's the next task once i have outputs for hiro to start working with
17:16:53 <mikeperry> ok
17:18:08 <mikeperry> I have been tweaking the flow control impl logic. it has been slow going, but the branch is at the point where dgoulet can hook up a token bucket to make it rate limit
17:18:45 <mikeperry> I still need to figure out some other things, like protecting against dropmark attacks
17:18:49 <mikeperry> but I will likely be distracted this week
17:19:06 <mikeperry> dgoulet: did you see the checklist I put on https://gitlab.torproject.org/tpo/core/tor/-/issues/40450 ?
17:19:22 <dgoulet> I did not but now I do, nice!
17:19:58 <mikeperry> dgoulet: so I can squash that down and it is more or less ready for a prelim review, rate limiting, and the oomkiller things on there
17:20:35 <dgoulet> great
17:20:43 <dgoulet> added on my todo
17:20:54 <mikeperry> I probably will just put it down for a bit so you can do that
17:22:58 <mikeperry> you can hit me up if you have questions. I should be able to text chat, maybe signal call. not sure if I will have bandwidth for a BBB thoh
17:23:14 <dgoulet> ack
17:23:35 <gaba> remember that next monday is a US holidays... do we still do the BBB s61 check that monday or we can move it to tuesday?
17:24:04 <mikeperry> oh yeah, lets move it
17:24:29 <mikeperry> juga: you had questions about the exit pinning thing for sbws?
17:24:47 <GeKo> juga might not be around due to hardware problems :(
17:25:09 <GeKo> but, yes, we are still stuck in sbws#40022
17:25:16 <mikeperry> mercury definitely in retrograde still
17:25:41 <GeKo> it seems arma2's idea is not the overall favorite of the network team
17:25:49 <GeKo> i mean that's something at least :)
17:26:03 <dgoulet> I recall asking what was ideal for juga with regards to .exit or single onion in that case
17:26:15 <dgoulet> pinning RP on a single onion I mean
17:28:22 <mikeperry> hrmm yeah. both seem like they will require some hax
17:28:29 <mikeperry> either approach, I mean
17:28:46 <dgoulet> pinning RP option for me is something I much prefer than bringing back .exit notation
17:29:11 <mikeperry> yeah, .exit is dangerous
17:32:10 <mikeperry> well we have maybe another week or two before flow control is ready. esp since I will be distracted this week
17:32:20 <mikeperry> but I may also just be mostly bored. we'll see, heh
17:32:58 <mikeperry> GeKo: anything on network-health front?
17:33:31 <mikeperry> saw you updated the experiment page. did you have a chance to talk with rob about when we need to take over that?
17:34:09 <GeKo> it seems we don't need to
17:34:15 <GeKo> see his last mail
17:34:26 <mikeperry> if I get mail, I will
17:34:32 <mikeperry> but i might not
17:34:33 <GeKo> heh
17:34:36 <GeKo> okay
17:35:00 <GeKo> so, i proposed the following
17:35:17 <GeKo> 1) continue the experiment another two rounds
17:35:33 <GeKo> 2) then turn the setup he has into fast exits
17:35:53 <GeKo> 3) then re-do the experiment with congestion control on
17:36:10 <GeKo> 4) do an analysis of both data sets
17:36:23 <GeKo> 5) decide whether we want to run the flooding permamently
17:36:35 <GeKo> and rob says that sounds good
17:36:46 <GeKo> in particular he'd bewilling to do 3), too
17:36:51 <GeKo> *be willing
17:37:00 <GeKo> so this seems some success to me :)
17:37:07 <mikeperry> ok nice
17:37:21 <GeKo> and i told him that 3) can take until like feb 2023
17:37:31 <GeKo> but he's fine with that
17:37:45 <mikeperry> again we will want to watch for emergent effects once CC is deployed.. hopefully we will be able to gets exits to upgrade quickly
17:38:00 <GeKo> yeah
17:38:30 <GeKo> dealing with that thread during last week was where i spent most of my s61 time
17:38:36 <GeKo> nothing more from my side
17:39:13 <mikeperry> ok
17:39:45 <mikeperry> I think that covers it then. dgoulet - I will try to make a v2 version of that flow control branch that is squashed. I'll let you know if I succeed
17:40:01 <mikeperry> so far internet has held up, but I am down to 1 bar. so who knows
17:40:01 <GeKo> s/if/when/ :)
17:40:03 <ahf> awesome
17:40:27 <ahf> anything else we need to talk about?
17:40:59 <GeKo> <- is fine
17:42:04 * dgoulet good
17:42:15 <ahf> let's call the meeting then
17:42:16 <ahf> thanks all
17:42:18 <ahf> #endmeeting