16:00:57 <GeKo> #startmeeting network-health
16:00:57 <MeetBot> Meeting started Mon Apr 19 16:00:57 2021 UTC.  The chair is GeKo. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:00:57 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic.
16:01:02 <GeKo> hello everyone!
16:01:06 <irl> hello!
16:01:11 <dennis_jackson> o/
16:01:18 <GeKo> we have a new week, and thus a new network-health meeting
16:01:29 <acute> hi!
16:01:31 <GeKo> our pad is at http://kfahv6wfkbezjyg4r6mlhpmieydbebr5vkok5r34ya464gqz6c44bnyd.onion/p/tor-nethealthteam-2021.1-keep
16:01:44 <GeKo> gaba: you around, too?
16:03:02 <ggus> o/
16:03:38 <GeKo> hi gus!
16:04:53 <GeKo> okay, let's get started
16:05:11 <GeKo> i've one item i thought about talking a bit
16:05:36 <GeKo> i've finally found time to start updating our team wiki
16:05:55 <GeKo> the landing page is in shape:
16:05:56 <GeKo> https://gitlab.torproject.org/tpo/network-health/team/-/wikis/home
16:06:07 <GeKo> and there i put some priorities for 2021
16:06:20 <GeKo> let me know if things should get added there
16:06:35 <GeKo> or if we want to set (slightly different prios)
16:06:38 <gaba> yes
16:06:39 <gaba> hi!
16:06:44 <GeKo> o/
16:07:14 <GeKo> this week i'll work on the bad-relays criteria page
16:07:29 <GeKo> i need to clean this more up and focus on the criteria and how we apply them
16:07:34 <dgoulet> that is nice
16:07:46 <GeKo> ggus: it would be nice if we could chat a bit about that topic at some point
16:07:59 <GeKo> because there is a community page, too, about bad-relays
16:08:04 <irl> I guess there should be 6 areas that network health focuses on, because Metrics is missing
16:08:24 <ggus> GeKo: yes, sure. i didn't have time to read the latest wiki page
16:08:26 <irl> ah no, it's number 3, just using different words that didn't immediately recognise
16:08:32 <ggus> but i'm happy to update the community portal one
16:08:32 <GeKo> yeah
16:08:42 <GeKo> irl: let me know if you have better words
16:08:47 <GeKo> or just change the wiki :)
16:09:10 <GeKo> ggus: the latest wiki page is a slightly cleaned-up version from the old trac one i believe
16:09:36 <GeKo> i have some ideas about splitting up the content for the team wiki page and the community portal
16:09:36 <irl> yeah i'll write up some text that covers what I think the metrics team did before, and see how much of it can fit in there
16:09:43 <gaba> irl: we were thinking of integrating metrics as a goal for this year
16:09:49 <GeKo> ggus: we could maybe chat about that tomorrow
16:09:52 <GeKo> ?
16:09:52 <gaba> we can have new goals metrics related for next year
16:10:26 <ggus> GeKo: yes, wfm
16:10:43 <GeKo> yes, i think for 2021 its mainly integrating metrics into network-health and figuring out a good workflow
16:10:56 <GeKo> maybe making progress on high-level metrics goals
16:11:03 <gaba> yes
16:11:07 <GeKo> ggus: great. i'll ping you
16:11:41 <irl> yeah i don't want to add new goals, just cover the things that already exist and should keep existing
16:13:19 <GeKo> okay
16:13:39 <GeKo> hm, i had another item which the pad ate
16:13:58 <GeKo> i wonder where we want to put issues related to the dir-auths
16:14:38 <GeKo> tpo/core/tor#40368 comes to mind
16:14:55 <GeKo> dgoulet: there is a dir-auth project in tpo/core
16:15:03 <GeKo> would that be a good place?
16:15:13 <GeKo> or should we have that somewhere in network-health land?
16:15:34 <GeKo> i am not sure what that dir-auth project is for, actually
16:15:35 <dgoulet> yeah I think so, that DirAuth project I think was created originally for the tools in dirauth conf and anything that we need to do on the dirauth side like add a new parameters and what not
16:15:42 <dgoulet> so moving that to health seems good to me
16:15:53 <dgoulet> since core/ doesn't really use it nor notice it :P
16:16:05 <GeKo> yeah
16:16:26 <GeKo> okay, i'll move it later then and probably write a mail to sina
16:16:54 <GeKo> okay, anything we should discuss related to anyone's updates?
16:18:28 <GeKo> then let's move on to the discussion part
16:18:46 <GeKo> dgoulet: i guess we need to further postpone the libressl thing?
16:19:06 <GeKo> (that is you helping nick to help)
16:19:11 <dgoulet> yeah... that one is hard
16:19:20 <GeKo> i know :(
16:20:10 <GeKo> so be it
16:20:15 <GeKo> dennis_jackson: you are up
16:20:33 <dennis_jackson> I'm a little concerned about the v2 switch-off, from a network-health perspective.
16:21:00 <dennis_jackson> Am I right in understanding that when the new stable hits in July, relays will no longer handle v2 requests from clients?
16:21:23 <dennis_jackson> Or is the new stable only disabling v2 on the client and on the hs's client?
16:21:46 <dgoulet> the former
16:22:03 <dgoulet> every v2 requests won't work _except_ rendezvous
16:22:18 <dennis_jackson> So if a substantial amount of traffic is still on v2, it will become concentrated on the relays which are slower to upgrade?
16:22:33 <dgoulet> correct
16:22:53 <dgoulet> that is why we plan to also release stable (for other versions) a bit after to also stop v2
16:23:58 <dennis_jackson> So this has been thought of and we anticipate a gradual switchover?
16:24:08 <dennis_jackson> If so, cool. I didn't find a ticket which is why I wanted to raise it
16:25:04 <dgoulet> yeah there might be a situation where many services converge to few intro points/HSDir but at least rendezvous where the bigger load happens will still work for all relays
16:25:36 <dgoulet> but yeah, it might be rough
16:25:59 <GeKo> is there anything we can help with form the network-health side?
16:26:02 <irl> any reason that we didn't have a consensus parameter to just turn off v2 services one day?
16:26:28 <dgoulet> no reason I recall
16:27:12 <dgoulet> GeKo: dunno, yet, I haven't thought this through about what health team can do but overload line might start to pop more during that time? I'm not sure tbh since load on intro/hsdir is really not that much but maybe
16:27:41 <GeKo> yeah
16:28:02 <GeKo> might be just a bumpy ride and we need to bite the bullet...
16:28:20 <dennis_jackson> Worst case, we get some fun data to look at.
16:28:29 <dennis_jackson> (but also some sad users and relay operators)
16:28:30 <GeKo> :)
16:28:31 <dgoulet> yeah it might be bumpy
16:28:33 <GeKo> :(
16:28:35 <dgoulet> in different ways...
16:29:08 <dgoulet> but that is what it is, network won't crash, I mean v2 traffic is still a tiny portion of what we are seeing and we'll roll out releases removing v2
16:29:31 <dgoulet> and worst case it crashes our Pi4 relay which personally I wouldn,t mind lol :P .... but yeah, 2021 will be interesting
16:29:43 <gaba> dgoulet: anything related to metrics that we should pay attention to wrt to v2?
16:29:43 <GeKo> dennis_jackson: i liked the experimental analysis pipeline you put up in tpo/metrics/statistics#40002
16:29:51 <GeKo> thanks for doing that
16:29:52 <dgoulet> having v3 traffic
16:29:56 <dgoulet> would be important
16:30:03 <GeKo> very exciting and seems a cool way forward
16:30:04 <dgoulet> in different graph (line)
16:30:09 <GeKo> keep poking at it :)
16:30:09 <dgoulet> yeah go dennis_jackson !!
16:30:15 <GeKo> +1
16:30:20 <dennis_jackson> thanks / np! :-)
16:30:34 <dgoulet> as network will move to 0.4.6 (and historically more than half upgrade in less than 2 months)
16:30:34 <dennis_jackson> It's pretty hacky, but I think using that pipeline / style would alleviate a lot of pressure on the metrics team / person
16:30:50 <dgoulet> if v2 traffic stays the same, we need to definitely check if relay overloading becomes a thing
16:31:33 <dgoulet> the likely big catalyst also for "something exciting" will be when TB ship without v2 stable :)
16:31:36 <dgoulet> anyway
16:32:17 <GeKo> dennis_jackson: yep
16:32:24 <yanmaani> dgoulet: cause the few remaining relays would get all the v2?
16:32:48 <GeKo> dennis_jackson: i think the plan is moving (slowly) towards that
16:33:19 <dgoulet> yanmaani: that is definitely one problem, but also most clients unable to reach v2 might bring unhappy people, rants, relay operators not upgrading, etc...
16:34:03 <dennis_jackson> I worry would have would be if it drives v2 users to switch to another browser or client config
16:34:29 <dennis_jackson> e.g. Brave, older Tor Browser
16:34:40 <dgoulet> won,t help much if the network doesn't allow you though hehe
16:35:11 <dennis_jackson> Well if you were a bad person, you'd find it easier to scoop all the v2 traffic I guess, but that's pretty speculative
16:35:20 <dgoulet> yes...
16:35:22 <dgoulet> definitely
16:35:32 <GeKo> okay, do we have anything else to discuss today?
16:35:43 <dennis_jackson> nothing from me
16:35:52 <ggus> i'm good.
16:35:53 <dgoulet> that is mostly why we will release all maintained version to cut it off so we dn't have relay that can hang around for many years supporting v2
16:35:56 * dgoulet is good
16:36:04 <irl> nothing from me
16:36:08 <gaba> fine
16:36:15 * ggus reading about dennis_jackson experimental analysis pipeline
16:36:20 * gaba too
16:37:01 <GeKo> k, thanks everyone. have a nice week
16:37:05 <dennis_jackson> cool! feel free to ping me if you have questions
16:37:05 <GeKo> #endmeeting