15:01:13 <karsten> #startmeeting metrics team
15:01:13 <MeetBot> Meeting started Thu Aug 15 15:01:13 2019 UTC.  The chair is karsten. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:01:13 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic.
15:01:21 <karsten> https://storm.torproject.org/shared/5h1Goax5eNusxjXJ_Ty5Wl7hFR1uqCReUiN8xdlBG8T <- agenda pad
15:01:34 <gaba> hi! can we talk about funding proposals first?
15:01:38 * karsten didn't send out a meeting reminder this time, glad you made it anyway!
15:01:45 <gaba> I need to leave in 30 min sharp
15:01:49 <karsten> sure!
15:02:18 <gaba> karsten, irl: do you both have a link to each proposal, right?
15:02:29 <irl> only the "scaling" one
15:02:51 * irl really thinks we should give the proposals names so we can talk about them *before* we get them submitted
15:02:54 <gaba> ok. I sent  you the one on IPv6
15:03:09 <gaba> you do not like "the scaling one" as a name :P
15:03:20 <irl> what if we apply to two places for different scaling things
15:03:44 <irl> ok, i see both proposals now
15:03:45 <gaba> then it gets converted into the scaling-grant-name proposal
15:04:50 <gaba> the scaling one is the one that is mostly work for metrics
15:05:12 <irl> karsten indicates he has connectivity issues
15:05:14 <karsten> sorry, got disconnected.
15:05:17 <karsten> back now.
15:06:18 <gaba> for the scalability one I would love karsten to read it and comment if ther eis any comment
15:06:23 <gaba> and we need estimation
15:06:40 <karsten> what's the timeline for these two?
15:06:49 <gaba> we also need tickets for scalability tasks but that is independent of this proposal
15:06:52 <irl> i only looked at the intro for the scaling one, the rest there are still problems with unless it's been rewritten to match
15:06:53 <karsten> (I didn't open any links yet.)
15:07:18 <irl> the ipv6 one is tricky, i can see exactly what we would need to do there but i don't know if we would have capacity to do it
15:07:24 <gaba> the scalability one is on rolling basis so we do not have a deadline. We want to get it out of the door soon.
15:07:34 <gaba> the ipv6 one is September 1st
15:07:52 <irl> we would need to work on adding new metrics to extra info descriptors for ipv4/v6 metrics and then archive/visualise them to say whether or not things work
15:08:01 <gaba> irl: yes, alsmith is working on the rewritten. we have a very good grants writer :)
15:08:04 <irl> this is probably the same metrics i suggested in my review of the happy eyeballs proposal
15:08:20 <gaba> irl: do not worry about capacity for now
15:08:57 <irl> if i ignore capacity, i agree with teor4 on basically all the ipv6 stuff we've ever discussed
15:09:05 <irl> and the proposal looks good from skimming it
15:09:18 <gaba> ok
15:09:19 <karsten> regarding tickets (as far as that's related), I'm blocking on receiving more input on what's needed.
15:09:39 <gaba> karsten: i do not undertand what you just wrote
15:09:44 <karsten> sorry.
15:09:51 <karsten> regarding those tickets that need to be written,
15:10:04 <irl> scaling tickets?
15:10:06 <karsten> I only have a single sentence stating something very general.
15:10:09 <karsten> yes, those.
15:10:21 <karsten> I'd need to hear more about the actual questions behind those to write something useful as ticket.
15:10:54 <karsten> maybe reading the proposal will shed some light on this?
15:11:04 <gaba> ok. I understood that those are issues already discussed before and that is why they did makes sense. Can you contact mikeperry with any question that needs to be answer to write them?
15:11:55 <gaba> I think it makes sense for you or irl to write them as they are stuff metrics will do. And Mike is quite over capacity right now with stuff that needs to be finish this month.
15:11:57 <karsten> I could try, but last thing I hear is that he's focusing on a sponsor thing.
15:12:08 <gaba> yes, but he can answer some questions
15:12:11 <irl> right now there is no urgent scaling task
15:12:14 <irl> for metrics
15:12:28 <karsten> okay, if that's the case, then let's not rush this.
15:12:43 <gaba> there is no urgent task but we need to move forward on this
15:12:46 <irl> i have some tickets for the "alternative reality simulator" on my whiteboard but i didn't put them in trac yet
15:13:20 <irl> (diff fast/guard cutoffs and how the aggregate stats of the network change by excluding relays, etc)
15:13:21 <karsten> gaba: can this wait for another two weeks?
15:13:42 <gaba> the creation of tickets?
15:13:45 <karsten> irl: sounds great, happy to take a look once it's on trac.
15:13:52 <karsten> yes, creating tickets and moving forward there.
15:14:15 <karsten> this will be more useful with mike's input.
15:14:30 <gaba> mike may be with a lot of other things until end of september
15:14:37 <gaba> i do not think we have to wait for mike here
15:14:45 <gaba> there is a tor/moz meeting next week too
15:14:59 <irl> (reminder: i am away tomorrow until the 28th for ccc camp)
15:15:06 <gaba> ok
15:15:13 <irl> (also so is acute)
15:15:38 <gaba> the most important thing related to scalability is for you to read the proposal and give feedback about it
15:15:47 <karsten> okay.
15:15:55 <karsten> irl: are you going to read email?
15:16:07 <irl> likely not, i'll be in a field
15:16:15 <irl> i can respond to signal messages in an emergency
15:16:27 <karsten> okay, then we should talk about reviews/releases later on.
15:16:33 <irl> sounds good
15:17:45 <gaba> ok
15:17:57 <gaba> TorDNSEL/Check replacement ?
15:17:58 <karsten> what about the ipv6 proposal?
15:18:09 <karsten> does this need review and feedback, too?
15:18:14 <gaba> yes
15:18:25 <gaba> the IPv6 proposal is the one that we are submitting in 2 weeks
15:18:39 <karsten> ok.
15:18:51 <karsten> TorDNSEL/Check replacement
15:19:18 <irl> (https://lists.torproject.org/pipermail/tor-dev/2019-July/013918.html has some thoughts i had on ipv6, including metrics we want)
15:19:55 <karsten> great, added to the pad.
15:20:12 <karsten> so, what's required for #29399?
15:20:36 <karsten> last week I said it's on our current roadmap.
15:20:41 <karsten> when does that end?
15:20:42 <irl> in stockholm, this task moved futher and futher along the roadmap until it nearly fell off
15:21:02 <gaba> uh, that is my fault on not following up on this. We really need to move this up in priorities
15:21:14 <irl> i found it very hard to say this was a priority above other things we have on the roadmap
15:22:03 <gaba> why?
15:22:16 <irl> because other things were either sponsor work or blocking other teams
15:22:29 <irl> this is an emergency that we've manufactured for ourselves
15:22:58 <irl> i would rather compromise on running an older debian for a few months than have another team fail to deliver for a sponsor
15:22:59 <gaba> mmm, i'm looking at the roadmap in trello
15:23:03 <gaba> which one is sponsor work?
15:23:52 <irl> in my decisions i treated scaling as "above non-sponsor work"
15:24:00 <irl> also we have sponsor30 and sponsor28
15:24:16 <gaba> yes, scaling is priority
15:24:27 <irl> right, so this is how it ended up at the end of the roadmap
15:24:36 <irl> it's still included but right now no effort is being made on it
15:24:37 <gaba> there is a few things on sponsor 30 (that still technically didn't start yet)
15:24:46 <karsten> so, the reason we're discussing tordnsel now is that buster got stable, right?
15:24:58 <irl> i believe so
15:25:00 <gaba> and s28 there is also a few things but we can move this after this other one.
15:25:24 <karsten> if that's the case, it's a very indirect change to the situation we were in before.
15:25:47 <karsten> we said that EOL in mid-2019 (?) is our hard deadline.
15:25:52 <karsten> mid-2020
15:25:54 <karsten> not 2019
15:26:08 <karsten> and we're already in a situation where we don't have backups for the machine.
15:26:35 <gaba> yes. I think TorDNSEL should be a priority after scalability now.
15:26:40 <karsten> but why?
15:26:50 <karsten> I'm saying that not much has changed.
15:27:20 <karsten> the priority considerations from two weeks ago are still valid.
15:27:21 <gaba> because it seems that we are blocking other team on this and we may have a bigger problem with it later
15:27:26 <gaba> karsten: yes
15:27:31 <irl> i also think that having it at the end of the roadmap is the right place for it, it started out at the beginning of the roadmap in stockholm and systematically everything else was a priority above it
15:27:35 <gaba> we need to plan to include the TorDNSEL in the roadmap
15:27:42 <irl> it is in the roadmap
15:28:05 <gaba> omg... i was looking at a filtered roadmap :|
15:28:13 <karsten> (you can filter the roadmap?!)
15:28:19 <irl> i think i had it starting end of october/start nov on the stickies
15:28:24 <gaba> (yes... it was focus on scalability)
15:28:55 <gaba> we have #29653 in the backlog
15:29:24 * phw status: merging and deploying #9316
15:29:34 <karsten> so, 5 seconds before gaba disappears.
15:29:41 <karsten> what do we say on the tordnsel ticket?
15:29:53 <karsten> "EOY seems realistic"?
15:29:54 <gaba> when do we start working on it? how long it will take to complete?
15:30:29 <karsten> likely not before Q4, depends on what version of our plans we're going to implement.
15:30:49 <karsten> but I think that's more information than required here.
15:31:26 <gaba> We could start in October and be done by end of December?
15:31:51 <karsten> sounds plausible to me.
15:32:13 <gaba> ok. Let's do that then. I'm moving the proposal for scaling to say that this is metrics can start jan 1st 2020
15:32:20 <gaba> and we wrap all this before
15:32:23 <irl> ok
15:32:38 <karsten> okay.
15:32:47 <gaba> can one of you update ticket #29399?
15:32:50 <irl> (gaba: don't forget your thing, it's been 30 min)
15:32:51 <gaba> with this info
15:32:58 <gaba> yes... runnig
15:33:00 <gaba> o/
15:33:03 <karsten> o/
15:33:06 <gaba> thanks!!
15:33:38 <karsten> should I update the ticket?
15:33:49 <irl> yes, ok
15:34:13 <karsten> alright,
15:34:34 <karsten> let's talk about the next two weeks in the field. :)
15:35:08 <karsten> sounds like reviews will be difficult during that time.
15:35:22 <irl> yes, that is likely
15:35:29 <karsten> can we get the snowflake stuff deployed before you leave?
15:36:03 <irl> i am probably going to be unable to do much more than an hour after the meeting
15:36:08 <karsten> ok.
15:36:13 <karsten> then this won't work.
15:37:08 <irl> is the deployment of collector blocking something?
15:37:33 <karsten> not really, it would just be about finishing this task.
15:37:41 <karsten> but it's fine to do it in two weeks.
15:37:47 <irl> ok
15:38:15 <irl> #31204 would be something to maybe also do in collector and deploy at the same time? i don't think there's a lot of code overlap there
15:38:37 <karsten> yes, true.
15:38:52 <karsten> moved to In Progress.
15:39:23 <karsten> what did you mention earlier about fast/guard flags?
15:39:48 <irl> basically it involves alternate consensus generation
15:40:20 <karsten> is that something I could try to move forward in the coming two weeks?
15:41:00 <irl> i think it might be easier if you start at the other end of this pipeline, which is onionperf data analysis
15:41:01 <karsten> and is it related to this roadmap item? Emulate different Fast/Guard cutoffs in historical consensuses (Aug 19)
15:41:12 <irl> yes, that's the roadmap item
15:41:19 <karsten> what does the Aug 19 part mean there?
15:41:31 <irl> i had things in months in the roadmap
15:41:34 <irl> in stockholm on stickies
15:41:45 <irl> it's a 10 point sticky though
15:41:52 <karsten> is this 2019-08-19?
15:42:38 <irl> i'm not sure why it's due then
15:42:39 <karsten> or 2019-08-31?
15:42:55 <irl> i think this is something gaba has added based on where things were on the roadmap wall
15:43:02 <karsten> ok.
15:43:09 <karsten> anyway, by onionperf data analysis,
15:43:20 <karsten> you mean this other roadmap item:
15:43:23 <irl> identifying poorly performing relays by correlating historical onionperf data
15:43:26 <karsten> Use emulated consensus with historical OnionPerf data to predict Tor performance with modified consensus Fast/Guard cutoff values
15:43:34 <irl> sort of
15:43:49 <karsten> ah, ok.
15:43:52 <irl> i've got plans for the modified alternate reality consensus, but so far not for doing the onionperf data analysis
15:44:35 <karsten> well, the task I suggested was to simply exclude measurements containing relays that would be dropped from the consensus with different fast/guard parameters.
15:44:40 <irl> so we need better tools for identifying poorly performing relays, recalculating network aggregates if we exclude certain relays, and working out some metric of "did we exclude enough poorly performing relays without sacraficing good ones"
15:45:00 <karsten> the task you suggested (I think) is to look at onionperf data and find correlations between slow relays and slow measurements.
15:45:25 <irl> the task is to look at onionperf data and see if it's possible to identify which relay it is that is badly performing
15:45:38 <irl> the data might be too sparse for this
15:45:44 <karsten> maybe, yes.
15:45:51 <karsten> but okay, worth taking a look.
15:45:59 <irl> in which case, yes, then the task is only recalculating based on excluding relays
15:46:19 <irl> we'd want to iterate this quickly, so the tools need to be fairly efficient for doing this
15:46:22 <karsten> I'll run with this.
15:46:26 <irl> awesome (:
15:46:27 <karsten> yeah.
15:46:43 <irl> we did talk about using a postgres db if it helps to speed things up
15:47:04 <karsten> did you start anything there?
15:47:15 <irl> nope, i've only looked at the fast/guard cutoff stuff so far
15:47:16 <karsten> I could probably build such a database.
15:47:23 <irl> ok cool
15:47:28 <karsten> but I don't want to redo work you already thought a lot about.
15:47:41 <irl> all the stuff i've done so far is dir-spec, not onionperf
15:48:28 <karsten> do you mind writing down two paragraphs about what you did?
15:48:41 <karsten> just to be sure that I'm not doing it once again?
15:48:45 <irl> i will put these tickets in trac before i pack up
15:48:52 <karsten> okay, great!
15:49:19 <karsten> roadmap!
15:49:24 <karsten> anything else we need to move around there?
15:49:41 <karsten> next week we won't have a meeting, right?
15:49:55 <karsten> or at least you and acute wouldn't be there.
15:50:02 <irl> i've been moving things as i go along, so everything there is correct on the roadmap
15:50:18 <karsten> well, aren't you in a field?
15:50:24 <irl> i will leave in-progress things in progress though i'm not working on them, as they are in progress i just don't have any work time
15:50:26 <karsten> are you going to work on roadmap stuff this week?
15:50:35 <irl> unless you think we should move them back
15:50:44 <karsten> I don't know how this works in a sprint.
15:51:01 <karsten> you could move them to the top of backlog, maybe?
15:51:17 <karsten> if the goal is to clear In Progress at the end of the sprint/week?
15:51:18 <irl> how about we update each of the tickets with the progress up to the point we are pausing, then put them back on the backlog?
15:51:45 <karsten> I think that would work.
15:51:58 <irl> i guess it is possible we would come back after 2 weeks to disasters and might need to reprioritise, and we don't want to lose state
15:52:06 <karsten> right!
15:52:28 <acute> will move tickets 30763 and 30792 to on review, as I have a merge request planned for today
15:52:50 <irl> ah ok cool
15:53:17 <karsten> okay, great!
15:53:30 <karsten> I think that's all for the roadmap then.
15:53:39 <karsten> any other topics for today?
15:53:46 <acute> a bit of news
15:53:52 <karsten> yes? :)
15:54:01 <acute> I exported all onionperf tickets from trac and imported them into gitlab
15:54:19 <karsten> exciting!
15:54:37 <irl> ah yes, i've been calling them onionperf#XXX instead of #XXX so we can differentiate
15:54:42 <irl> like on the roadmap
15:55:10 <karsten> makes sense.
15:55:34 <karsten> anything we need to do on the trac side?
15:55:40 <acute> it makes sense to keep them in both places
15:55:55 <acute> the import process does not work for all the comments on the ticket, just for the initial description
15:56:08 <irl> the trac component is archived, the tickets closed, but they are still referenced
15:56:19 <karsten> okay.
15:56:44 <acute> any new OP issue should be opened in gitlab
15:56:45 <irl> i don't know if gaba would be interested in acute's export/import procedure
15:56:50 <irl> should ask when we are back
15:56:55 <karsten> yep!
15:57:20 <acute> :)
15:57:21 <karsten> let's see how this works.
15:57:37 <karsten> looking forward to moving more components to gitlab, but waiting patiently. ;)
15:57:50 <irl> i've found some scary issues that need resolving first
15:58:01 <irl> like doing git push deletes all merge requests
15:58:14 <irl> (i know the fix though)
15:58:17 <karsten> heh
15:58:26 <karsten> good to give it some testing this way.
15:58:34 <irl> yeah
15:58:47 <karsten> alright, we're running out of time for today.
15:58:56 <karsten> enjoy camping!
15:59:09 <acute> thanks!
15:59:17 <irl> (:
15:59:18 <karsten> and talk to you in 2 weeks!
15:59:22 <irl> yep
15:59:30 <karsten> #endmeeting