15:00:30 <karsten> #startmeeting metrics team
15:00:30 <MeetBot> Meeting started Thu Mar 30 15:00:30 2017 UTC.  The chair is karsten. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:00:30 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic.
15:00:40 <karsten> https://pad.riseup.net/p/3M7VyrTVgjlF <- agenda pad
15:00:44 <karsten> what's missing?
15:01:34 <iwakeh> more than enough for 45min ;-)
15:01:50 <karsten> indeed. shall we start?
15:01:58 <iwakeh> yep.
15:02:06 <karsten> okay.
15:02:10 <karsten> - Amsterdam meeting report (karsten)
15:02:17 <karsten> I pasted some links to notes to the pad.
15:02:43 <karsten> lots of good discussions and no worries about running out of work soon.
15:03:09 <hiro> hehe
15:03:18 <karsten> the joint ooni meeting brought up some good ideas for getting more user input.
15:03:33 <iwakeh> cool
15:03:55 <karsten> but in general, take a(nother) look at these notes and see what we could use in the future.
15:04:14 <karsten> maybe we can do this when we meet in berlin.
15:04:24 <iwakeh> yes.
15:04:39 <karsten> okay, I don't have more to say there. let's move on, unless there are questions.
15:05:10 <karsten> - In-memory statistics (iwakeh)
15:05:20 <iwakeh> Yes,
15:05:22 <karsten> want to give a quick update?
15:05:39 <iwakeh> I changed some parameters and run
15:05:56 <iwakeh> (more than what I mailed earlier)
15:06:17 <iwakeh> run the simulation, which is quite
15:06:20 <karsten> so, this is about the simulation of obfuscating dirreq stats.
15:06:28 <iwakeh> helpful for writing the the report.
15:06:30 <karsten> by adding laplace noise and/or binning.
15:06:37 <karsten> removing binning, that is.
15:06:38 <iwakeh> yes.
15:07:05 <karsten> did you come up with good parameters yet?
15:07:22 <iwakeh> No, but its just running.
15:07:38 <iwakeh> I multitask a bit
15:07:52 <karsten> ok. so, I have ignored that simulation yesterday and today. should I resume doing that tomorrow, or rather not yet?
15:07:54 <iwakeh> and look at the (possible )patch
15:08:20 <iwakeh> hmm, I think I need
15:08:41 <iwakeh> to clarify some of the suggestions I made
15:08:50 <iwakeh> (just noticed your mail)
15:09:11 <iwakeh> just wait a little
15:09:17 <iwakeh> until monday?
15:09:20 <karsten> sure!
15:09:51 <iwakeh> Very useful to do the simulation!
15:09:55 <karsten> oh, yes.
15:10:04 <karsten> very dangerous not to do it. ;)
15:10:11 <iwakeh> true,
15:10:23 <iwakeh> better to tweak in advance.
15:10:30 <karsten> yes.
15:10:47 <karsten> okay, anything we should discuss today rather than next week, wrt in-memory stats?
15:11:01 <iwakeh> I'm fine for today.
15:11:07 <karsten> ok. :)
15:11:13 <karsten> moving on:
15:11:17 <karsten> - OnionPerf deployment (hiro)
15:11:24 <hiro> yes
15:11:26 <karsten> hiro: how are our op-?? going?
15:11:45 <hiro> so here is the thing, onionperf needs the connect and the listen port to be the same
15:11:56 <karsten> why?
15:11:59 <hiro> if they are not the same there needs to be some portforwarding
15:12:07 <karsten> ah, yes.
15:12:22 <iwakeh> really?
15:12:29 <hiro> and to have portforwarding as you saw (maybe) with Mart email I would need to do some tweaking with the modules in the kernel of the VM
15:12:43 <hiro> so I am waiting on some confirmation on the last questions I sent him
15:12:58 <karsten> the connect port is something that the exit node sees,
15:13:05 <karsten> which then tries to connect to that port.
15:13:08 <karsten> which should be 80.
15:13:22 <karsten> but we want to listen on a high port to avoid permissions issues.
15:13:25 <karsten> like 8080.
15:13:37 <hiro> ok
15:13:37 <karsten> now we need to port-forward 80 to 8080 internally.
15:13:43 <hiro> yes
15:13:51 <karsten> didn't you say you're using nginx for that?
15:14:08 <hiro> nginx doesn't do port forwarding... only proxy
15:14:23 <hiro> it will have to be done w iptables
15:14:44 <karsten> and proxy does not work? or does it affect measurements?
15:14:53 <hiro> so I can say on nginx if something is on 8080 serve it on 80
15:15:06 <irl> does tgen act as an http server?
15:15:12 <irl> i didn't think it did that
15:15:31 <karsten> not http.
15:15:34 <karsten> I think.
15:15:43 <irl> this would also cause nginx to buffer data before passing it, which will mess up timings
15:15:45 <hiro> no it doesn't but if I have tgen running on 80 and I try to do 8080 to 80 on nginx i see something on the measurements
15:16:02 <karsten> irl: okay, that makes sense.
15:16:10 <karsten> hiro: but did you see successful measurements?
15:16:15 <karsten> also, hi irl! :)
15:16:20 <irl> hi karsten (:
15:16:29 <hiro> nope :( i think tgen will see the port blocked by nginx
15:16:41 <karsten> I think, overall, having an iptables rule for this would be best.
15:16:45 <hiro> yep
15:16:58 <irl> i can test this out on the one at aberdeen
15:17:12 <irl> karsten: https://onionperf.erg.abdn.ac.uk/ is now running
15:17:25 <hiro> it's a quick nat rule I just want to understand exactly what we will miss as per Mart email
15:17:36 <karsten> irl: neat!
15:17:56 <hiro> on a side note I gave irl a VM in hk for pathspider :)
15:18:07 <karsten> irl: first question: can you change the subdomain to contain the source name in first place? so, something.erg.abdn.ac.uk, where something is the source name?
15:18:24 <irl> source name?
15:18:39 <iwakeh> replace onionperf.
15:18:48 <karsten> hmmm, hiro, where did you configure the source name?
15:19:29 <hiro> it's an onionperf option
15:19:32 <hiro> hold on
15:20:19 <hiro> -n
15:20:26 <irl> what should it be set to?
15:20:27 <hiro> when you start onionperf
15:20:43 <karsten> irl: depending on what name you want to use.
15:20:47 <hiro> the hostname you want to appear in the files
15:20:55 <karsten> irl: ideally, users would be able to make _some_ sense of it.
15:21:06 <karsten> when comparing your instance's measurements to other measurements.
15:21:23 <karsten> that's why we picked op-hk for tor's onionperf instance in hong kong.
15:21:27 <karsten> and op-nl and op-us.
15:21:48 <karsten> but that was before rob came along with phantomtrain which currently happens to run in ca and which ran in us before.
15:22:13 <karsten> anyway, onionperf as source name is not so good. anything else is better.
15:22:18 <hiro> -n STRING, --nickname STRING
15:22:18 <hiro> the 'SOURCE' STRING to use in .tpf torperf stats files
15:22:18 <hiro> produced by OnionPerf (default: op-nl)
15:22:21 <irl> op-ab?
15:22:35 <irl> or op-gb?
15:22:48 <irl> (ab for Alba - the theorietical ISO code)
15:22:54 <iwakeh> op-uk?
15:23:06 <irl> uk isn't the ISO3166-1
15:23:11 <iwakeh> oh.
15:23:16 <irl> it's gb
15:23:17 <hiro> what would be the country code for scotland? ;)
15:23:20 <irl> .ab
15:23:32 <irl> there was talk over the independence debate
15:23:33 <iwakeh> well, then ab.
15:23:39 <irl> heh
15:23:44 <iwakeh> op-ab
15:23:50 <karsten> yeah, op-ab sounds good.
15:24:39 <karsten> hiro: I have another request:
15:24:42 <hiro> sure
15:24:57 <karsten> we're currently running a torperf instance on one of ln5's hosts.
15:25:02 <hiro> yes
15:25:08 <hiro> shall we shut it down?
15:25:09 <karsten> which we want to retire very soon.
15:25:20 <karsten> well, he'd like to run an onionperf instance on that host.
15:25:26 <hiro> ah ok
15:25:28 <hiro> nice
15:25:33 <karsten> can you talk to him and tell him how to set it up?
15:25:38 <hiro> sure
15:25:58 <karsten> he might set it up and then hand it over to their local admins.
15:26:11 <karsten> but it would be good to have your input for the initial setup.
15:26:29 <karsten> okay, let me make an introduction for this topic. (you may have met him in ams.)
15:26:34 <karsten> thanks!
15:26:40 <hiro> sure, thanks
15:27:10 <karsten> okay, anything else related to onionperf?
15:27:12 <irl> ok, hostname is now op-ab but the dns will take a while to replicate, i've added the name op-ab
15:27:20 <irl> how does this feed collector?
15:27:43 <irl> or shall we have that discussion next week once it's run for a while
15:27:44 <iwakeh> collector fethces the files.
15:27:45 <hiro> I think we are good on my side. Hope will have that port forwarding sorted tomorrow so we can start anew for next week
15:27:46 <karsten> irl: I need to tell it to fetch from it.
15:27:56 <karsten> hiro: great!
15:28:34 <karsten> okay, moving on?
15:28:50 <irl> yep
15:29:05 <karsten> - Monthly report for March (karsten)
15:29:19 <karsten> I'm collecting items for the report.
15:29:22 <karsten> I have:
15:29:27 <karsten> - Atlas blog post
15:29:27 <karsten> - CollecTor and Onionoo pages on Metrics
15:29:28 <karsten> - Wrote tutorials for metrics-lib and created a page on Metrics (#21379). https://metrics.torproject.org/metrics-lib.html
15:29:48 <karsten> what else am I missing? we're looking for completed tasks or completed substeps.
15:29:53 <iwakeh> that looks like the latest javadoc for metrics-lib :-)
15:30:02 <iwakeh> the ams meeting plans?
15:30:22 <karsten> added javadocs.
15:30:30 <karsten> meeting plans?
15:30:33 <iwakeh> the results from ams.
15:30:47 <iwakeh> the trac pages and notes?
15:30:50 <Samdney> do we have results(?)
15:31:01 <iwakeh> well, plans are a result
15:31:28 <iwakeh> aren't they?
15:31:45 <karsten> these notes are basically brainstorming notes.
15:31:54 <karsten> which may turn into plans, but I think they are not plans yet.
15:31:57 <iwakeh> ok.
15:32:04 <karsten> which is different from the OKR thing, if we can agree on that.
15:32:15 <karsten> maybe we should add that?
15:32:27 <hiro> the plans are like long term okrs?
15:32:30 <iwakeh> well, something about meeting ooni folks and nrl?
15:33:12 <karsten> okay, adding a note for those meetings.
15:33:38 <karsten> hiro: I think long-term OKRs exist as a concept, but we didn't do those.
15:33:51 <karsten> like, we could define OKRs for a year, not just a quarter.
15:33:57 <karsten> but we're not there yet.
15:34:16 <iwakeh> that's needs more thinking.
15:34:29 <hiro> karsten: ok. I thought the "plans" from the meeting were things that we could do long term.
15:34:49 <Samdney> +1
15:35:14 <karsten> they are certainly input for long-term plans, but I don't think they're plans yet.
15:35:23 <hiro> ok thanks
15:35:35 <iwakeh> they need some sorting and arranging to be defined OKRs.
15:35:43 <karsten> let's talk about OKRs in a bit.
15:35:47 <karsten> for Q2.
15:35:49 <iwakeh> but good guidelines.
15:36:01 <karsten> Sebastian: are you around?
15:36:56 <karsten> - a couple of topics that came up during AMS (Sebastian)
15:37:20 <karsten> Sebastian just dropped three questions/tasks on the pad.
15:37:30 <karsten> but apparently doesn't have internet for the next 30 minutes.
15:37:32 <karsten> :D
15:37:35 <hiro> maybe he is still on the train with limited connectivity
15:37:38 <karsten> yep.
15:37:56 <iwakeh> well, maybe he manages to return soon.
15:38:04 <karsten> he said 30 minutes.
15:38:13 <karsten> anyway, let's move that to the end.
15:38:18 <karsten> maybe 30 minutes are in fact only 15.
15:38:27 <karsten> - Objectives and Key Results for Q2 (karsten)
15:38:36 <karsten> I pasted something on the pad.
15:38:44 <karsten> which is the result of discussions with iwakeh and isabela.
15:39:00 <karsten> this is a possible plan for Q2/2017.
15:39:16 <karsten> where the main focus is finishing external deliverables that all end in q2.
15:39:33 <karsten> but we also added a few more tasks.
15:39:53 <karsten> the idea is that each key result (sub item) is measurable somehow.
15:40:04 <iwakeh> OKR style.
15:40:13 <karsten> and at the end of q2 we'll see how much of these tasks we completed.
15:40:24 <karsten> where it's okay to complete only 0.7 or 0.8 of all the things.
15:40:42 <karsten> I'd like to put this plan on the metrics team wiki page and also tell the other teams about it.
15:40:57 <karsten> this plan or whatever plan we come up with by, say, tomorrow evening.
15:41:17 <iwakeh> should we just add comments to this pad?
15:41:21 <karsten> yes, please!
15:41:32 <iwakeh> (not that I had one, yet ;-)
15:41:53 <karsten> heh. yes. please add comments, ideally after the meeting or tomorrow morning.
15:41:59 <iwakeh> so Draft until tommorow evening
15:42:06 <karsten> even a comment saying "these are perfect" is fine.
15:42:10 <iwakeh> tomorrow
15:42:12 <karsten> well, submit tomorrow evening was my plan.
15:42:16 <iwakeh> ok :-)
15:42:21 <karsten> because that's when we run out of march.
15:42:48 <karsten> ok!
15:43:02 <karsten> - Berlin meeting planning (karsten)
15:43:09 <karsten> two possible dates:
15:43:15 <karsten> - Wedneday, April 26
15:43:15 <karsten> - Thursday, April 27
15:43:29 <karsten> it will be a one-day meeting.
15:43:49 <karsten> any early thoughts on these dates?
15:43:55 <karsten> should we do a doodle?
15:44:02 <irl> i wonder about cost/benefit for my travel to berlin for a single day meeting
15:44:50 <hiro> i have something on the 26th, so I could travel the 27th...
15:44:54 <karsten> fine question. maybe we can start a list of possible benefits/topics to talk about.
15:45:17 <iwakeh> maybe on a meeting pad?
15:45:25 <irl> that sounds good
15:45:27 <hiro> yes!
15:45:31 <Samdney> y
15:45:34 <karsten> okay. want to create one?
15:45:42 * iwakeh creating ...
15:45:43 <karsten> we can leave that open until next week and talk more.
15:45:52 <irl> ok
15:45:55 <Samdney> ok
15:46:08 <hiro> ok
15:46:08 <iwakeh> https://pad.riseup.net/p/SFYYn1aEpkPm
15:46:29 <karsten> great!
15:46:53 <karsten> shall we briefly talk about Sebastian's topics, even though he's not here?
15:47:15 <karsten> - it'd be great if we had a feature where we can see which dirauth dictated the median for a given relay wrt bandwidth weight (perhaps also with #relays over time)
15:48:10 <karsten> hmm, I'd have to speculate use cases here.
15:48:28 <karsten> does this make perfect sense to anybody else here, without further explanation?
15:48:36 <iwakeh> not really
15:48:55 <irl> should dirauth maybe be bwauth?
15:49:00 <karsten> in which case I'd say let's move this to metrics-team@ or to next week.
15:49:03 <karsten> irl: yes.
15:49:08 <karsten> still. :)
15:49:08 <iwakeh> yes
15:49:14 <irl> no idea
15:49:22 <karsten> - webstats takeover and using them for graphs (mail from isa iirc)
15:49:22 <irl> list sounds good to discuss
15:49:25 <karsten> ok.
15:49:39 <karsten> so, we'll need to take over webstats.
15:49:51 <karsten> I wonder if we should just re-implement it in collector.
15:49:52 <iwakeh> those logs etc?
15:50:09 <irl> collector sounds like the "right" place, is it the easiest?
15:50:14 <karsten> rather than become responsible for that fine collection of scripts and reimplement it a few months later.
15:50:26 <iwakeh> yes, collector seems right.
15:50:41 <irl> http://www.java2s.com/Code/Java/Development-Class/ParseanApachelogfilewithRegularExpressions.htm
15:50:43 <irl> easy
15:51:24 <karsten> it shouldn't be too hard.
15:51:40 <karsten> of course it's yet one more thing.
15:51:51 <iwakeh> collector it is.
15:52:01 <iwakeh> collecting more and more.
15:52:06 <iwakeh> ;-)
15:52:12 <karsten> I mean, it is the right place longer term.
15:52:32 <iwakeh> how urgent is the 'takeover'?
15:52:37 <karsten> the alternative is to move it over later. but I'm not sure if that wouldn't just create more work overall.
15:52:40 <karsten> overdue, I think.
15:53:03 <karsten> let me find out more and let's discuss this again next week.
15:53:09 <iwakeh> yes.
15:53:20 <karsten> the next:
15:53:23 <karsten> - bwauth feeding servers monitoring
15:53:33 <karsten> seems out of scope.
15:53:51 <karsten> it's part of operating the tor network.
15:53:58 <iwakeh> ah, ok.
15:54:18 <karsten> and we're not particularly great at monitoring services.
15:54:20 <iwakeh> maybe find out who should take over this.
15:54:35 <karsten> hiro: maybe this is something for tor's nagios?
15:54:41 <hiro> was thinking about it
15:54:55 <hiro> tor nagios or something of the like
15:55:27 <karsten> okay, let's just invite Sebastian to next week's meeting.
15:55:33 <karsten> and talk about this more.
15:55:43 <karsten> look how fast 45 minutes pass.
15:55:50 <karsten> are we out of topics?
15:56:11 <iwakeh> apparently
15:56:25 <hiro> yep
15:56:29 <karsten> cool!
15:56:33 <Samdney> :)
15:56:41 <karsten> please add comments to the OKRs.
15:56:59 <karsten> including any obvious objectives or key results that we missed.
15:57:16 <iwakeh> oh, adding topics?
15:57:26 <karsten> only if we missed obvious ones.
15:57:38 <karsten> ah, housekeeping tasks should not be added, I learned.
15:57:49 <karsten> like "fix new bugs as they come in".
15:57:50 <iwakeh> well, they take up time
15:57:57 <karsten> yes, we need to account for that differently.
15:58:20 <karsten> (this is just the short version of isabela's statement about this.)
15:58:27 <iwakeh> good housekeeping saves a lot. but that's always like this, sigh.
15:58:40 <karsten> oh, that's not what she says.
15:59:00 <iwakeh> ok
15:59:02 <karsten> I guess it's more that we should put fewer O's and KR's on the list to account for housekeeping.
15:59:28 <karsten> but I also see your point how that gets less attention if it's not on the list.
15:59:38 <karsten> anyway, we're experimenting here. the Q3 list will be better.
15:59:40 <karsten> ;)
15:59:44 <iwakeh> ;-)
15:59:49 <karsten> anyway, let's talk more next week!
15:59:50 <hiro> when I was in google I learned it is about how you frame it
16:00:07 <karsten> oh, great, you have done this in the past!
16:00:23 <hiro> so if you say make the code base this and that way so that we will have fewer bugs it's fine
16:00:32 <hiro> but if you say solve bugs as they come in it is not ...
16:00:38 <karsten> makes sense.
16:00:48 <iwakeh> hmm.
16:00:57 <hiro> yeah, but depends a lot on how people that have to judge what you do react to it
16:01:10 <iwakeh> yes, context is important.
16:01:44 <hiro> so i think in our case it is more up to us to work better? and present better what we do?
16:01:50 <nmago> Hi!  I would like to ask, if I wrote proposal, have I to send it to gsoc site as draft?
16:02:11 <karsten> hiro: better visibility to other teams.
16:02:23 <karsten> hiro: being more explicit about our plans, also to ourselves.
16:02:40 <karsten> okay, I'll end this meeting now. 45 minutes are over. ;)
16:02:44 <hiro> ok :)
16:02:46 <iwakeh> sure.
16:02:47 <hiro> talk next week
16:02:48 <Samdney> :)
16:02:48 <karsten> thanks, everyone! :)
16:02:52 <karsten> #endmeeting