13:59:29 <karsten> #startmeeting Metrics Team
13:59:29 <MeetBot> Meeting started Thu Dec  3 13:59:29 2015 UTC.  The chair is karsten. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:59:29 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic.
13:59:38 <karsten> hello.
13:59:41 <tomlurge> hi there!
13:59:43 <karsten> who's here for the meeting?
13:59:46 <karsten> hi tomlurge!
13:59:49 <karsten> https://pad.riseup.net/p/zUNzEIFRq5S4 <- agenda
14:00:00 <Letty> hi karsten
14:00:01 * qbi is, but need to do some other work also.
14:00:08 <karsten> hi Letty and qbi!
14:01:02 <karsten> feel free to add more topics.
14:02:53 <karsten> in the meantime I'll start with the first topic:
14:02:58 <karsten> * Misc team-related stuff (karsten)
14:03:19 <karsten> so, there was no meeting with other vegas team leads two weeks ago, even though it was planned.
14:03:35 <karsten> but I asked the applications team about their meeting logs.
14:03:42 <karsten> http://meetbot.debian.net/tor-project/2015/
14:03:53 <karsten> "app-dev" is what they use as handle.
14:04:20 <karsten> qbi: does that help?
14:05:13 <karsten> (this was in response to your question from last time.)
14:05:16 * qbi looks.
14:06:12 <qbi> karsten: Yes, this does look helpful. Thanks.
14:06:22 <karsten> cool.
14:06:52 <karsten> are there any other questions I can help answer that have to do with other teams?
14:07:40 <karsten> otherwise I'd say let's move on to the next topic.
14:07:55 <karsten> virgil and Sean are not around, it seems.
14:08:04 <karsten> let's move that to the end in case they show up later.
14:08:13 <karsten> * Disagreement of directory authorities (Letty)
14:08:50 <Letty> http://letty.io/tor/
14:08:53 <karsten> we already talked this morning, and an angry person on the pad (linked from here) complains that we should learn american.
14:09:20 <Letty> sounds like a nice person..
14:09:28 <Letty> last time i used an aster plot and we thought about having a barchart as an additional visualization.
14:09:34 <Letty> i've implemented the barchart and karsten and i decided to just focus on that vis because its easier to understandsounds
14:09:51 <tomlurge> that looks nice!
14:09:57 <qbi> recived --> received
14:10:15 <Letty> :o thanks!
14:10:41 <tomlurge> gives a totally different impression about the ratio of disagreement than teh circular version from 2 weeks ago. i guess this one is more accurate?
14:10:49 * qbi still fails to understand the visualisation although it looks nice. ;)
14:10:50 <Letty> yes
14:11:04 <tomlurge> then it’s MUCH better
14:11:27 <karsten> so, I started writing some text to explain this visualization.
14:11:48 <karsten> qbi: ^ want to help me make that text better?
14:11:59 <qbi> Where does the text live?
14:12:13 <karsten> Letty: mind if I put it on the pad that americans don't understand?
14:12:39 <karsten> Letty: though pads are cheap, and I can create a new one.
14:12:52 <Letty> you can use the readme markdown file in my git repo: https://github.com/Letty/healthy_onion
14:13:22 <tomlurge> … will the chart become interactive?
14:13:35 <Letty> Karsten: i can also use my pad and share this. no one will complain
14:13:38 <karsten> ah, it could live there once it's ready. let me put it on a pad for now until that is the case.
14:14:10 <Letty> tomlurge: of course! i hate non interactive visualizations
14:14:33 <karsten> https://pad.riseup.net/p/tTXlrySWiliw <- draft text
14:14:45 <tomlurge> letty: cool :-)
14:15:06 <Letty> https://pad.dsst.io/p/lettysplayground < new pad
14:15:12 <karsten> ah
14:15:58 <karsten> now it's there, too.
14:17:09 <Letty> tomlurge: i want to make some mouseover effects, like showing the concrete numbers and we want to allow users to zoom in the graph
14:17:58 <tomlurge> good text. i feel informed. of course explaining the algorithm givescbonus points
14:20:07 <karsten> needs much more work.
14:20:10 <karsten> but it's a start.
14:21:00 <karsten> okay, I think we both have a few action items.
14:21:05 <karsten> Letty: ^
14:21:11 <Letty> yes
14:21:14 <tomlurge> letty: the numbers are static I guess. in that case I’d rather put them on a scale at the bottom. a mouseover, to be usefull, could link to a list of those relays (although I bet you would first have to aggregate that. probably not worth the effort). but I was more hoping for a visualization of these bars over time. how the ratios change
14:22:38 <karsten> so many requests in one line. ;)
14:23:14 <karsten> let me answer the one about a list of relays:
14:23:15 <Letty> tomlurge: bars over time could be hard to read, because you will have a bar for every flag. maybe a line for every flag over time is easier? (different colors for different flags)
14:23:37 <karsten> what you see on that graph are a few thousand relays. a list of them would be ... long.
14:24:32 <tomlurge> karsten: yep. maybe just the rotten ones… but as already said: just an idea, probably not very useful
14:24:56 <tomlurge> letty: i magine the bars pumping as time goes by, leaving marks of maxima
14:25:55 <Letty> tomlurge: ahh! now i understand
14:27:53 <karsten> tomlurge: interesting idea about that list. one issue is that it would require much more data.
14:28:21 <karsten> tomlurge: but I see the desire to learn more about problematic cases.
14:28:36 <tomlurge> exactly
14:29:19 <karsten> https://consensus-health.torproject.org/consensus-health.html
14:29:24 <karsten> takes long to load ^
14:29:35 <karsten> in particular, https://consensus-health.torproject.org/consensus-health.html#relayflags
14:29:50 <karsten> there are all the details.
14:30:13 <tomlurge> takes long to load indeed
14:30:20 <tomlurge> omg
14:30:21 <karsten> yep.
14:30:30 <Letty> :D
14:30:49 <tomlurge> well, “ich ziehe meinen antrag zurück”
14:31:05 * tomlurge has to learn english
14:31:13 <karsten> well, it may be possible to include some details in the graph,
14:31:23 <karsten> but I'd say let's save that for version 2. ;)
14:31:43 <tomlurge> “version 2” … I heard that before ;-)
14:31:59 <tomlurge> but you’re right!
14:31:59 <karsten> okay, should we move on to the next topic?
14:32:30 <Letty> yes
14:32:52 <karsten> * Analytics server (tomlurge)
14:33:30 <tomlurge> yep. well, JSON converters seem in good shape now, although they gave me out of memory troubles on a first test run
14:33:44 <karsten> -Xmx4g ?
14:33:57 <tomlurge> which made me procrastinate there and switch to my new hobby: Avro serialization
14:34:24 <tomlurge> which is very cool and soon in the state to give me … memory troubles too. i hope
14:35:00 <tomlurge> that Avro thing is necessary to do more than just SQL on the data (to answer Karstens question right away)
14:35:09 <karsten> thanks. :)
14:35:42 <karsten> so, regarding converters,
14:35:43 <tomlurge> i heard of that Xm thing. good news here: i got a friend to help me alittle with java stuff over the weekend
14:35:56 <karsten> I assume they'll take a while to run.
14:36:07 <tomlurge> so, all in all, i’m vaguely optimistic
14:36:21 <karsten> what you can also do is set an option in metrics-lib.
14:36:23 <karsten> let me look...
14:36:41 <karsten> setMaxDescriptorFilesInQueue()
14:36:45 <tomlurge> karsten: yes, but i was positively surprised with the MongoDB aggregation, so I decided not to be too nervous…
14:36:50 <karsten> DescriptorReader.setMaxDescriptorFilesInQueue()
14:37:03 <karsten> just set that to 5
14:37:15 <karsten> it's currently at 100.
14:37:24 <karsten> well, by default at 100.
14:37:38 <karsten> do that before you call readDescriptors().
14:37:43 <tomlurge> ah, good hint! okay. seems like i should have asked you right away. didn’t want to bother you with rookie questions
14:38:05 <karsten> and give it more memory anyway, because why not.
14:38:44 <tomlurge> right. after that’s done it will all be smooths sailing (as w.white used to say) i hope(…)
14:39:50 <tomlurge> those metrics lib bugs though… is tordnsel important?
14:40:08 <karsten> is there a fix for that yet?
14:40:31 <tomlurge> no, but i wonder about he consequences of just commenting it out for the time being
14:40:47 <karsten> it's fine.
14:40:55 <tomlurge> will that bite me when i try to do interesting things?
14:41:14 <tomlurge> okay. v2 then
14:41:23 <karsten> yep, v2.
14:41:38 <tomlurge> cool. that’s all from my side
14:41:45 <karsten> ok. great!
14:42:23 <karsten> I think we ran out of topics then.
14:42:47 <karsten> until anything else comes up?
14:42:51 <karsten> unless*
14:42:58 * karsten should learn american.
14:43:42 <Letty> :D i'm wondering what language american is.. american spanish? amercian english? american polish?
14:44:13 <karsten> oh, that's a fine question. I forgot to ask.
14:44:18 <tomlurge> anything with lot’s of chewing gum in you mouth
14:44:34 <karsten> hehe
14:44:39 <karsten> okay, let's end this meeting early then.
14:44:47 <karsten> in case others want to talk about tor dev.
14:44:53 <karsten> thanks, everyone!
14:44:58 <karsten> #endmeeting