14:29:39 #startmeeting metrics team 14:29:40 Meeting started Thu Aug 10 14:29:39 2017 UTC. The chair is karsten. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:29:40 Useful Commands: #action #agreed #help #info #idea #link #topic. 14:29:42 hello. 14:30:33 hi there! 14:30:44 hi iwakeh! 14:30:50 trac seems down. 14:30:55 yep. 14:31:01 being worked on. 14:31:16 yes, the error messages vary :-) 14:31:17 https://storm.torproject.org/shared/Ou-1QRctynWbF4yedi-MfDsjImFMFSIEP20fbVGCPRa <- agenda pad 14:31:22 heh 14:31:36 okay, shall we go through the agenda items? 14:31:42 fine 14:31:45 * Onionoo backup host (iwakeh): running fine (see stats below) 14:31:49 yay! 14:32:02 yes, it helped to use more memory ;-) 14:32:07 how much? 14:32:12 just 4g 14:32:15 ah ok. 14:32:20 cool! 14:32:20 but the host has 12g 14:32:28 did you compare results? 14:32:38 like, does it diverge from omeiense? 14:32:41 not yet 14:32:51 okay, that's one possible next step. 14:33:01 that's part of the 'next steps' question. 14:33:07 right. 14:33:13 but I#M reluctant 14:33:15 it's not accessible from outside, right? 14:33:34 to run the web b/c it is accessible from outside. 14:33:52 hmm? 14:34:02 Which of course makes it more convenient to 14:34:14 retrieve the data and shut-down the service again. 14:34:59 I'm unclear what the issue is. 14:35:16 There is no firewall. 14:35:20 I assume. 14:35:32 no redirection etc. 14:35:39 did you try? 14:35:43 yes, 14:35:53 I ran the web and accessed it locally 14:36:04 and just for fun I tried the 14:36:16 host:8080 and got the pages. 14:36:19 ok. 14:36:24 well, not the end of the world. 14:36:40 and because of that I shut down the web immediately. 14:36:56 btw, on omeiense we're running the web udner a different user. 14:37:04 I noticed :-) 14:37:07 onionoo-unpriv 14:37:15 yes, good idea. 14:37:36 okay, do you want to discuss those things (firewall, redirection, etc.) with the admins? 14:37:43 (different user, ...) 14:37:55 same setting as omeiense? 14:38:05 that would work, yes. 14:38:48 ok. 14:38:59 next step after those two: include in rotation. 14:39:12 ok. 14:39:29 do you want to do the configuration stuff before comparing results? 14:39:53 hmm, and when results are different? 14:40:03 well, and before including it in the rotation. 14:40:13 though, you could also listen on localhost only. 14:40:20 I'd like to know it first, but 14:40:36 it could be done parallel as I need to mail with the admins. 14:40:48 I guess. 14:40:50 right. and if you only open the port on localhost, you should be fine. 14:41:06 yes, I can set that. 14:41:14 ok. 14:41:40 that are some good next steps, I think. 14:41:45 true. 14:41:49 moving on? 14:41:56 yes. 14:42:01 * metrics-lib internal&LogDescriptor (iwakeh): #22983 14:42:12 ready for review. 14:42:15 is that the ticket in needs_review from today? 14:42:30 correct and #22428 14:42:36 is to follow soon. 14:42:39 ok! 14:42:59 I'll look tomorrow. 14:43:00 there are quite a few code 'reductions' 14:43:11 possible with the new approach. 14:43:18 okay. :) 14:43:59 let's talk more after the review? 14:44:04 sure. 14:44:11 great! 14:44:18 moving on: 14:44:19 * Check logs of TB downloads (iwakeh) 14:44:47 still wondering about the keep salt in-memory with subsequent 14:44:55 script runs. 14:45:05 seems quite tricky. 14:45:25 that's why I suggested the logio module. 14:45:38 which could answer our question, too. 14:45:38 which doesn't hash IPs. 14:45:42 true. 14:45:47 different approach. 14:46:01 yes, but with less answers. I intend 14:46:17 to just drop a mail to the admins and maybe they have a solution. 14:46:23 sounds good! 14:46:35 after all a salt removed and hidden from us should work too. 14:47:00 maybe. 14:47:15 well, we all trust root there, don't we. 14:47:50 it's the part where non-hashed IPs hit the disk that I'm worried about. 14:48:06 oh, they shouldn't hit the drive unhashed. 14:48:16 only the salt is stored during the process. 14:48:35 and then erased. 14:48:50 and read from disk before writing a log line? 14:49:12 hmm, yes. 14:49:23 sounds okay. 14:49:48 fine. 14:49:52 alright. asking the admins sounds like a fine next step. 14:49:56 next? 14:50:02 yep. 14:50:08 * putting ExoneraTor on Tor Metrics (karsten) 14:50:14 so, I worked a bit on exonerator in the past days. 14:50:30 is there a ticket? 14:50:46 and I have a version with a servlet that produces a JSON file with results and the original servlet using that JSON file rather than querying the database directly. 14:50:54 there's no ticket yet. I didn't know how far I'd get. 14:51:01 but now I have something running here. 14:51:01 ok 14:51:16 fine, a Metrics sub-page. 14:51:24 the question is: should we put ExoneraTor on Tor Metrics, say, under https://metrics.torproject.org/exonerator.html? 14:51:43 and keep https://exonerator.torproject.org/ running for now, just serving the data? 14:51:52 hmm, 14:52:00 maybe a sub-tab? 14:52:13 It's sort of like the graphs. 14:52:19 ah. 14:52:30 I enter a configuration/data and receive a result. 14:52:41 regarding the place on Tor Metrics, 14:52:50 I thought about putting it under Home >> Operation >> ExoneraTor. 14:53:15 that page would pretty much be what's currently on https://exonerator.torproject.org/, with different header. 14:53:25 and it would show results. 14:53:36 we'd just keep exonerator.tp.o running for the JSON data. 14:53:47 so, Tor Metrics would query that host and get the JSON object with results. 14:53:55 is this the same model as for collector then? 14:54:00 hi irl! 14:54:04 hi! 14:54:04 pretty much, yes. 14:54:10 well, no 14:54:28 not exactly, as the user gives an ip and receives a result. 14:55:00 i mean, in the split of backend vs. frontend it's similar, vaugely in a handwave way 14:55:11 yes. 14:55:16 that's also what I meant. 14:55:26 and a bit like onionoo. 14:55:32 ah, I was looking from the user perspective, usually 14:55:33 where the specification is now on Tor Metrics. 14:55:44 and the service is still running on onionoo.tp.o. 14:55:50 we're 'hiding' the more tachnical data away. 14:56:23 but, for a start we could just use the onionoo/collector approach. 14:56:38 what would be the step after that? 14:56:42 (when you say for a start) 14:57:11 see, if complaints come in and move 14:57:20 it to some more prominent place. 14:57:45 you mean easier to find on Tor Metrics? 14:57:52 on the metrics website, it's more discoverable 14:57:57 yes, easier to find for non-techs. 14:58:03 it's already linked from the operations page, it's just where the link goes that will change 14:58:08 old location would redirect anyway 14:58:14 true. 14:58:14 true. 14:58:23 I mean, if we can make it even easier to discover, let's do that. 14:58:28 (: 14:58:52 right now, we focus mainly on graphs. 14:58:56 on the start page. 14:59:11 we could change that and also add direct links to services. 14:59:18 I think, it'll be fine with the links and redirection. 14:59:21 well, extend, not necessarily change. 14:59:26 as a fun thing, "this ip address was last seen as a tor relay on DD-MM-YYYY" link to exonerator in the footer of the metrics website, link to exonerator 14:59:41 maybe it'll always say "never" for most users, but it's an interesting thing people might click 15:00:00 hehe. and it'll totally kill the database to make that many requests. ;) 15:00:09 ok, maybe not then (: 15:00:13 not yet. 15:00:14 ;) 15:00:15 oh, another reason for refacturing ;-) 15:00:25 the db, that is. 15:00:28 yes! 15:00:32 I didn't touch that part. 15:00:42 could be a bigger project. 15:00:54 anyway. 15:00:58 but for the integration now the plan seems set 15:01:06 alright, I'll proceed with this plan. 15:01:14 and open a ticket once I have something to share. 15:01:20 :-) 15:01:50 speaking of, 15:02:01 should we consider doing something like this with atlas? 15:02:16 that is, move it to Home >> Operation >> Atlas ? 15:02:19 on my roadmap is updating the bootstrap and jquery things 15:02:28 we could retheme it to the metrics theme, i don't see why not 15:02:38 we have a ticket for this 15:02:47 cool! 15:02:50 I mean, we could even include it on the page. 15:02:59 and avoid updating jquery and bootstrap in 2 places. 15:03:11 not suggesting this as short-term thing. 15:03:17 but, looking at the next weeks or so. 15:03:23 yep, i get it, that is what i meant 15:03:35 instead of updating it, just go for integrating it with metrics website 15:03:45 which updates it as a side effect 15:03:51 yes. worth thinking about! 15:04:00 i'll update the ticket once trac is back 15:04:20 in theory, we could add atlas.git as submodule and let metrics-web pull it in when producing a .war file. 15:04:28 not sure what's easiest. 15:04:32 okay. 15:04:50 we can discuss this on the ticket 15:04:56 i like it though 15:05:07 cool! 15:05:10 last on the list: 15:05:13 * future of Compass (karsten) 15:05:25 we should decide what to do with compass. 15:05:32 I hear it's still being used. 15:05:38 is gsathya still around? 15:05:39 but it's not really maintained anymore. 15:05:44 I don't think so, no. 15:06:08 though I believe there's a fork of some sort. 15:06:10 i know people do still use it 15:06:22 we could take out the link? 15:06:22 created by nusenu who cares a lot about tor network health. 15:06:39 ah, or find a new active maintainer? 15:06:49 if we want to keep it, we might want to review those changes and merge it back. 15:06:55 or decide it's the new compass. 15:07:04 i could look at that, i can python 15:07:25 that would be neat! 15:07:36 maybe you can talk to nusenu about this, too. 15:07:55 I noticed this branch/fork when looking at a compass ticket. let me find that. 15:08:08 #10306 15:08:12 #10306: Show relays by nickname substring 15:08:20 (pretending to be the bot) 15:08:25 hehe 15:08:28 bleep bloop 15:08:38 (full url ;-) 15:08:45 okay, do you want to take a look when trac is back? 15:08:50 (sorry, I'm still learning..) 15:08:51 ok, i'll follow up on that 15:08:57 cool! 15:09:47 (: 15:09:48 alright. 15:09:53 out of items. 15:09:55 anything else? 15:10:00 yep :-) no. 15:10:01 one quick onionperf 15:10:02 (glad to see you back here, irl, by the way!) 15:10:08 yes? 15:10:15 i have restored my instance and i'm doing some debugging 15:10:35 i'm also looking at writing icinga plugin to make sure it doesn't disappear for extended periods again 15:10:42 heh 15:10:43 will follow up on onionperf deployment ticket if that's still going 15:11:09 sure. either the existing ticket or a new one. 15:11:36 i will also probably be along to the montreal meeting, need to sort out some admin things but should have an idea by the end of the week 15:12:05 i think that's all my things 15:12:12 be along means? 15:12:32 present in person on the days of the meeting, including the team meeting 15:12:43 ah, neat! 15:12:53 regarding dates, 15:13:09 I'll only be around for 3 days, and iwakeh won't make it at all. 15:13:21 see the trac page for details. 15:13:23 :-( 15:13:28 yes, :( 15:13:37 ah ok 15:13:51 still, maybe some remote participation can be organised 15:14:06 irc/skype/jitsi/whatever 15:14:09 yes! 15:14:13 yes. 15:14:32 okay. 15:14:39 are we done? 15:14:44 I think so. 15:14:46 (: 15:14:59 great. thanks, iwakeh and irl! 15:15:06 thanks and bye, bye :-) 15:15:08 more via trac once it's back. bye, bye! 15:15:19 #endmeeting