[09:08:38] Analytics-Engineering, Wikimedia-Hackathon-2015: Analytics-Engineering availability to support VE A/B test during hackathon - https://phabricator.wikimedia.org/T99014#1292919 (Qgil) Is this a hackathon activity or an activity that needs to be done despite the team being busy with the hackathon? The descri... [09:30:56] ssh shrek [09:31:00] oops :) [10:04:59] Analytics-Tech-community-metrics: Community Bonding evaluation for "Allowing contributors to update their own details in tech metrics" - https://phabricator.wikimedia.org/T98045#1292995 (Sarvesh.onlyme) [11:12:16] Analytics-Engineering, Wikimedia-Hackathon-2015: Analytics-Engineering availability to support VE A/B test during hackathon - https://phabricator.wikimedia.org/T99014#1293087 (Qgil) It is time to promote #Wikimedia-Hackathon-2015 activities in the [[ https://www.mediawiki.org/wiki/Wikimedia_Hackathon_2015... [11:55:56] Hello everyone, I just tried to upload a cohort of 5cswiki users and it is still pending after an hour. Anyone else encountering this problem? [11:56:14] (to wikimetrics) [13:36:00] Analytics-Cluster, Graphoid, Services, Varnish: Graphoid access logs are missing from Hadoop - https://phabricator.wikimedia.org/T99372#1293481 (Ottomata) Interesting. Is graphoid not served by one of the regular varnish clusters? That is: bits, misc, mobile, text, or upload. [13:38:05] morning! [13:39:05] Analytics-Cluster, Graphoid, Services, Varnish: Graphoid access logs are missing from Hadoop - https://phabricator.wikimedia.org/T99372#1293493 (mobrovac) >>! In T99372#1293481, @Ottomata wrote: > Interesting. Is graphoid not served by one of the regular varnish clusters? That is: bits, misc, mo... [13:39:09] good morning! [13:39:15] (to wikimetrics)§ :) [13:39:36] Copying my earlier message: Hello everyone, I just tried to upload a cohort of 5cswiki users and it is still pending after an hour. Anyone else encountering this problem? [13:40:32] milimetric: I read your e-mail in the mailing list, it says it should be OK by now, right... or do you know if anyone else is experiencing difficulties [13:40:34] ? [13:41:39] Hey Vojtech-Dostal, hang on, lemme make sure the timing makes sense here [13:41:54] so my message was from last Thursday, when were you having problems? [13:42:07] Analytics-Cluster, Graphoid, Services, Varnish: Graphoid access logs are missing from Hadoop - https://phabricator.wikimedia.org/T99372#1293507 (Yurik) @mobrovac - we log all external access to hadoop for analysis, and keep logs for about a month. This way we can have reasonable usage statistics, e... [13:42:21] yurik: https://en.wikipedia.org/wiki/List_of_most_expensive_paintings#Scatter_plot [13:42:25] :) [13:43:16] milimetric, https://en.wikipedia.org/wiki/User_talk:Primaler#Graphs :D [13:43:35] I know, who is this? I was just on his talk page [13:44:29] milimetric, i mean - i added a comment at the end of his page :) [13:44:31] no idea [13:44:38] but guess what - huwiki is huge! [13:44:40] I know, I saw your post, he seems great [13:44:45] they have ~3000 graphs [13:44:49] woa!! [13:44:51] automated! [13:44:58] on each city page [13:45:08] no idea where they pull data, but the graphs are generated with lua :) [13:45:40] yeah, I was thinking the same thing you told Primaler, that he should factor out the data [13:45:47] I think that's the next barrier here [13:46:04] is there a way to make Lua parse tables in the same article? [13:46:05] yep, convert existing data into json [13:46:15] there should be, right? [13:46:46] milimetric, even though it is possible to load wiki page into lua and parse it and extract table, overall its a ... bleh ) [13:47:14] i think a better path would be to come up with a wiki-table -> json conversion util [13:47:23] that can be written in lua [13:47:46] and once json is generated, it should be placed into a separate page [13:48:03] and lua will convert it into table and into a graph [13:48:15] you know, I guess this is just the same old datasets problem [13:48:24] datasets? [13:48:31] yeah, we brought it up with Wikidata [13:48:43] and they said they're focusing on factoids at first [13:48:49] but wanted to do "tables" or "datasets" later [13:49:01] I think through queries [13:49:12] yeah, but i'm not sure they will want it arbitrary [13:49:28] well, we seem to be attacking this dataset problem from too many angles if you ask me [13:49:44] we should just make one basic arbitrary dimension dataset storage and retrieval wiki [13:49:46] like in that list of paintings, the data is very random - can be expressed with tons of different ways [13:49:50] and then put layers on top of that [13:49:55] something like a blob storage? [13:49:58] json [13:50:10] well, I like yaml over json, but yeah [13:50:20] yaml/json - representation [13:50:25] yaml would be much more human readable and just as good for machines [13:50:40] although if yaml, we could allow comments [13:50:44] i'd estimate yaml to be 2x more likely to be adopted by a regular person [13:50:49] true [13:50:51] agree [13:51:07] i guess to use it we would need yaml content handler [13:51:20] and for now keep it on wiki [13:51:20] yes :) but that shouldn't be hard. [13:51:26] yeah, shouldn't be [13:51:32] yeah, that's what bothers me, keeping it on wiki [13:51:49] we have the same exact problem with datasets now that we had with plain data factoids before wikidata started [13:51:53] well, i strongly believe in reusing tech - we should simply allow cross-site templates [13:52:11] once that works, the data will simply reside on commons [13:52:15] cross-site templates would definitely stop the bleeding, I agreed with you when you said that in the demo [13:52:16] and so will some scripts [13:52:30] yeah, and that'll be a good first step towards a proper dataset solution [13:52:44] wikidatasets.org :) [13:53:05] meh, enough with the domains, commons or source is good enough :) [13:53:33] now if only i can find volonteers to work on all that :) [13:53:34] i think it would need a really different user experience from what we have now, but I guess extensions could do that [13:53:45] yeah, or if we could convince someone at WMF ;) [13:54:11] well, content handler is easy [13:54:16] its the cross-wiki templ [13:54:19] thats the hard prob [13:54:30] i think roan was trying to solve it at one ponit [13:54:31] point [13:54:56] yurik: https://gerrit.wikimedia.org/r/#/c/211720/ [13:55:13] there will be a little more to do on the analytics side to get the data into the tables, but this is the start [13:56:12] ottomata, does it have to be called parsoid? [13:56:24] because other services like mathoid etc are using it [13:56:27] it has to have a name, [13:56:28] hm. [13:56:37] i just created the topic in kafka. oops, should have waited [13:56:46] no, but it sounds like the puppet stuff is poorly named then too [13:56:49] what should it be called? [13:56:50] -oid? [13:56:52] webrequest_oid? [13:56:53] haha [13:56:55] services? :) [13:56:59] webrequest_services [13:57:01] that would be fine [13:57:08] oids is good too :) [13:57:12] shorter :) [13:57:15] ottomata, ^ [13:57:23] webrequest_oids :D [13:57:25] i like [13:57:31] yeahHHhhH, but probably more services will use this, right? and they won't all be called -oid? [13:57:52] not necessarily, but its a good indicator overall :D [13:58:00] services is too long:D [13:58:07] esp with webrequest_ [13:58:15] webrequestoid [13:58:38] heh, naw not too long. [13:58:45] long == descriptive and not confusing :) [13:58:46] ottomata, up to you :) [13:58:49] i don't mind _oid, though [13:59:02] oki ) [13:59:04] actually, let me talk to bblack first, i think he had mentioned somehting about wanting to change this setup anyway [13:59:16] bah, i shoudln't have been so hasty [13:59:18] i created the topic in kakfa [13:59:23] and i can't delete them! :( [13:59:25] oh well. [14:01:01] tis ok [14:02:21] Vojtech-Dostal: there was a problem, I had to restart the queue. It was separate from last Thursday. Please refresh your cohort page and see if it was validated ok and file a bug in phabricator if it wasn't. Thank you for the report! [14:03:13] milimetric: Wonderful! :) Thank you very much [14:03:16] it works now [14:03:54] ottomata, what will happen if some service is hosted separatelly? [14:05:46] yurik: eh? [14:05:48] whatcha mean? [14:06:09] ottomata, if you call it webrequest_services, and we build another service that goes onto a separate varnish cluster [14:06:28] will we want to create a new branch, or will we simply add a new one like webrequest_maps [14:06:42] well, we could still make the other cluster use the same topic name [14:06:45] the topic name is our choice [14:06:47] gotcha [14:07:07] but, now that you mention this, i'm not so sure I want to add a lot of new topics for this [14:07:09] but you can't split one cluster into multiple topics? [14:07:25] the more of those we have, the more painful it is to deal with jobs in the cluster, because of the way oozie organizes some things [14:07:36] i'm sure gabriel will want all parsoid stuff separate [14:07:38] yurik: sorta we could, but we'd have to run multiple varnish instances [14:07:42] sorry [14:07:44] varnishkafka instnaces [14:07:50] i see [14:07:51] and match on requests somehow [14:07:53] like [14:08:04] -m 'uri_path == blabla' [14:08:09] or however varnish likes that [14:08:35] yurik: in general, we need a way for apps to log to kafka that is not part of a webrequest [14:08:39] right now, we only log all webrequests [14:08:49] but it woudl be much nicer if we could do eventlogging style logging from apps [14:08:54] we sooorta can at the moment [14:08:57] but not in seprate topics [14:09:36] but in general, i would like to see an easy way for teams to build apps that can log and then collect whatever they need [14:10:39] milimetric: iiintersting: http://docs.stratio.com/modules/sparkta/development/usingSparkta.html [14:11:23] * milimetric looking [14:12:05] ottomata, i thought mobrovac built some system to log stuff from nodejs services [14:12:11] not sure where that stuff goes at the moment [14:12:19] (i think its broken and doesn't go anywhere) [14:15:50] ottomata: yeah, this is the right model. I think this is superior to lambda architecture [14:16:10] this spark ecosystem seems to be maturing quickly though [14:16:28] pipeline db is thinking in similar terms but sparkta goes all the way to the viz layer [14:16:55] yeah, i'm not so sure how I feel about doing all of that via declarative configuration, but it could be super cool and remove a lot of error prone code [14:17:18] well, that code is getting written somehow [14:17:53] it's just that right now instead of declarative json or sql (in pipeline db's case) it's a bunch of strung together pieces that all break and cause me nightmares [14:18:26] yeah [14:18:48] i'm thinking more along the lines of this for new stuff, like pageview aggregates, or the mobile apps uniques, etc. [14:18:49] side note: why in the world does everyone keep using json. They're aware of YAML right?! :) [14:18:52] wondering how much we could use it for that [14:19:05] yeah, we should try it [14:19:16] haha, good question :) json i think is better for data transport, but for config stuff yaml is so much better [14:19:17] how much could we use it for that? A lot. All the much [14:19:43] right, agreed on data transport, but these are declarative definitions. Same with Druid [14:19:45] so weird [14:20:13] aye [14:20:13] yeah [14:24:44] Analytics-Cluster, Graphoid, Services, Patch-For-Review, Varnish: Graphoid access logs are missing from Hadoop - https://phabricator.wikimedia.org/T99372#1293592 (BBlack) >>! In T99372#1293507, @Yurik wrote: > Also, it seems parsoidcache is a bad name for generic services varnish cluster :) It... [14:37:32] Analytics-EventLogging, operations: Allow eventlogging ZeroMQ traffic to be consumed inside of the Analytics VLAN. - https://phabricator.wikimedia.org/T99246#1293621 (Ottomata) Thank you! confirmed. [14:38:21] ottomata: sparkta looks fun :) [14:39:21] yeah! [14:59:11] Analytics-Cluster, Graphoid, Services, Patch-For-Review, Varnish: Graphoid access logs are missing from Hadoop - https://phabricator.wikimedia.org/T99372#1293684 (Ottomata) Yuri, according to @gwicke and @bblack, most Parsoid requests should already be coming from the text varnish cluster, which... [15:01:07] Analytics-Cluster, Graphoid, Services, Patch-For-Review, Varnish: Graphoid access logs are missing from Hadoop - https://phabricator.wikimedia.org/T99372#1293689 (GWicke) @ottomata, @yurik is looking for graphoid requests, which aren't exposed via text yet. However, that will change soon with ht... [15:04:19] Analytics-Cluster, Graphoid, Services, Patch-For-Review, Varnish: Graphoid access logs are missing from Hadoop - https://phabricator.wikimedia.org/T99372#1293690 (Ottomata) Oh duuhhh. ok. Let's just wait for that then. [15:19:20] whoa joal, this might be simpler than rest proxy: http://hermes-pubsub.readthedocs.org/en/latest/contents/tutorials/01_getting_started.html [15:23:32] ottomata: no validation in hermes from what I see :( [15:24:02] aye [15:24:08] but maybe simpler to work with, no sure [15:24:17] yeah, for sure [15:24:21] or add meta.wm.org validation [15:24:23] who knows [15:24:28] but ja [15:24:50] On the other end I have managed to validate event in scala using https://github.com/fge/json-schema-validator [15:41:49] ottomata, so whats the verdict? can logstash be used for all the good logging? [15:41:54] for the services [15:42:38] yurik: logstash? [15:42:50] yurik: they will be moving graphoid into text cluster this week [15:42:58] ottomata, well, any kinds of logging really [15:43:03] to you should get those requests in webrequest_source = text [15:43:08] ? [15:43:15] yurik: not sure what you are asking [15:43:28] joal: cool! [15:43:36] ottomata, re service logging, such as errors in logstash [15:43:47] "all the good logging" -- yes. But don't go nuts logging every thing that ever changes. Operational logging in production should really be warnings and errors but with enough context to knwo why things broke [15:44:01] ottomata, or were you checking just on using hadoop for app-specific logging? [15:44:07] yurik: sure, if you are writing graphoid stuff and want operational error logging, logstash sounds good [15:44:24] yurik: the thing I was talking about, more generic app logging [15:44:27] would be mroe like eventlogging works [15:44:31] with defined schemas [15:44:43] used for collecting data and doing analysis [15:44:56] a/b testing, performance timing, funnel analysis, whatever [15:45:10] yeah, so that's something i might be interested in as well - do you think it will be possible to do this for the services, or the infrastructure is not in place yet? [15:47:09] yurik: infrastructure not in place [15:47:15] there is a lot of though around it [15:47:27] we are working on porting the eventlogging transport to kafka now [15:47:30] which will make somethings nicer [15:47:41] but, it will still use a limited number of topics [15:47:46] so your logs will be all mixedup with others [15:47:53] i would like folks to be able to do things like [15:48:08] consume --topic 'graphoid-timing' | do-whatever-you-want [15:48:10] or. [15:48:16] have all that data in hadoop automaitcally [15:48:18] like webrequset stuff [15:48:22] already divided up by schemas [15:48:34] it might be up to you to map a hive or whatever schema to it ,so you can query [15:48:40] but maybe we can automate that too, dunno [15:48:49] or, use that stuff in spark-streaming, or storm, or something [15:48:53] lots of stuff to think about [15:51:27] mforns: I merged your change, but I have meetings until my EL maintenance window. So do you want to deploy it together then? It's in roughly 3 hours [15:51:51] milimetric, sure! [15:52:04] milimetric, I will stay late today [16:07:09] Analytics-Cluster, Analytics-Kanban: Productionize reporting of app session metrics [8pts] {hawk} - https://phabricator.wikimedia.org/T97876#1293815 (madhuvishy) a:madhuvishy [16:56:29] joal: you in meeting? want to talk about and then maybe deploy those memory setting changes? [16:56:39] ottomata: in tasking yeah [16:56:42] hm, k [16:58:08] ottomata soooory :) [16:58:15] s'ok [16:58:21] we can do later today, or maybe tomorrow, eh? [17:19:06] Analytics-Wikimetrics, Community-Wikimetrics: Date selection form element is not visible on mobile - https://phabricator.wikimedia.org/T99539#1293958 (Capt_Swing) [17:25:49] ottomata, hi, is there reason i cannot connect to hue? [17:26:05] i used to work with it, but now i can't login [17:26:46] oh yes. sorry one sec [17:26:53] consequence of upgrade a few weeks ago [17:27:20] yurik: try now [17:27:42] ottomata, thx, works [17:29:09] Analytics-Engineering: Analytics-Engineering availability to support VE A/B test during hackathon - https://phabricator.wikimedia.org/T99014#1293987 (ggellerman) [17:31:11] Analytics-Engineering: Analytics-Engineering availability to support VE A/B test during hackathon - https://phabricator.wikimedia.org/T99014#1283579 (ggellerman) This task is mostly to raise awareness for the support needs of an A/B test that coincides with the Hackathon but is not related to the Hackathon.... [17:32:27] ottomata, sorry, another question - i know we spoke about it before: I have a "summary" table in hive, but the summary is too big to be processed in python too, so now i need to run a query on top of the summary table to generate the actual CSV text that i upload to the wiki. At the moment i do it from python. What's the best path for me to do it? [17:32:50] you can make csv from hive [17:32:54] woudl that do? [17:33:06] if that is complicated...you might be able to adapt your python to spark [17:33:09] ottomata, sure, as long as it is reliably can executed from python [17:33:13] ? [17:33:35] yurik: two suggestsion: make your hive query write out CSV data [17:33:38] or [17:33:41] e.g. at the moment i run hive query from python, and wait for it to complete. The .hql file runs and inserts results into the summary tabel [17:33:46] oh from python [17:33:56] via some jdbc connector? or just shelling out? [17:34:01] shelling out [17:34:04] ah [17:34:15] well, i mean, you can write hive queries that generate csvs [17:34:21] so you could do that all in the query [17:34:23] now i need to run another hive query, but this time getting csv [17:34:25] rather than post-processing with python [17:34:54] my existing process looks inside the actual tables because i use text table [17:34:56] via /mnt [17:36:40] ottomata, is this method still works: hive --database wmf_raw -e "query goes here" > file_name.tsv [17:36:46] yurik: does your python process do more aggegation or trasnformation of your summary table, or does it just convert to csv? [17:36:58] no , yurik, how about [17:37:14] -e "SeLECT ... INSERT INTO table FORMATTED AS CSV" [17:37:18] (pseudocode) [17:37:59] ottomata, https://github.com/wikimedia/analytics-zero-sms/blob/master/scripts/zero-counts3.hql [17:38:05] this is the script i run by shelling out [17:38:11] http://stackoverflow.com/questions/18129581/how-do-i-output-the-results-of-a-hiveql-query-to-csv [17:38:21] it creates TSV files basically [17:38:36] than i process the results by reading files directly via /mnt [17:38:41] and do additional processing [17:38:53] can you do the additional processing in hive rather than in python? [17:38:56] with the new query, too much data, so i would have to use alternative method [17:39:49] doubtful - I need to do additional processing [17:39:54] at least i did before [17:40:10] maybe for the new one i can use the resulting data directly and simply upload it [17:40:26] but it doesn't matter because i still have to load the result data into python and upload it to the wiki [17:53:30] yurik: what additional processing are you doing? [17:54:22] ottomata, things like normalizing totals per 30 days instead of per month, ignoring days with missing data if the gaps are shorter than N days, etc [17:55:36] yurik: try spark with python [17:55:51] you can probably read your summary table in it, and then use much of your existing python code [17:59:29] ottomata, is spark like pandas? [17:59:49] sorrrta, it has dataframe type stuff [17:59:58] but, it runs on the yarn cluster using hdfs [18:00:04] it's scala API is better than its python [18:00:09] but for what you are doing, the python one will probably be just fine [18:00:18] yurik: try it on stat1002 [18:00:20] run [18:00:22] pyspark [18:00:31] (that runs in local mode, not YARN cluster mode, but is faster for just prototyping) [18:00:44] yurik: where is a summary table file, and a link to some of your python code? [18:00:45] ottomata, i would much rather run a hive query on top of existing data, and it would have been great if the results could go directly into python, bypasing external storage (temp tables or files) [18:00:56] than the data is small enough that i could massage it myself [18:01:10] yurik: sure [18:01:11] that would be fine [18:01:13] ottomata, https://github.com/wikimedia/analytics-zero-sms/blob/master/scripts/zero-counts3.hql is the query [18:01:24] you could use a python jdbc connector probably [18:01:29] instead of shelling out to hive [18:01:32] i will look at jdbc [18:01:36] thx [18:01:37] ! [18:01:47] https://cwiki.apache.org/confluence/display/Hive/HiveClient#HiveClient-Python [18:01:48] the scripts are in the same repo [18:02:07] yurik: do you hav;e a summary table file in hdfs already? [18:02:09] i want to look at it [18:04:02] ottomata, sure - there are two - yurik.zero_webstats3 (that's the one that i already process in python directly by reading /mnt files). And the bigger one is yurik.countrycounts [18:04:26] that one is huge - per domain per request originating country per https per proxy [18:04:27] danke [18:04:40] i will look at the smaller one [18:05:04] the code is undergoing some rewrite atm [18:06:03] the existing code places stats into yurik.zero_webstats (older), and sums it up and uploads it to the zero portal using weblogs2.py [18:06:07] ottomata, ^ [18:06:48] also, yurik [18:06:48] fyi [18:06:48] https://spark.apache.org/docs/1.3.0/sql-programming-guide.html#hive-tables [18:08:32] you mean that will work right now without any extra work like installing weird libs? [18:08:43] right. [18:08:45] right now :) [18:11:05] also works pretty well on larger datasets [18:11:14] in yarn mode, it runs across the clsuter [18:11:15] just like hive [18:11:20] probably even will be faster [18:15:17] yurik: https://gist.github.com/ottomata/b71dcaedfc1eb93c8707 [18:18:39] ottomata, thanks!! will try [18:19:09] ottomata, the thing that sucks is that i can't test/dev it locally ( [18:19:26] i wonder if there is a way to set up intelij to remote debug python [18:19:47] over the tunnel [18:21:03] ottomata, update; got the install working [18:21:04] yurik: mw vagrant has spark? [18:21:07] extern "C" goddammit [18:21:15] general recommendation for people, actually [18:21:26] if you absolutely have to integrate with a C library try to avoid the ones written by Perl developers [18:21:35] ottomata, don't know, i could install everything local (i run ubuntu) [18:21:37] it's like a how-to of unreadable, eyeball-searing code [18:21:40] but i would rather tunnel remote [18:21:49] into stat, and debug remotely [18:21:51] much easier [18:31:41] yurik: you can probably install spark pretty easily [18:31:45] you can run it in local mode without hadoo [18:31:48] hadoop [18:32:06] there are .debs [18:32:15] you could add the wmf apt repo and even apt-get install it [18:36:08] milimetric: can I get a python brain bounce? [18:36:10] batcave? [18:37:17] ottomata: sorry in ameeting [18:37:20] and then deploying EL [18:38:45] ok [19:03:38] mforns: deploy time? [19:03:47] to the batcave [19:03:50] :] [19:04:09] ottomata: if you don't mind we'll talk / deploy memory tomorrow ? [19:05:01] time for me to get dinner, and spent some time with my son :) [19:05:18] Talk to you guys tomorrow! [19:05:41] yeah lets do toorrow [19:05:47] goodnight joal! [19:05:49] cool :) [19:05:53] Thx dude ! [19:06:08] BTW: Spark + parquet is really fast :) [19:06:15] ottomata -^ [19:08:36] awesoome [19:22:28] Analytics-Cluster, Graphoid, Services, Patch-For-Review, Varnish: Graphoid access logs are missing from Hadoop - https://phabricator.wikimedia.org/T99372#1294563 (akosiaris) As I already commented on the change, that cluster will be around for the foreseeable future as it serves as the entrypoin... [20:08:10] Analytics-EventLogging, Analytics-Kanban: Event Logging doesn't seem to handle unicode strings - https://phabricator.wikimedia.org/T99572#1294745 (Milimetric) NEW [20:08:31] anyone know of tools to get actual word count for (large numbers of) Wikipedia article revisions? [20:08:55] (as opposed to something just derived from character count) [20:10:40] halfak: ^ maybe knows something? [20:24:45] ragesoss, sounds like this could be done, but it'll take a bit of python [20:25:44] halfak: yeah, I don't think it's hard, just wondering if someone had already done it. I couldn't find anything, but LiAnna told me that Nimish had pulled word count numbers for her for the education program at some point. [20:26:03] ragesoss, I'm not aware of any nice way to do it from the DB. [20:26:18] But writhing a python script to scan the dumps shouldn't be too hard. [20:26:22] (Also, I think it could probably be done without python. ;) ) [20:26:49] Oh definitely. I just don't think you can do it without programming. [20:26:57] :) [20:27:05] halfak: *wave* [20:27:12] I hear you're also arriving in lyon late on friday [20:27:12] o/ valhallasw :) [20:27:20] Yup. that's right. [20:27:36] I'll be flying right from a conference in Florence. [20:27:47] I might explode with sleep when I arrive. [20:27:58] interested in sharing a ride? I'm arriving around 2200 with KL1419 [20:28:56] Yeah! I should land at 2230. Air France 7650 [20:30:19] \o/ [20:32:53] valhallasw, looking to meet up with prateek too. [20:32:59] We should be on the same flight [20:33:01] sounds good. [20:45:15] Analytics-Wikimetrics, Community-Wikimetrics: Date selection form element is not visible on mobile - https://phabricator.wikimedia.org/T99539#1294862 (Capt_Swing) Turns out, not a bug. From @milimetric: "There's something slightly different going on here. The reason you're not seeing dates is because th... [20:45:38] Analytics-Wikimetrics, Community-Wikimetrics: Date selection form element is not visible on mobile - https://phabricator.wikimedia.org/T99539#1294864 (Capt_Swing) Open>Invalid [20:47:12] nick mforns_brb [20:57:22] nick mforns [20:57:49] Hi, I've tried to get vagrant up and running to work on the wikimetrics website but once I do vagrant reload I get an "The SSH command responded with a non-zero exit status." error and the website isn't up on localhost:5000. [20:59:56] memeht: that sounds like your VirtualBox VM didn't start for one reason or another [21:00:41] bd808: It did start up, I could see the mediwiki site on 127:1.0.0/wiki/Main_Page [21:01:17] ah, so just some PUppet provisioning failing then perhaps [21:01:27] bd808: I have a log of the vagrant stdout, would that help? [21:01:46] memeht: yeah. paste it somewhere and give the link here. I'll take a look [21:01:50] bd808: Yes, I get the error on vagrant reload --provision [21:01:53] bd808: k [21:09:17] bd808: http://pastebin.com/Gp2P6VsD [21:13:58] memeht: looks like the default config for running alembic is not correct -- __tablename__ = db.config['ARCHIVE_TABLENAME']; KeyError: 'ARCHIVE_TABLENAME' [21:15:19] MediaWiki-Vagrant imports that Puppet module from operations/puppet.git; I've not worked on it [21:15:49] bd808: would you know of the correct config or should I hop on the dev channel? [21:45:07] memeht: In production it gets set here -- https://github.com/wikimedia/operations-puppet/blob/f173fb61a80490b5b97bb4152dbf673bf4151835/manifests/role/wikimetrics.pp#L232 [21:45:44] mediawiki-vagrant probably needs something similar added to the hiera config [21:46:26] hmmm --- https://github.com/wikimedia/mediawiki-vagrant/blob/107da09920ec412f98f0d4b2c4f9141a47fe9945/puppet/modules/role/manifests/wikimetrics.pp#L61 [21:52:38]