[06:58:30] 10Analytics, 10MediaWiki-API: Copy cached API requests from raw webrequests table to ApiAction - https://phabricator.wikimedia.org/T155478#2944303 (10Tgr) [07:09:22] 10Analytics, 10MediaWiki-API: Copy cached API requests from raw webrequests table to ApiAction - https://phabricator.wikimedia.org/T155478#2944334 (10Tgr) [08:06:30] morning! [08:07:03] One Clickhouse issue solved and another one pops up :) [08:07:20] but at least the linking part is now good [08:08:23] the main issue now is that building on Debian Sid, since the final clickhouse-server binary depends on GLIB-2.23, and we don't have it on Debian Jessie (2.19) [08:08:54] I thought that the "GLIBC_COMPATIBILITY" flag that Clickhouse offer would have helped [08:08:57] but not much [08:09:06] anyhow, following up with upstream :) [08:09:40] I can see another Oozie error but I'll wait a bit before taking any action, this is new for me and I don't want to cause any issue :) [08:18:34] 06Analytics-Kanban, 13Patch-For-Review, 15User-Elukey: Puppetize clickhouse - https://phabricator.wikimedia.org/T150343#2944414 (10elukey) Updates: upsteam [[ https://github.com/yandex/ClickHouse/commit/e2bc0c7471b1ca56242d159a046ca5e1b5cfd6a1 | released ]] a fix for the linking issue. This solves the proble... [08:21:15] elukey: if you build on sid, applications will pick up functions and symbols introduced after glibc 2.19, you'll need to build on jessie to avoid that [08:21:55] that's why proprietary software shipped as binary-only is usually built on really ancient distros (to support the smallest common denominator) [08:24:46] moritzm: o/ Hello! I wanted to follow up with you later on today :) [08:25:56] last time that I tried to build using jessie I didn't satisfy all the build deps, but I can try again and see [08:34:15] moritzm: main problem - clickhouse require at least gcc-5 to build :( [08:36:41] is it written in C++? [08:37:11] yess! [08:37:31] they offer two ways to build [08:38:22] 1) all-in-wonder binary with all the deps statically built, having to consider only GLIBC compatibility at load/runtime [08:38:55] 2) usual shared libs dependency, but less flexible for "older" systems [08:39:31] I am trying 1) but I keep incurring in weird build issues [08:39:53] you won't really get around the GCC requirement, there's a range of features from the C++11 standard which are only supported in GCC and later [08:40:48] fun fact is that we are running (Experimentally) clickhouse on druid hosts using their binaries (extracted directly from the debs that they provide) [08:41:06] and I don't see the same GLIBC dependency issue [08:41:15] but GCC itself is fairly lightweight itself (since it needs to be bootstrapped for other archs), so it should be possible to simply rebuild gcc-5 on jessie [08:41:55] they probably built on some distro which has an older glibc, but still GCC 5 [08:42:25] do the also need a current version of libstdc++ in their all-in-one-binary? [08:43:45] so from debian/control: cmake, gcc-6, g++-6, libicu-dev, libreadline-dev, libmysqlclient-dev, unixodbc-dev, libglib2.0-dev, libltdl-dev, libssl-dev [08:45:10] maybe they don't use the standard library, then [08:45:51] I'd try to build gcc-5 for jessie. then it should all work out I guess [08:46:30] would it need to be on our repos to be able to pdebuild (or similar) > [08:46:33] ? [08:47:38] 10Analytics-Tech-community-metrics, 06Developer-Relations: Measuring Time To First Code Change (TTFCC) - https://phabricator.wikimedia.org/T137201#2944450 (10Qgil) At this point, I would either push this proposal to #devwish17 to check what is the interest or decline the task. [08:49:14] we should avoid to add it to our repos, better build it in a jessie VM or chroot [08:50:21] ah yes I suspected that, just wanted to be sure :) [09:43:49] moooorning a-team!! [09:43:56] Hiiii a-team :) [09:44:38] o/ [09:44:41] elukey: sorry you were alone yesterday, I also checked to see if anyone jumped in to the standup, but I guess we were on recovery mode :) [09:44:48] :) [09:45:12] joal: if you restart the failed oozie job, can you tell me how so I'll know it too ? :) [09:45:48] elukey: Yessir, it's the exact same error we had last week [09:46:00] elukey: simple rerun through hue should hopefully do [09:47:44] ah nice [09:47:46] can I do it? [09:48:16] Pleasde elukey :) [09:48:18] of course it is enough, we don't need to change parameters like when we get validation errors [09:48:20] grrr [09:48:33] elukey: you have ti :) [09:49:51] Hi fdans, you went back home without trouble? [09:50:49] joal: I did! but I'm afraid I'm still combating jetlag [09:50:57] fdans: So am i ! [09:51:14] fdans: Woke up this night at 1am, spent 2hours reading ... [09:51:28] yeah I slept from 10am to 4pm on Sunday [09:51:32] HUGE mistake [09:51:43] !log restarted mediacounts-archive-wf-2017-01-16 [09:51:44] Logged the message at https://www.mediawiki.org/wiki/Analytics/Server_Admin_Log [09:52:10] indeed fdans :) [09:52:14] Thanks elukey [09:55:54] joal: is pageview data on hadoop considered "private" ? [09:55:59] yessir [09:56:17] ah ok, I suspected that, I am reviewing an access request [09:56:34] I still haven't found a good distinction between analytics-users and private [09:56:37] elukey: it contains enough info on some single lines to allow good fingerprinting, so it's private [09:57:44] yes now I remember all the attributes, it makes sense [09:57:50] thanks :) [09:58:29] joal: from the clickhouse point of view, slowly making progresses with upstream :) [09:58:56] now it should be a matter of having a Jessie VM or similar and build gcc5 or gcc6 manually [09:59:38] elukey: I backloged your discussion with moritzm, thanks a lot for making progress :) [10:02:36] joal: it is a very interesting task, I am learning tons of new things. TIL that -no-pie is a linker flag that avoids relocable code issues [10:02:49] the fix was simply to add it in the main CMake file [10:02:52] and that's it [10:03:16] (the overall deb that they make is still horrible but I'll try to follow up again with upsteam :P) [10:03:48] elukey: awesome :) [10:09:53] if possible, try to use PIE though. it's a useful security issue [10:10:47] Hi moritzm [10:10:58] moritzm: I hope you're suffering less jetlag than I do :) [10:11:22] moritzm: Do you recall my ask for opening from analytics vlan to new labsdb one? [10:11:42] our second son is seven weeks old, jetlag is a NOP to me :-) [10:11:53] riiiiight moritzm :) [10:12:04] I'll make the ferm change later the day, currentl need to catch up with some updates [10:12:04] moritzm: congrats, and good luck ;) [10:24:55] moritzm: oh yes I am absolutely in favor of using shared libs, if the problem will turn up to be only having gcc-5 or 6 built manually I'll definitely do it [10:28:41] sounds good, GCC as used in Debian stretch defaults to building PIE executables anyway [10:28:51] it's a Debian-specific patch [10:34:35] 10Analytics, 10ChangeProp, 10Citoid, 10ContentTranslation-CXserver, and 11 others: Node 6 upgrade planning - https://phabricator.wikimedia.org/T149331#2944616 (10hashar) [10:36:56] moritzm: one of the main issues that I found with the clickhouse build was that they were trying to statically compile multiple libs with PIE, ending up in linker errors on debian [10:37:14] so they added the -no-pie when using the statically compile all option [10:41:16] that makes sense for this use case [10:48:55] 10Analytics, 10ChangeProp, 10Citoid, 10ContentTranslation-CXserver, and 11 others: Node 6 upgrade planning - https://phabricator.wikimedia.org/T149331#2944635 (10hashar) The CI Jessie instances are now using NodeJS version 6 as provided by apt.wikimedia.org. I kind of freaked out yesterday until I saw thi... [10:50:46] 06Analytics-Kanban, 06Operations: Open temporary access from analytics vlan to new-labsdb one - https://phabricator.wikimedia.org/T155487#2944637 (10JAllemandou) [10:59:11] 06Analytics-Kanban, 06Operations, 10netops: Open temporary access from analytics vlan to new-labsdb one - https://phabricator.wikimedia.org/T155487#2944653 (10elukey) p:05Triage>03Low a:05MoritzMuehlenhoff>03None [11:00:07] joal: can you update --^ with more details about the project etc.. ? [11:08:11] 06Analytics-Kanban, 06Operations, 10netops: Open temporary access from analytics vlan to new-labsdb one - https://phabricator.wikimedia.org/T155487#2944677 (10JAllemandou) [11:08:17] elukey: --^ enough? [11:11:48] joal: much better thanks :) [11:16:58] !log Remove mediawiki-history-beta datasource from druid [11:16:59] Logged the message at https://www.mediawiki.org/wiki/Analytics/Server_Admin_Log [11:20:05] 10Analytics, 10Analytics-Cluster: Review Druid's logging configuration - https://phabricator.wikimedia.org/T155491#2944710 (10elukey) [11:22:01] (03PS1) 10Joal: Update pageview druid job to overwrite files [analytics/refinery] - 10https://gerrit.wikimedia.org/r/332446 [11:23:57] * elukey lunch! [11:24:02] (03PS5) 10Joal: Add oozie job loading MW history in druid [analytics/refinery] - 10https://gerrit.wikimedia.org/r/328154 (https://phabricator.wikimedia.org/T141473) [11:24:09] enjoy elukey :) [14:19:45] * joal takes a break - see you at standup a-team [14:20:13] morning a-team [14:22:16] o/ [14:37:20] (03CR) 10Mforns: [C: 032] "LGTM!" [analytics/refinery] - 10https://gerrit.wikimedia.org/r/332446 (owner: 10Joal) [14:43:06] hi team :] [14:46:34] o/ [15:04:36] weirdness of the day: analytics1002 does not pubblish resource manager's graphite metrics [15:13:37] ah found it [15:17:12] joal: forgot to mention - this week we might need to restart the whole cluster for jvm updates :) [15:20:01] probably later, the hadoop cluster is still on trusty, it usually takes Ubuntu quite a while to issue updates [15:21:12] I really hope that during the next two quarters we'll migrate the whole cluster to jessie :D [15:45:26] Hi milimetric [15:45:26] milimetric: I have newly computed metrics ready to be uploaded for dashiki in stat1004:/home/joal/code/metric_result [15:45:26] milimetric: Those have been renamed and header-updated :) [15:57:06] (03CR) 10Nuria: [V: 032] Update pageview druid job to overwrite files [analytics/refinery] - 10https://gerrit.wikimedia.org/r/332446 (owner: 10Joal) [15:57:28] joal: cool, I'll sync them over [15:57:36] Thanks milimetric :) [16:00:01] need to reboot - problems with hangouts [16:01:55] milimetric: standdduppp [16:06:24] 10Analytics: Measure Community Backlog. - https://phabricator.wikimedia.org/T155497#2945157 (10Nuria) [16:39:55] 10Analytics: POC of clickhouse + denormalize edit data on labs that can be used as a backend for quarry - https://phabricator.wikimedia.org/T152788#2945357 (10Nuria) [16:50:51] 10Analytics: POC of clickhouse + denormalize edit data on labs that can be used as a backend for quarry - https://phabricator.wikimedia.org/T152788#2945411 (10chasemp) I believe we discussed (for now) extracting data from the (new) labsdb hosts (1009-1011) instead of `sanitarium` (which would be a different pros... [16:59:38] 10Analytics, 10Analytics-Cluster: Review Druid's logging configuration - https://phabricator.wikimedia.org/T155491#2945457 (10Nuria) To be clear indexing logs are present but just in one of the three machines. Pulling would need to be set up for application logs and indexing logs, specially application logs as... [17:10:47] 10Analytics: POC of clickhouse + denormalize edit data on labs that can be used as a backend for quarry - https://phabricator.wikimedia.org/T152788#2945470 (10Nuria) [17:21:09] joal: just found http://kylin.apache.org/ [17:37:48] joal: after taking an hour to remember how to use rsync, I got the files up :) https://analytics.wikimedia.org/dashboards/standard-metrics/ [17:37:56] thanks for the update! [17:40:29] 10Analytics: Meta-statistics on MediaWiki history reconstruction process - https://phabricator.wikimedia.org/T155507#2945580 (10mforns) [17:49:07] 10Analytics, 10ChangeProp, 10Citoid, 10ContentTranslation-CXserver, and 11 others: Node 6 upgrade planning - https://phabricator.wikimedia.org/T149331#2945609 (10mobrovac) a:03mobrovac [17:50:05] Thanks a lot milimetric :) [17:50:52] nuria: Would you have a minute before goals meeting? [17:51:20] joal: ay, probably not, but plenty time after [17:51:26] k [17:54:09] joal:goals meeting is to plan for this quarter, if you want to talk more long term we can do so later/if there is time/ [17:54:31] was not really about long term :) [17:58:21] [17:58:58] 06Analytics-Kanban, 06Discovery, 06Discovery-Analysis (Current work), 03Interactive-Sprint, 13Patch-For-Review: Add Maps tile usage counts as a Data Cube in Pivot - https://phabricator.wikimedia.org/T151832#2945642 (10Milimetric) cc-ing Mikhail, whose username is @mpopov [17:59:54] always batcave... [17:59:56] ? [18:02:42] 10Analytics, 10MediaWiki-API, 13Patch-For-Review: Run ETL for wmf_raw.ActionApi into wmf.action_* aggregate tables - https://phabricator.wikimedia.org/T137321#2945660 (10Milimetric) @Tgr I'm aiming to review this by the end of the week, please ping me if I slip up. [18:16:04] 10Analytics: Pull data for edit reconstruction from labs. - https://phabricator.wikimedia.org/T152788#2945705 (10Nuria) [18:35:58] 10Analytics: Hadoop cluster expansion.Add Nodes - https://phabricator.wikimedia.org/T152713#2945797 (10Nuria) We should upgrade cluster and build the new nodes in debian [18:37:14] 10Analytics, 06Editing-Analysis: Move contents of ee-dashboards to edit-analysis.wmflabs.org - https://phabricator.wikimedia.org/T135174#2290688 (10Jdforrester-WMF) p:05Triage>03Normal [18:37:51] 10Analytics: Hadoop cluster expansion.Add Nodes - https://phabricator.wikimedia.org/T152713#2945823 (10Ottomata) [18:39:35] 06Analytics-Kanban, 06Discovery, 06Discovery-Analysis (Current work), 03Interactive-Sprint, 13Patch-For-Review: Add Maps tile usage counts as a Data Cube in Pivot - https://phabricator.wikimedia.org/T151832#2945827 (10Yurik) Looks awesome! Could we add the more recent data to it too? [18:52:41] 06Analytics-Kanban, 06Discovery, 06Discovery-Analysis (Current work), 03Interactive-Sprint, 13Patch-For-Review: Add Maps tile usage counts as a Data Cube in Pivot - https://phabricator.wikimedia.org/T151832#2945911 (10Nuria) @yurik: Data comes from the source table @mpopov created. Once he updates it I c... [18:57:02] 10Analytics, 06Fundraising-Analysis: Build small pipeline of fundraising banner data into druid so it can be visually analyzed - https://phabricator.wikimedia.org/T155519#2945972 (10Nuria) [18:57:18] 10Analytics, 06Analytics-Kanban: Build small pipeline of fundraising banner data into druid so it can be visually analyzed - https://phabricator.wikimedia.org/T155519#2945985 (10Nuria) [18:58:07] ottomata: and .. what about the replacement of 1001? is that blocked on something? [18:58:32] 1001? is done? oh its in ops hands now [18:58:39] they just need to physically remove it [18:59:05] Gone for tonight a-team :) [18:59:51] laters! [19:01:14] going afk too! [19:09:22] nuria: just making sure you saw this: https://phabricator.wikimedia.org/T131280#2929889, I think we should make sure to look at it next quarter [19:10:03] milimetric: i also talked to kati love about similar [19:10:12] cool [19:27:59] 10Analytics: Making geowiki data public - https://phabricator.wikimedia.org/T131280#2946141 (10Nuria) Note to self: this is the most similar report publicly available now: https://stats.wikimedia.org/wikimedia/squids/SquidReportPageEditsPerCountryOverview2014Q4.htm [20:29:06] Hi nuria milimetric joal mforns ottomata!! Hope everyone had a good trip home, if u were traveling :) I'm here now to bother you with another question....!!! It is: what are the details of the permissions for access to pivot.wikimedia.org? I can get in with my Wikitech ID... Does that mean that anyone who registers on wikitech gets access? I imagine not, that rather there's another step in the [20:29:08] process... If there's somehwere in a codebase that I could just check myself, also, I'm happy to do so... Thx!!!!! [20:33:21] AndyRussG: you need a wikitech login that is in the "wmf" or "nda" LDAP groups [20:33:39] AndyRussG: if people need access to it, the process is to follow the steps in this task: [20:33:47] AndyRussG: https://phabricator.wikimedia.org/T150790 [20:36:27] milimetric: ok fantastic!! Yeah that sounds just right :) thx! [20:39:51] 10Analytics, 10Fundraising-Backlog, 13Patch-For-Review: Productionize banner impressions druid/pivot dataset - https://phabricator.wikimedia.org/T155141#2946531 (10AndyRussG) I successfully uploaded data for all of December. Looks good!! I think it'd be quite easy to productionize this so the data accrues d... [21:10:00] AndyRussG|away: you need an NDA [21:10:33] AndyRussG|away: if you are a wmf employee you already have signed one so just file a ticket for access and tag it "ops access request" [21:10:56] AndyRussG|away: ah sorry, mi irc client trick me milimetric had alredy responded [21:15:06] AndyRussG|away, milimetric : updated docs on wikitech with this info: https://wikitech.wikimedia.org/wiki/Analytics/Cluster/Druid#Pivot [21:15:31] good call [21:20:28] 06Analytics-Kanban, 06Discovery, 06Discovery-Analysis (Current work), 03Interactive-Sprint, 13Patch-For-Review: Add Maps tile usage counts as a Data Cube in Pivot - https://phabricator.wikimedia.org/T151832#2946848 (10debt) Let's go ahead update the data for Dec 2016, @mpopov and then we can look at sett... [21:21:45] 06Analytics-Kanban, 06Discovery, 06Discovery-Analysis (Current work), 03Interactive-Sprint, 13Patch-For-Review: Add Maps tile usage counts as a Data Cube in Pivot - https://phabricator.wikimedia.org/T151832#2946854 (10Yurik) we could probably even have half of january as well :) [21:36:05] 06Analytics-Kanban, 06Discovery, 06Discovery-Analysis (Current work), 03Interactive-Sprint, 13Patch-For-Review: Add Maps tile usage counts as a Data Cube in Pivot - https://phabricator.wikimedia.org/T151832#2946933 (10mpopov) Okie dokie, thank you @Nuria! I will see if I can replicate your Druid-loading... [21:36:49] bearloga: will be here for a bit, let me know if you need help [21:43:53] nuria: thanks! :) [22:09:04] 10Analytics, 10Fundraising-Backlog, 13Patch-For-Review: Productionize banner impressions druid/pivot dataset - https://phabricator.wikimedia.org/T155141#2947156 (10ggellerman) p:05Triage>03Normal [23:55:13] 10Analytics, 10EventBus, 06Services (doing): EventBus produces non-canonical page urls - https://phabricator.wikimedia.org/T155066#2947714 (10Pchelolo) a:03Pchelolo > The EventBus::getArticleURL explains this workaround through something about RESTBase not expecting slashes, but that doesn't seem relevant... [23:56:08] 10Analytics, 10EventBus, 06Services (doing): EventBus produces non-canonical page urls - https://phabricator.wikimedia.org/T155066#2947731 (10Pchelolo) Removing the #analytics tag since it's not related to analytics.