[00:22:50] drdee: ready for new patch sewt [00:22:53] *set [00:29:50] drdee: please review https://gerrit.wikimedia.org/r/#/c/32137/ [00:29:56] when you have time [00:31:19] average_1rifter, I think he's getting dinner now. He said he'll be back later. [00:36:40] louisdang: no problemo, I'm going to be here for a while [00:36:45] louisdang: thanks for telling me [00:36:53] average_1rifter, np [00:57:02] getting some coffee now, then i;l do the review and then some groceries :) [00:57:14] drdee: nice, thanks :) [01:32:32] average_1rifter, can you make me the maintainer in the change log for all versions so that we can sign it using PGP? [01:35:43] yes [01:35:48] drdee: yes [02:38:32] average_1rifter: that's the only comment I have, after fixing that ill merge it [02:40:13] drdee: yes, just refactoring debianize.sh to accept parameters [02:40:21] so I can have something like [02:40:22] sorry, I meant git2deblogs [02:40:23] so I can have something like [02:40:36] cool [02:40:44] ./git2deblogs --generate --force-maintainer-name="..." --force-maintainer-email="..." [02:43:14] awesome! [02:44:42] i am still confused about the PLACEHOLDERS in configure [02:45:04] when are those replaced? [02:45:08] so basically there are two directories [02:45:10] udp-filters [02:45:11] and [02:45:19] udp-filters-0.3.etc [02:45:21] depending on the version [02:45:28] now, udp-filters is the one in which we develop [02:45:37] and udp-filters-0.3... [02:45:47] is the one in which dpkg-buildpackage is built [02:46:02] all replacements are made in udp-filters-0.3.. [02:46:58] so inside debianize.sh, it will change directory to udp-filters-0.3.. [02:47:05] and it will run some oneliners to replace versions [02:47:12] afterwards [02:47:26] so one of the versions it will replace is in configure.ac [02:47:46] then dpkg-buildpackage will at some point run compile.sh [02:47:53] and compile.sh will run autoconf [02:47:58] now that autoconf takes stuff from configure.ac and creates configure [02:48:10] so the version which was replaced by debianize.sh is now in configure too [02:48:17] but only in the udp-filters-0.3.. [02:48:45] the same goes for src/udp-filters.c [02:48:51] src/udp-filter.c I mean [02:49:13] where the version is being replaced directly by debianize.sh [02:49:51] that version in configure.ac is important because I think it will be passed to help2man and the manpage generated contains it [02:50:43] the only reason you see PLACEHOLDERS inside configure in the repo right now, is because I ran compile.sh manually inside udp-filters [02:50:57] which is the development directory [02:52:05] so compile.sh from udp-filters development directory ===> autoconf ====> generates configure with "PLACEHOLDER" string.. literally inside configure [02:57:39] thx :) [13:47:47] good morning!!! [13:58:01] yoyoyoyoyo [13:58:07] good morning milimetric [13:58:39] woooooo! [13:58:41] :) [13:58:54] two storms down, no basement flooding yet [14:03:44] you are on a roll! [14:22:46] yo ottomata [14:23:21] so i debugged with louisdang a sqoop job that runs using oozie inside hue :) [14:23:31] it works. [14:28:40] yooooo [14:28:48] yeah? [14:28:57] any luck? [14:29:10] it works [14:29:31] it was non-trivial as there are a lot of parts that all need to be synced up [14:29:52] but we can now create a workflow that would import data from enwiki on a weekly basis [14:30:05] and run it automatically like a cronjob [14:30:31] you should take a look at http://incubator.apache.org/ambari/ [14:30:55] might be useful, it's puppet stuff for deploying, monitoring and configuring hadoop clusters [14:32:08] its puppet? [14:32:41] yes [14:35:55] wow looks thourough [14:36:02] i'm just skimming through some puppet manifests [14:36:18] poorly documentd, geez, why don't people document things anymore? (or ever?) [14:36:49] :) [14:37:01] brb coffee and then oozie time! [14:37:16] there's something similar that dschoon liked, can't remember its name though [16:01:39] ottomata: http://lists.vireo.org/pipermail/debiannyc/2012-November/001794.html I just posted this and got about 4 responses offlist. Would you be interested in hearing about them? one of them is free: http://www.wixlounge.com/ [16:03:37] hmm, wix looks cool [16:03:41] we should go check it out someday! [16:04:27] yeah! [16:04:37] if you can GET to Manhattan. can you? [16:04:54] for all I know you live in a yurt right now with only a laptop strapped to your back and as much as you can carry on a bike [16:06:24] oh one more thing ottomata would you like to see http://www.telecharge.com/BehindTheCurtain.aspx?prodid=9238 with me sometime? [16:06:41] it has Aasif Mandvi from the Daily Show [16:07:24] sounds kinda interesting and intense, hmmm, mayyyyybe [16:07:36] i'm out of town til after thanksgiving [16:07:37] HA! [16:07:38] AND [16:07:44] I might spend a night or two in a yurt! [16:08:36] ottomata: cool! I spent a night in a Mongolian-style yurt while walking across England this summer and it was neat-o burrito. [16:09:12] ottomata, see also https://app.asana.com/0/828917834272/1055914021119 for storm [16:12:59] morning average_drifter, how is the debian package doing? [17:29:18] ottomata, quick question: why do the hive.stats.* properties from http://analytics1001.wikimedia.org:8888/beeswax/configuration not match /etc/hive/conf/hive-site.xml ? [17:30:25] hm, i dunno, i suppose they should huh? [17:30:40] i would think so :D [17:31:01] i am trying to get hive metrics written to msyql as you can imagine [17:31:05] and nothing is written atm [17:31:22] or do i need to restart hue? [17:31:49] ich weiss nicht, mein freund! [17:31:54] i'll try [17:31:58] but let us see [17:31:59] ok [17:32:24] no restarting hue does not solve the issue either [17:33:04] you don't have metrics thigns in hive-site.xml [17:33:04] do you? [17:33:45] i am just looking as well [17:33:54] * drdee is flabbergasted [17:34:09] is puppet running by default?! [17:34:21] at least that explains why it doesn't show the properties :) [17:34:25] dunno [17:34:58] why yes I think it is! It shoudl be running the wmf puppet, not kraken [17:34:59] grrr [17:35:31] ok stopped it [17:35:36] edit away :) [17:38:29] ty [17:47:23] haha, drdee: [17:47:24] https://github.com/jordansissel/fpm [17:48:26] looks cool [17:48:55] yeah, this is using it [17:49:02] https://github.com/wmf-analytics/storm-deb-packaging [17:49:05] (i forked) [17:49:09] gonna chnage some versions and such [17:49:10] looks nice! [17:49:11] k [17:49:12] and easy! [17:49:23] presentation too: [17:49:23] https://docs.google.com/present/view?id=0Aa9liCTsAyzRZGNtd3dkOTRfMTdmczY2azlkcg [17:49:23] still no hive metrics in mysql btw [17:49:37] also no errors to be found [17:50:54] yeah, that's how far I got when I tried it out [17:56:38] https://plus.google.com/hangouts/_/2e8127ccf7baae1df74153f25553c443bd351e90 [18:01:47] ottomata, milimetric ^^ [18:02:51] trying [18:32:45] hi drdee, got anything for me? [18:34:06] yeah let's do oozie stuff! [18:34:24] how about creating a serious workflow that would import a number of tables from wiki? [18:34:32] sure [18:43:11] so louisdang let's create an ether pad with what needs to happen: [18:43:12] http://etherpad.wikimedia.org/oozie [18:43:26] ok [18:53:55] louisdang: i left you some question on the etherpad doc [18:54:07] drdee, responded [18:54:48] thx [18:57:24] louisdang: do you have enough information to get started? [18:57:33] drdee, one more question on etherpad [18:57:38] ok [18:59:17] otstormata is storm nice to you? [19:04:16] yes at least packaging it is! [19:04:25] i need to change a few of the storm-package-deb settings [19:04:28] which is why I forked it [19:04:34] that repo is so nice [19:04:39] so much thanks to that guy [19:04:39] very cool [19:04:50] nathan? [19:14:48] no [19:15:01] https://github.com/phobos182/storm-deb-packaging [19:15:03] phobos182? [19:15:04] dunno [19:18:12] drdee: pong pong [19:18:21] hey! [19:18:30] you were so quiet today :) [19:18:59] drdee: oh sorry [19:19:00] drdee: https://gerrit.wikimedia.org/r/#/c/32137/ [19:19:02] drdee: new patch set [19:19:13] that's what i am used to :) [19:20:10] merged [19:22:44] :) [19:43:07] you can ask ottomata to install the new udp-filters package on stat1 [19:43:25] then we can geocode all the existing sampled traffic files [19:44:23] fyi: analytics1011-1022 are racked, im working on flashing the drac on them so they have remote access [19:44:29] its slow though, will take most of afternoon. [19:46:12] yep [19:46:30] WOOT WOOT, thanks so much RobH [19:54:28] drdee: so i would normally discuss the tech stuff with andrew, but he isnt about [19:54:37] so when he is, will update him as well [19:54:46] but im not partitioning these, just getting them remotely manageable for him [19:55:01] ok [19:56:03] since the old disk layout isnt going to work, you guys hvae 4 more 2tb disks than the old servers had =] [20:04:16] another 4? per server? [20:05:02] * drdee is thinking that the analytics should wear t-shirts of Cookie Monster eating hard-drives [20:11:22] I'm out to get some stuff, bb in ~1h [20:11:55] k [20:12:14] btw, is there any free/cheap sms sending service with a public API that I could use ? I need to write something to hook into irssi so I can get texted on my phone [20:12:39] maybe whatsapp [20:12:54] ok, thanks [20:21:38] AAARRGGHHHHHHH [20:22:13] in mysql giving rights to user@localhost != user@127.0.0.1 [20:23:00] ottomata: [20:23:07] drdee: AAARRGGHHHHHHH [20:23:07] [3:22pm] drdee: in mysql giving rights to user@localhost != user@127.0.0.1 [20:23:18] i got hive stats to work [20:24:03] ? [20:24:06] whatcha doing? [20:25:51] i got hive stats to work [20:26:06] i reassigned the asana task to you with the steps [20:26:20] in the end this was my problem: in mysql giving rights to user@localhost != user@127.0.0.1 [20:26:40] so mysql kept complaining about access denied [20:26:50] and i am like i am sure i granted access to the hive user [20:26:53] etc etc [20:27:11] yeah, localhost != 127.0.0.1 according to mysql [20:27:14] so it was a permission problem? [20:27:44] lemme do that now, I just finished the storm packaging! [20:27:47] so easy! so nice [20:28:03] i made a few tweaks to some of the packaging stuff to make it not conflict with existing libzmq package, etc. etc. [20:28:08] now its in the kraken apt repo [20:28:15] installing storm should just be apt-get instal lstorm [20:28:16] weeeee [20:28:30] https://app.asana.com/0/828917834272/2399816680078 [20:28:43] ja see it [20:28:45] oh you knew this localhost != 127.0.0.1 according to mysql [20:28:47] ???? [20:29:14] that i knew, yeah [20:29:23] hive errors are in /tmp/ not only is the username different (usernames in mysql are the full user@host) [20:29:31] /tmp! [20:29:33] crazy [20:29:34] but also [20:29:43] mysql sees 'localhost' as connections coming through the mysql.sock file [20:29:46] we can fix that (i guess) [20:29:50] and '127.0.0.1' as localhost netowrk connections [20:30:48] not sure if it makes sense to me but i figured it out the hard way (without google) ;) [20:31:43] hmmm, weird, hmm [20:31:53] so I did grant these perms on my local when I was trying to figure this out [20:38:57] cool, drdee, done. puppetized [20:39:01] check to make sure it still works [20:39:13] aight, let me know when puppet has run [20:39:29] should we fix the log location of hive or just leave it at the default? [20:40:14] ottomata, RobH was looking for you, the Dell's are racked up and he was fiddling with them [20:40:21] its run [20:40:23] Whoaaaaaa [20:40:30] yeah my adium crashes sometimes [20:40:36] and when I reopen it [20:40:39] i forget to join the chat rooms [20:40:41] it doesn't auto join me [20:40:48] RobH [20:40:54] tell me about things! [20:41:20] hmmm, well they are per-user logs [20:41:23] which is good in /tmp [20:41:41] k [20:42:48] ottomata: 1011-1022 racked [20:42:54] setting up the drac stuff now [20:42:57] will be awhile longer [20:43:07] (but will have them accessible today for you) [20:45:00] ok cooooooooool [20:45:04] exciting! [20:50:39] ottomata, are you gonna play with the dell's now or is that too early? [20:51:06] need drac set up, probably only working for an hour or so more today [20:51:11] so sounds like a tomorrow project [20:51:32] yea i will be settin gup drac stuff for another 30 to 45 minutes atleast [20:51:49] these didnt come enabled, so have to boot off iso, enable trial drac port, then go back and upload the actual drac key [20:51:54] cool, i might start playing with them for a min, just to get my bearings and see if the scripts we have currently will work [20:51:57] and start doign them tomorrow [20:52:04] well, none are ready yet ;] [20:52:06] ah annoying, but cool [20:52:09] yeah yeah [20:52:10] when they are [20:52:15] im batching, doing all the flashing, then all the programming, then all the uploads [20:52:24] yayyyy [20:52:26] thanks [20:52:37] you wanna spend your final hour on the jobhistory server stuff? [20:52:46] sure, actually, aeyh thats a good idea [20:52:52] reading some storm docs now, but lemme put that aside for now [20:52:53] cool [20:53:47] so it seems that all the logs of finished jobs are stored on hdfs://var/log/hadoop-yarn/apps//logs/ [20:53:52] yeah [20:57:44] which log file shows the "Failed redirect for container" message? [20:58:17] good q [21:01:59] hadoop has some many log files…….. :\ [21:02:51] probably the node manager log files... [21:05:40] tailing all logs now, don't see anything [21:06:20] so periodically, i see this message in /var/log/hadoop-mapreduce/yarn-mapred-historyserver-analytics1001.log [21:06:26] INFO org.apache.hadoop.mapreduce.v2.hs.JobHistory: Starting scan to move intermediate done files [21:10:36] * drdee is reading yarn-mapred-historyserver-analytics1001.log  [21:10:50] (i just restarted historyserver, fyi) [21:10:54] but that's not periodically i would say, it's all the time [21:11:09] but there is not a single mention of redirect error [21:11:16] maybe that error is caught by jetty [21:11:26] as it is a web / http error [21:11:49] eah [21:12:32] i am just generating ideas here :) [21:14:56] what is your suspicion? [21:15:11] i think that the jobs are not being moved properly [21:15:26] from /user/history/done_intermediate to /user/history/done [21:15:26] not sure why t hough [21:15:40] i just set those values explicitly [21:15:54] i'm gonig to restart all hadoop and see if it makes a difference [21:15:59] how about tweaking /etc/hadoop/conf/log4j.properties [21:16:03] and get more data? [21:21:43] growl internet [21:22:03] not sure if you received this msg: [21:22:07] drdee: how about tweaking /etc/hadoop/conf/log4j.properties [21:22:08] [4:16pm] drdee: and get more data? [21:22:19] i think that the jobs are not being  moved properly [21:22:19] 4:15 [21:28:19] this is the relevant source code: https://svn.apache.org/repos/asf/hadoop/common/tags/release-0.23.0/hadoop-mapreduce-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/java/org/apache/hadoop/yarn/server/nodemanager/webapp/ContainerLogsPage.java [21:34:13] back [21:35:05] welcome [21:35:10] :) [21:35:15] ok so I'm going to do the editor configs now [21:35:42] ? [21:35:52] drdee: wikistats editor separte config .pm files [21:36:01] *separate [21:37:55] getting a new instance of wikistats ready that only consumes the sampled edits traffic [21:38:16] where do I get the edits traffic logs ? [21:38:22] are they somewhere on stat1 ? [21:38:55] or maybe I have to grep from existing log specific urls which are the ones from the edits [21:40:24] stat1:/a/squid/archive/edits/ [21:40:30] alright [21:40:31] or something like that [21:45:48] ok, restarting hadoop one more time [21:48:11] ottomata, are we on a roll today or what?? [21:48:25] i am all PUMPED [21:48:53] yeehaaaawwww [21:49:20] we put the history in the yarn in the server in the netty going backy to the hue weathering the storm racking up some boxes all back 2 back with [21:50:27] hahaha [21:50:29] drdee is freestyling! [21:50:59] so the failed redirect thing still happens [21:51:05] cRAp [21:51:14] bout….. the great news is that we really are getting closer with configuring the entire stack [21:51:40] wait [21:51:46] what redirect? [21:51:54] i mean from where? [21:52:05] so, for new jobs that have not yet rotated into job history [21:52:13] i et the failed redirect for container [21:52:16] when I try to view the logs [21:52:22] me too [21:52:36] so i'm not sure what that is [21:52:55] i think maybe you can't view completed job logs until the job has been moved into job history? [21:53:33] yeah, my job just got rotated, and I can find it in history server [21:53:35] yeah…. or maybe it' still syncing up old logs? [21:54:10] yeah, i can get logs while the job is running [21:54:14] http://analytics1009:8042/node/containerlogs/container_1352411161525_0003_01_000001/otto/syslog/?start=-4096 [21:54:23] but once it finishes [21:54:24] not [21:54:32] and I have to wait til job is put into history [21:54:35] which i guess is fine [21:54:39] that's even what the error says [21:54:42] Unknown container. Container either has not started or has already completed or doesn't belong to this node at all. [21:54:46] it has already completed [21:55:02] so I guess that is normal [21:55:51] ok [21:55:55] this one is still empty: http://analytics1001.wikimedia.org:19888/metrics [21:56:12] i was hoping it would be fixed now as well [21:56:56] wasszat? [21:57:35] oh whoa [21:57:38] this is useful: [21:57:38] http://analytics1001.eqiad.wmnet:19888/conf [21:57:43] yup [21:57:59] this is super useful information [21:58:11] info on what conf was executed with a particular job [21:58:19] all kinds of instrumentation data [21:58:30] really useful for debugging [21:58:57] louisdang: the job history server is now also working so you should be able to debug stuff much faster [22:01:06] drdee, cool thanks ottomata and drdee [22:01:41] ottomata: can we deploy the new version of udp-filters ? [22:02:07] bwerrrrrrrr? new version! [22:02:08] cool [22:02:12] just on stat1 for now [22:02:12] ummmmmmm yes but not tonight! [22:02:19] alright [22:02:35] ottomata, pretty please :) can you just install it on stat1? no puppet [22:03:10] oh on stat1 no probs, [22:03:14] thx! [22:03:17] ummmm wait, can't you just use the bin? [22:03:21] do you need it 'installed'? [22:03:28] mmmmm good point [22:03:30] you can just compile it there, right? [22:03:35] or even copy over the bin [22:03:40] average_drifter: can you just copy the bin to stat1? [22:04:21] yes [22:05:24] so just do that and go ahead with writing a bash script to geocode all existing files in /a/squid/archive/ [22:19:22] got a package on build1, and binaries [22:23:08] ottomata: Ok, so the new servers are onlie and remotely accessible [22:23:22] their dhcp files are wrong and outdated [22:23:48] (they are com2 115200, so they will go in linux-host-entries.ttyS1-115200 file [22:30:23] brb [23:59:50] as might be deduced from the hour, Internet Installation was a failure. [23:59:56] i think we will tragically have to switch to comcast :(