Skip to main content

Browse the archive

https://github.com/wikimedia/operations-puppet
12 April 2024, 10:39:33 UTC
sort by:
Revision Author Date Message Commit Date
48be930 Antoine Musso 15 April 2014, 14:33:31 UTC contint: extract android SDK dependencies to a module The Android SDK requires a few specific packages and ant 1.8+. It is being used by other projects than contint (ex: on Tools Labs) so this is factoring out the code to a new androidsdk module. * creates module androidsdk * get rid of contint::android-sdk, moved to androisdk::dependencies * Tools Labs might want the packages to be ensure => latest, so have the dependency class to accept a different ensure value for packages it defines. * be paranoid and only define Package['ant'] when it is not there already. * remove obsolete ant18 class since Ubuntu Precise ship ant 1.8 by default * Move out android SDK include from gallium in favor of an include in contint::packages (one less line in site.pp). Change-Id: Ic37af0fd68fbabe0c8defbdc865364e142118290 15 April 2014, 14:37:38 UTC
5dcc016 Alexandros Kosiaris 15 April 2014, 13:30:34 UTC Revert "Adding host deleteme.wikimedia.org" This reverts commit c0ffd928fabb645ffcfb2c61bb357e8c773d0cf7. Change-Id: I0737cbb70e4093e567e9e4aa8894d56dd8c1032d 15 April 2014, 13:32:57 UTC
0bb629b andrewbogott 15 April 2014, 12:38:53 UTC Remove virt12-15 from puppet 13, 14, and 15 never existed. I'm removing just 12 as a test in anticipation of decommissioning all of the pmtpa virt hosts. RT 7280 Change-Id: Ia7e0f1a9a2f337656f25c228c59cd41b58632fb3 15 April 2014, 12:52:15 UTC
dc0b55c dzahn 15 April 2014, 08:15:25 UTC add monitoring for Ubuntu mirror being in sync see RT #3793 for an issue in 2012 when this was not updated and went unnoticed for a long time Matanya added a script that i slightly fixed in I4d6e47ed7776a This commit replaces that script entirely with a much better version that has no fetching stuff for remote urls and as such is more secure This is a slightly modified version of the script found in: https://wiki.ubuntu.com/Mirrors/Monitoring%20Scripts moving that around to run via NRPE and adding it to installserver module, Ubuntu mirror class Change-Id: I6024b3b57906ba0b023a94ba1375be71174a6e72 15 April 2014, 12:28:08 UTC
a3491ac dzahn 15 April 2014, 12:04:49 UTC decom db78 (was once fundraising db) RT #6543 says it should be shipped to new dc to be reused, just like db64-67,db70 which we just did Change-Id: I803949a07cf27660e115c746c9976bfe0d1acecd 15 April 2014, 12:07:12 UTC
db03c29 dzahn 31 March 2014, 13:52:37 UTC remove db67 from coredb,decom db64,db65,db66,db70 - remove db67 from role/coredb config - remove db67 from site.pp - remove db64,db65,db66,db67,db70 from DHCP ok to remove this researchdb? RT #7129 Change-Id: I2b6bf8555b0f3af120272e7dfb36f8b7e39aa23f 15 April 2014, 08:29:11 UTC
8ccc96e Ariel T. Glenn 15 April 2014, 05:38:43 UTC subversion role lint, pass svn hostname to cert check Change-Id: I27ab362127cf3bb0735670ead112731286fb3df9 15 April 2014, 05:38:43 UTC
0e6621c Antoine Musso 14 March 2014, 17:02:23 UTC openstack: generic::upstart_job() now uses boolean values The define has been updated with 4887dc90 to use boolean value. Follow up 7afabf6 added a back compatibility fix to accept string. I guess we really want to use boolean values. Change-Id: I2c66c5e6a630b382c6f22b822cfbabcb5fd03484 14 April 2014, 21:40:15 UTC
da248c5 Andrew Otto 14 April 2014, 20:36:10 UTC Updating zookeeper submodule with proper class name Change-Id: I3d096e139eac976e06ebfd644b4b63d35dd6d8f1 14 April 2014, 20:36:10 UTC
d87488e Andrew Otto 14 April 2014, 20:12:08 UTC Sending Zookeeper JVM stats to Ganglia Change-Id: I7737addae75c0d5e1c6768fd88b3aa6712efe8e1 14 April 2014, 20:18:00 UTC
c0ffd92 Alexandros Kosiaris 14 April 2014, 17:38:01 UTC Adding host deleteme.wikimedia.org Really really temporary, revertion is imminent Change-Id: I6165e7de17bc3e83a30a71066811a2f9a291b519 14 April 2014, 17:40:37 UTC
6c0898b Ryan Lane 17 December 2013, 23:13:06 UTC Use eth0 IP rather than localhost for multi-region For testing multi-region support in labs it's necessary to use the eth0 IP address rather than localhost so that the services will be reachable from the OpenStackManager instance. Change-Id: Ie92c14c0db940e6ccf5e8cc18add1e35c73d975a 14 April 2014, 17:35:31 UTC
12f81a5 Ryan Lane 17 December 2013, 17:47:27 UTC Enable a labs site override option for nova config It's necessary to test multiple regions of openstack within labs so this change allows the configuration to pretend it's a different site. Change-Id: I4308d0ac71051044179acfc408c32607ea751033 14 April 2014, 17:06:51 UTC
bdf1129 dzahn 14 April 2014, 16:09:01 UTC remove subversion role from formey already migrated to antimony and now http config on formey is broken anyways because the certs have been cleaned svn.wikimedia.org is an alias for antimony.wikimedia.org. see how antimony includes the same role RT #6134 Change-Id: Ic83de3a15fcfff90c41226740e98b36db053bde7 14 April 2014, 16:11:19 UTC
8c25524 Andrew Otto 14 April 2014, 15:57:30 UTC Can't use domain names because ip6tables gets upset Using stat1001 and stat1003 IP addresses Change-Id: I04e00712a30b2353daf1eff656fabb843ccba285 14 April 2014, 15:57:30 UTC
2d1ceab Andrew Otto 14 April 2014, 15:46:17 UTC Allowing rsync from public stat* servers, as well as internal Change-Id: I494b4c71e95d7e17bb001a2bae1d7afd30b5b18f 14 April 2014, 15:49:55 UTC
79bbd64 Alexandros Kosiaris 14 April 2014, 15:19:52 UTC Dont check speed on ifaces that dont report one Also fix a typo (EDIT_CODE vs EXIT_CODE) Change-Id: Iad58b129f9e81b3c2fad5a2f905cfdfdec2232e1 14 April 2014, 15:21:50 UTC
798f878 Ariel T. Glenn 14 April 2014, 14:44:22 UTC ms1001 should no longer sync from nas, long since obsolete Change-Id: Iff43edd5fedb996f98942ce0a1d4537457fc7e22 14 April 2014, 14:44:22 UTC
02d51ae andrewbogott 14 April 2014, 14:35:05 UTC Merge "Exclude virbr0 from eth interface tests." into production 14 April 2014, 14:35:05 UTC
0c7d5bd Alexandros Kosiaris 14 April 2014, 12:59:31 UTC Correct land_polygons file names Change-Id: I12f7df5b4601fc502ffec8baf503ac42bb8c82b6 14 April 2014, 12:59:31 UTC
4dc4757 Alexandros Kosiaris 14 April 2014, 12:55:10 UTC Merge "Import coastlines and land polygons into OSM db" into production 14 April 2014, 12:55:10 UTC
aac1b27 Alexandros Kosiaris 03 April 2014, 21:57:00 UTC Import coastlines and land polygons into OSM db * Using a manually and previously downloaded WGS84 coastline file from http://openstreetmapdata.com/data/coastlines we populate a coastlines table in OSM dbs * Using a manually and previously downloaded WGS84 land polygon file from http://openstreetmapdata.com/data/coastlines we populate a land_polygons table in OSM dbs * While at it, fix sysctl::parameter priority * Rename osm::populatedb to osm::planet_import and move the shapefile import functions to osm::shapefile_import * Update docs and rspec tests Change-Id: I457b92a8a9339a80874db3e7e13fc364fd8b0822 14 April 2014, 12:08:50 UTC
adeb7ac dzahn 31 March 2014, 08:42:39 UTC lint role/keystone (labs) 57 x "tab character found" 48 x "double quoted string containing no variables" 23 x "=> on line isn't properly aligned" this trades unaligned => for lines over 80 chars though :p Change-Id: Id8a90030be50456d6c783bbcbe0ae44ef20f59d7 14 April 2014, 12:06:22 UTC
4610124 dzahn 14 April 2014, 12:03:14 UTC Merge "Remove deprecated roles, and lint remaining labsproxy.pp" into production 14 April 2014, 12:03:14 UTC
95482e7 dzahn 31 March 2014, 09:08:39 UTC lint role/deployment 20 x "tab character found" 37 x "double quoted string containing no variables" 2 x "quoted boolean value found" 1 x "ensure found on line but it's not the first attribute" -- 13 of 50 x "=> on line isn't properly aligned" (i know matanya will hate partial work, but try aligning all => in this one correctly, heh, and you will see. it could be correct but look super ugly anyways, i would call it a good example of the limits of following style) :) Change-Id: Ia5ac299c4661beef05ac2c6ff704838e98a99193 14 April 2014, 11:52:59 UTC
17c11f3 dzahn 31 March 2014, 08:51:42 UTC Remove deprecated roles, and lint remaining labsproxy.pp Change-Id: Id2a3bcdd75050c4e5a74302b27e17bb4cec5fe5b 14 April 2014, 11:43:17 UTC
aefe212 dzahn 14 April 2014, 11:15:22 UTC Merge "puppetize apache-graceful-all" into production 14 April 2014, 11:15:22 UTC
0758a87 Faidon Liambotis 14 April 2014, 10:56:08 UTC ssl: fix whitespace-damaged certificates Several of our certificates were with DOS line-endings (CRLF), which results in funky bundles and openssl troubles when parsing them. Fix that. Three more certificates had an inexplainable space at the start of the file. openssl also borked on that and it's surprising how our webservers don't. Fix this as well. Change-Id: Iab073c7375eff19659d38fbb215df3b6e2b84cfb 14 April 2014, 10:58:19 UTC
1c6288a Dzahn 14 April 2014, 10:50:37 UTC Merge "apt mirror: add monitoring script to verify mirror is up to date" into production 14 April 2014, 10:50:37 UTC
d8df62f Matanya Moses 11 February 2014, 18:53:54 UTC apt mirror: add monitoring script to verify mirror is up to date RT #3793 Change-Id: I4d6e47ed7776a92310ffba7dc63e6de77576a710 14 April 2014, 10:33:33 UTC
8dbd419 Matanya Moses 03 April 2014, 12:43:30 UTC decom : brewster RT #6133 Change-Id: I4dc76fe4a60679916dfc34508501015c1450d068 14 April 2014, 07:45:44 UTC
89add6c Alexandros Kosiaris 14 April 2014, 07:15:13 UTC Merge "Create and import shapelines from a pre-existing dump" into production 14 April 2014, 07:15:13 UTC
9224557 Ori Livneh 14 April 2014, 04:33:58 UTC Graphite/Carbon: less logspam Carbon's logging is rather verbose by default. We don't really need to know every time something connects to something else. Debugging is typically done via tcpdump anyhow. Change-Id: I56feba5e20de2ea932927a6951d3a4bd5347c2a2 14 April 2014, 04:33:58 UTC
8dc02a3 Ori Livneh 14 April 2014, 04:27:18 UTC Graphite/Carbon: set max open files via Upstart 'limit' directive Upstart does not respect /etc/security/limits.{conf,d}. If you want to set limits for Upstart jobs, you have to do it explicitly in the job configuration, via the 'limit' directive. This patch set ulimit -n to 65k for Carbon instances and removes the limits.d file. I'll purge the file from tungsten manually. Change-Id: I4fa1ccf58ff700d4521a5cf19de5309b57aa9296 14 April 2014, 04:30:25 UTC
3bb4856 Andrew Otto 11 April 2014, 21:04:21 UTC Running CirrusSearch-log at 0 minute every hour Change-Id: Iebcc0260f5d481a35195f7035d4baf013d5c805b 11 April 2014, 21:04:21 UTC
0ee8ba0 Andrew Otto 11 April 2014, 21:01:30 UTC Running CirrusSearch-slow.log logster job every 60 minutes This should make the hourly average more accurate, and keep the alert from flapping whenvever there is a single log Change-Id: If2b79f5928522bf9063fc473f09f3cfa13e5d0bc 11 April 2014, 21:01:30 UTC
2688aae Andrew Otto 02 April 2014, 22:35:57 UTC Adding logster module and using it to monitor CirrusSearch-slow.log This is for RT 7118 Change-Id: I92e4b9c4fd29f802de38ac127d4dcbc58addddf1 11 April 2014, 18:05:42 UTC
77e34e4 Nik Everett 09 April 2014, 18:43:14 UTC Elasticsearch config to better spread shards The default Elasticsearch configuration was putting too many of enwiki's shards on one matchine. This makes sure they are more even. Change-Id: I02d15cab1678c9a56a7b9e92854d9235af93b2ce 11 April 2014, 17:59:20 UTC
aba119f Andrew Otto 11 April 2014, 17:19:11 UTC Enabling base::firewall on stat1003 Change-Id: Ic734b51a3f1071cbbebb442caa1fb88f956f6399 11 April 2014, 17:19:11 UTC
97be7d6 Andrew Otto 11 April 2014, 15:09:15 UTC Using exec path rather than fully qualifying I'm getting a puppet error on this one Change-Id: I61174c0dd63813d39398d678c2453e3e0aac58a1 11 April 2014, 15:09:15 UTC
6f75727 Andrew Otto 11 April 2014, 15:05:34 UTC Need to fully qualify all executables in unless command Change-Id: I95eadc296b4b2e162f550e93d3e45aba12bf5d06 11 April 2014, 15:05:34 UTC
a9d3a12 Andrew Otto 11 April 2014, 14:40:03 UTC Adding hdfs user to stats group This is so that hdfs user may write to the /var/log/kraken directory. I am doing this with an exec because I don't want to start managing the hdfs user in puppet just for this. It is currently added by the cdh4 packages. Change-Id: Ifdf1ce4deb6adcc3a6b67c3a47e3a16e9f3fa121 11 April 2014, 15:00:13 UTC
46571b6 Ori.livneh 11 April 2014, 09:41:16 UTC Merge "Revoke my key, I'm relocating to SF" into production 11 April 2014, 09:41:16 UTC
5ca0cef Ori Livneh 11 April 2014, 02:18:52 UTC Merge "uWSGI: strip .ini suffix from instance names" into production 11 April 2014, 02:18:52 UTC
ac38565 Ori Livneh 11 April 2014, 02:16:59 UTC uWSGI: strip .ini suffix from instance names The Upstart instance name for uWSGI apps is derived from the basename of the configuration file, which trims the path but not the extension. Make it trim the extension as well. Change-Id: Ia401ac5a97e0c2d7c56f648bc8a9bcd0113d5a04 11 April 2014, 02:16:59 UTC
6117d07 Andrew Otto 10 April 2014, 21:21:01 UTC Installing java version 7u25-2.3.10-1ubuntu0.12.04.2 for elasticsearch Change-Id: I50507fe73d6989d774f0bf9774d9b0b1381f9521 10 April 2014, 21:21:01 UTC
386756c Andrew Otto 10 April 2014, 20:11:54 UTC Allowing wikimetrics $debug parameter to be set from labsconsole Change-Id: I97b8706350942a7038a6e4787bcdfceac4e44709 10 April 2014, 20:12:44 UTC
6737441 Max Semenik 10 April 2014, 19:28:14 UTC Revoke my key, I'm relocating to SF ...and leaving my desktop behind. Change-Id: I675a590b21f7083b38182b5a77242811f8c1b665 10 April 2014, 19:28:14 UTC
2e5decd Andrew Otto 10 April 2014, 18:51:52 UTC Installing jq on all elasticsearch server nodes Change-Id: I5143f22c6cd1f218c2756548bcacbd848301d62f 10 April 2014, 18:51:52 UTC
24b1465 Andrew Otto 10 April 2014, 17:54:14 UTC Adding howief, jmorgan and msyed to bast1001 This allows them to access stat1003. They have confirmed to have read the Server Access Responsibilities document. Change-Id: Ifc550c1f99b39bdef0ce7440108a6a9dc1bf72d5 10 April 2014, 17:54:14 UTC
5f4f04f Andrew Otto 10 April 2014, 17:38:48 UTC Meant to use raid1-30G, not raid1-1partition for elastic nodes Change-Id: I66ca7e82dcb05adbeea43ba6fe857edbc7ccd8b2 10 April 2014, 17:38:48 UTC
0a551cb Andrew Otto 10 April 2014, 17:02:45 UTC Using raid1-1partition.cfg for elastic* nodes Change-Id: I20cc3a47498b189dafb776aedb0973f72b23ee01 10 April 2014, 17:02:45 UTC
4cf2d44 Alexandros Kosiaris 10 April 2014, 16:25:03 UTC Correct LVS sysctl oid checks net.conf.ipv4 => net.ipv4.conf Change-Id: Ia798bad09640c885e0368a61030516f32fa5ded3 10 April 2014, 16:25:03 UTC
b251b46 Andrew Otto 27 March 2014, 22:09:50 UTC Puppetizing Camus cronjob Change-Id: Idfca2c8b6a8ec77a536487b4e14934fb4ab1e93d 10 April 2014, 15:47:03 UTC
dc612d9 Andrew Otto 10 April 2014, 13:40:56 UTC Setting checkout_submodules and gitfat_enabled for test/testrepo for testing Change-Id: Ic766490151a420c5574e519e1b0f4ec856b7a661 10 April 2014, 13:41:37 UTC
87f90ec Alexandros Kosiaris 10 April 2014, 13:31:47 UTC Merge "Move check_rp_filter_disabled check to role::lvs" into production 10 April 2014, 13:31:47 UTC
5933f73 Alexandros Kosiaris 10 April 2014, 13:25:00 UTC Move check_rp_filter_disabled check to role::lvs nrpe checks should defined in the role classes anyway, plus the previous setup actually implemented then on the icinga host and not the actual LVS servers Change-Id: I8f528babbd89c341d4ba3015f93c7c17fdadb16f 10 April 2014, 13:26:51 UTC
3b5ae31 Brandon Black 10 April 2014, 13:24:17 UTC remove resolv.conf dependencies on dobson+mchenry(.pmtpa) Change-Id: Ib92c580892d577e4acb3cab4a8eb0763ecf4ec1b 10 April 2014, 13:24:46 UTC
6f5b686 andrewbogott 10 April 2014, 13:20:05 UTC Exclude virbr0 from eth interface tests. virbr0 is installed by the libvirt package but we don't use it for anything. Change-Id: Ib3aa11c530b891565d373c18d12d858b9333db9b 10 April 2014, 13:20:05 UTC
5722b12 Antoine Musso 10 April 2014, 12:27:40 UTC jenkins: logrotate main file on a daily basis The Jenkins main log is rotated on a weekly basis. Our setup produce a lot of logs, so it seems better to rotate on a daily basis (keeping 20 days of archives instead of 52 weeks). The current log file for four days is 3.1 GBytes. Also adds ensure => present to /etc/default/jenkins Change-Id: Id6134525872f292056fd02c7fe290345d2223c14 10 April 2014, 12:36:08 UTC
b27ff11 Brandon Black 10 April 2014, 11:57:54 UTC add esams.wmnet to strontium puppet allow_from Change-Id: Ic1d2b2577b477cc4ce47b8fb4f1ec119100a83ec 10 April 2014, 11:58:39 UTC
4578fd4 Alexandros Kosiaris 10 April 2014, 11:55:22 UTC Correct check_sysctl source path Fix typo Change-Id: I08d997211a038851feb7597e18ce15d437902161 10 April 2014, 11:56:04 UTC
ea2e314 Alexandros Kosiaris 10 April 2014, 11:21:57 UTC Merge "Set up a check for LVS rp_filter is disabled" into production 10 April 2014, 11:21:57 UTC
72bd9d3 Alexandros Kosiaris 10 April 2014, 09:35:09 UTC Set up a check for LVS rp_filter is disabled Check all interfaces using an inline_template Change-Id: I562c12bb67954af65ed2a748da3d3990c80f5f6f 10 April 2014, 10:01:56 UTC
691e1d3 Sean Pringle 10 April 2014, 09:59:02 UTC Enable all MariaDB block-based joins algorithms Change-Id: I93f9d399f49ac9ad7c7e4336124154c2b2575a75 10 April 2014, 09:59:13 UTC
51131f5 Alexandros Kosiaris 10 April 2014, 09:58:23 UTC Merge "Have check_sysctl populated on all hosts" into production 10 April 2014, 09:58:23 UTC
8373a25 Alexandros Kosiaris 10 April 2014, 09:16:16 UTC Have check_sysctl populated on all hosts It was populated wrongly in 5f07e69 on the icinga host(s) but this needs to be executed via NRPE on the actual host being monitored. Moving it to the base module and have it populated on all hosts Change-Id: I8777d94d846e0c70a3f532bd80c9be921ad82b86 10 April 2014, 09:52:15 UTC
66f915e Brandon Black 10 April 2014, 09:49:51 UTC remove dysfunctional DNS recursor from LVSes Change-Id: I8ff2ac78cc602217271e2fd665bde5756b1821ae 10 April 2014, 09:49:51 UTC
8f5f197 cpettet 10 April 2014, 08:50:04 UTC collector definition. CPU and Network collector enable. Change-Id: I3d02225713fd81c1018c05d320a490375dfa6341 10 April 2014, 09:23:53 UTC
e966110 Brandon Black 10 April 2014, 08:57:16 UTC add lvs300[1234] to lvs::configuration Specifically, the high-traffic[12] classes and the pybal BGP peer addrs Change-Id: I471f25d00be71e9d24bc29faadf032b381765169 10 April 2014, 08:57:19 UTC
781af0c Brandon Black 10 April 2014, 08:34:25 UTC add public service IPs to lvs300[1234] Change-Id: I0f791b0566c7eeee835b67631559ce72df4e2aad 10 April 2014, 08:36:38 UTC
9c32c40 Alexandros Kosiaris 10 April 2014, 08:34:36 UTC Merge "contint: apply maven settings on labs slaves" into production 10 April 2014, 08:34:36 UTC
47c2ff6 Antoine Musso 09 April 2014, 10:49:28 UTC contint: apply maven settings on labs slaves The production slaves run with the jenkins-slave user which has a maven setting to point maven to our web proxy. The slaves (running jobs with jenkins-deploy) are missing that configuration which cause issue whenever maven attempt to access URLs hosted in labs (such as toolslabs.wmflabs.org). Add a required 'homedir' parameter to contint::maven_webproxy. Set it accordingly for prod and slave labs. Change-Id: I433366ea51740c9a8256a755e8171fab90fb80ee 10 April 2014, 08:23:07 UTC
04f4de6 dzahn 04 April 2014, 12:22:40 UTC puppetize apache-graceful-all goal: shutdown fenari one of the blockers: move Apache config deploy away from it to new deployment host therefore: take the original "apache-graceful-all" (direct scp from fenari) and install it on tin first where the other scripts (sync-apache etc) are already add to deployment. you don't need to say we should do this with trebuchet, already agree, but this is specifically just to move the existing stuff and add an existing old file that appears in docs but not in puppet aware that the content of the file will also needs fixing because of the pathes Change-Id: I005152af28e30735983a094aa4d529ce55b26d6f 10 April 2014, 08:17:35 UTC
afa873f Sean Pringle 10 April 2014, 07:55:13 UTC Enable engine condition pushdown for dbstore Change-Id: I6198f10bc4811f4a01a9471ae7042c60b16ec666 10 April 2014, 08:03:23 UTC
0e2d424 cpettet 10 April 2014, 07:59:41 UTC Merge "module to manage new python-diamond package" into production 10 April 2014, 07:59:41 UTC
4c37492 cpettet 08 April 2014, 15:02:06 UTC module to manage new python-diamond package Change-Id: Iff78d1a20ea40140a437f8ce317fbdc65e58f741 10 April 2014, 07:56:22 UTC
16dc143 Alexandros Kosiaris 10 April 2014, 07:52:42 UTC ignore vnet interfaces in check_eth No point in checking them since these are assigned to VMs and are not physical interfaces Change-Id: Ie1390936353aa0b4f24b3a13687097618c6d0607 10 April 2014, 07:54:58 UTC
4bf8cf4 dr0ptp4kt 09 April 2014, 17:30:57 UTC Add proxy support for 437-01. Change-Id: I85ca72f3ddc89b926d9553441c94fc5ddda1b8e5 10 April 2014, 07:19:06 UTC
d73bcc7 Ori Livneh 09 April 2014, 23:28:24 UTC Merge "Increase count of carbon cache instances to eight" into production 09 April 2014, 23:28:24 UTC
535d5eb Ori Livneh 09 April 2014, 23:28:07 UTC Merge "Corrections to the inline documentation for role::graphite" into production 09 April 2014, 23:28:07 UTC
212385d Ori Livneh 09 April 2014, 23:00:33 UTC Increase count of carbon cache instances to eight tungsten has 16 CPUs with four cores each. Apart from the six carbon-cache instances, the jobs that generate significant load are mwprof (one process, single-threaded), txStatsD (one process, single-threaded), graphite-web (four uWSGI worker processes), and carbon-relay (one process, single-threaded). Profiling the system shows that we are not exploiting parallelism as much as we could -- whenever I looked there are always multiple processors that are just idling. If the incoming Graphite metric load was enough to overwhelm four instances, let's not stop six. Change-Id: I978efd095392c0e262a22ad7deca6cff9d1bb736 09 April 2014, 23:00:33 UTC
b11b4f8 Bryan Davis 09 April 2014, 22:26:06 UTC Fully qualify reference to hhvm class Use '::hhvm' instead of 'hhvm'. Without this, puppet fails on deployment-apache01 with this error: err: Could not retrieve catalog from remote server: Error 400 on SERVER: Duplicate definition: Class[Applicationserver::Hhvm] is already defined; cannot redefine at /etc/puppet/modules/applicationserver/manifests/hhvm.pp:17 on node i-0000011b.eqiad.wmflabs Change-Id: I9077e448b21cdd804ff5896af78d4c6a5d468a35 09 April 2014, 22:26:06 UTC
c6601b8 Andrew Otto 09 April 2014, 21:35:07 UTC Adding analytics networks to $mw_appserver_networks variable This allows trebuchet deployments to analytics nodes Change-Id: I2891dcb8ef51706ca20770253eb7ea1eebea0cba 09 April 2014, 21:37:35 UTC
a5799c0 Andrew Otto 09 April 2014, 19:54:36 UTC Using kraken/deploy repository for deployments instead of kraken Change-Id: I76bd027a37a251ade6388b1ccddbb5355dbdfee0 09 April 2014, 19:54:36 UTC
bd13662 Ori Livneh 09 April 2014, 18:05:42 UTC Corrections to the inline documentation for role::graphite * The number of Carbon cache instances is now six, not four. * Removed TODO about automatically generating the list of destinations. I've thought about this before and was prompted to think about it again by I57d5d85bf. I agree that it's good to be DRY -- configurations that have to match in value should ideally be specified once, to avoid the possibility of error due to mismatching values. But I think the ways of achieving that here entail adding a layer of abstraction on the Graphite module, and that would come at the cost of diminished flexibility (if we design the module to generate configs for a single-node Graphite setup) or added complexity (if we design the module so that it is able to generate configuration files for single-node and multi-node setup), which would in turn mean a greater surface for bugs. So I don't think it's warranted. Note that this is the decision Graphite itself made: its config parser could also be made smart enough to automatically determine the set of caches a relay instance should forward to. But Graphite declines to be too clever about this, and I think it's the right call, to be honest. Change-Id: Icc7fb9adb81eae13dcf532bbd25aede74d94289b 09 April 2014, 18:12:52 UTC
8529042 Ori Livneh 09 April 2014, 16:31:37 UTC mwprof: change collector port to 3811 tungsten receives profiling data on port 3811. A small python script listens on that port and multiplexes the data stream to localhost:3812 and professor:3811. Now that professor has been decommissioned, the Python script is no longer needed, and mwprof can bind to 3811. Change-Id: I192160be50e40ea7db0012ea2b967882ec7418b6 09 April 2014, 16:39:03 UTC
1e3fff7 Andrew Otto 09 April 2014, 16:06:56 UTC Installing make on stat nodes Aaron Halfaker wants this to manage dataset dependencies Change-Id: I1534f734ef5aa1f1781e940731f62cc3e592d8aa 09 April 2014, 16:06:56 UTC
5c2ba2b Giuseppe Lavagetto 09 April 2014, 14:09:33 UTC Adding two carbon-cache workers to graphite. Graphite has had its share of problems in the last couple of days; since it seems to originate from a slowness in getting data from carbon-cache, and the workers are overloaded (100% CPU on tungsten), we think this may ease the load and allow to make graphite-web and gdash usable again. Change-Id: I57d5d85bfaad6133aab5fac3e0638b95cc2bf526 Signed-off-by: Giuseppe Lavagetto <glavagetto@wikimedia.org> 09 April 2014, 15:51:36 UTC
91ecae9 Dzahn 09 April 2014, 15:46:33 UTC Merge "decom: ms6" into production 09 April 2014, 15:46:33 UTC
cf0af38 Filippo Giunchedi 09 April 2014, 15:26:23 UTC RT #7246 add Filippo Giunchedi to icinga cgi allowed commands Change-Id: I4d12616aff90dc7899a203204738e5594cc809d0 09 April 2014, 15:26:54 UTC
37406c7 Andrew Otto 09 April 2014, 15:23:56 UTC Moving sqstat to analytics1003 It looks like it is working Change-Id: I98de7e30daddd5efe8f459d7c02a724a6f7c5b14 09 April 2014, 15:24:39 UTC
8b36418 Filippo Giunchedi 09 April 2014, 15:14:49 UTC RT #7246 add fgiunchedi to icinga contact groups Change-Id: I41a53683d842657cbccb88342b3b97258e904345 09 April 2014, 15:18:16 UTC
b3f371d Giuseppe Lavagetto 09 April 2014, 14:43:57 UTC Fix graphite-web cronjob. We should use shutil.move instead of os.rename, that does not work across filesystems. Change-Id: I55db9f76f9cc78a3a1ae6d6f505d19395a893fdf Signed-off-by: Giuseppe Lavagetto <glavagetto@wikimedia.org> 09 April 2014, 14:56:49 UTC
bfbb946 Filippo Giunchedi 09 April 2014, 14:56:44 UTC Merge "RT #7243 add filippo to admins::roots" into production 09 April 2014, 14:56:44 UTC
5c24c1d andrewbogott 09 April 2014, 14:56:38 UTC Merge "Check the proper cert on the cache boxes" into production 09 April 2014, 14:56:38 UTC
d26b70e andrewbogott 09 April 2014, 14:32:07 UTC Check the proper cert on the cache boxes Change-Id: Iad22593c95a43c851c95526140504a0f32ef970d 09 April 2014, 14:38:18 UTC
25f593c Filippo Giunchedi 09 April 2014, 14:30:12 UTC RT #7243 add filippo to admins::roots Change-Id: I45226234db9e2c62cc2536d9fd77934e1e086ca1 09 April 2014, 14:33:35 UTC
332e3ed Alexandros Kosiaris 09 April 2014, 14:31:52 UTC Merge "Some extra improvements for check_eth" into production 09 April 2014, 14:31:52 UTC
1cb337d andrewbogott 09 April 2014, 14:29:35 UTC Merge "This is still wrong but slightly better!" into production 09 April 2014, 14:29:35 UTC
back to top