[01:04:09] is login working? password submit form hangs indefinitly [01:04:36] [[Special:UserLogin]] that is [01:04:46] on wikitech.wikimedia.org [01:10:11] yurik: confirmed [01:10:12] looking [01:10:45] ah [01:10:55] backups [01:11:06] who needs them? [01:11:29] yurik , it finally completed for me, after 53 seconds [01:11:35] it's back to normal now [01:11:38] the backup finished [01:11:41] works for me now [01:11:44] oh yeah,it just finished too [01:12:02] we need to backup and purge all of the server admin log crap [01:13:20] Ryan_Lane, thx, another thing - could you add dr0ptp4kt to loginviashell? [01:13:48] i tried adding him to medawiki-api, but it required that [01:14:45] one sec [01:16:02] yurik: done [01:16:30] Ryan_Lane, thanks! [01:16:33] yw [08:17:33] !log bots root: killing swap on bsql01 and usin 10gb of that space for db files [08:17:36] Logged the message, Master [09:36:21] hashar, does beta still use git-deploy? [09:36:35] MaxSem: nop [09:36:38] MaxSem: we had a sprint before EQIAD migration [09:36:44] maxsem@deployment-bastion:~$ sync-dir wmf-config [09:36:44] No syntax errors detected in /home/wikipedia/common/wmf-config [09:36:44] no dsh on this host, aborting [09:36:48] to test out git-deploy on beta, but that has been abandonned [09:36:55] MaxSem: hooo sorry [09:37:06] MaxSem: so yeah that is slightly different. One simply has to sudo as mwdeploy [09:37:08] and git pull [09:37:14] lol [09:37:23] /home/wikipedia/common is linked to /data/project [09:37:27] which is shared by all instances [09:37:42] hmm, I sudoed as root - did I break anything?:P [09:37:45] the apaches boxes symlink it as /usr/local/apache/common-local or something [09:37:51] yeah you broke everything :-] [09:38:08] yay [09:38:15] alias mwdeploy='sudo su --login --shell /bin/bash mwdeploy' [09:38:18] works well :-] [09:38:45] just shown everything back to mwdeploy:mwdeploy [09:38:54] will eventually one dir migrate to scap / sync-dir [09:39:01] but that need an update of all those scripts [09:39:21] okay, the changes have been deployed [09:39:40] mobile login works now, but did I break anything else? [09:39:52] just other users will have to sudo root [09:39:55] I will fix the perms [09:40:04] I should get the change automatically deployed [09:40:35] we have a jenkins slave on deployment-bastion now so that should not be too hard [09:41:18] fixing perms right now [10:51:50] Hi! My shell request is fullfilled here: https://wikitech.wikimedia.org/wiki/Shell_Request/DixonD But I can't ssh bastion.wmflabs.org [10:52:13] Could anybody help me? [10:59:26] DixonD1 hi [11:00:06] petan: Hi [11:00:32] seems you already have the permissions [11:00:35] !access [11:00:35] https://labsconsole.wikimedia.org/wiki/Access#Accessing_public_and_private_instances [11:00:39] did you read this? [11:00:46] Actually I don't see my name in the bastion project [11:00:57] mhm... let me check [11:01:18] @labs-user DixonD [11:01:19] That user is not a member of any project [11:01:22] indeed [11:01:38] !log bastion +DixonD [11:01:41] Logged the message, Master [11:01:46] DixonD1 ur there [11:02:00] @labs-user DixonD [11:02:00] That user is not a member of any project [11:02:10] it takes some time to bot to update its db [11:02:20] but you should be able to ssh to bastion now [11:02:24] @labs-user Beetstra [11:02:24] Beetstra is member of 2 projects: Bastion, Bots, [11:02:32] :-) [11:02:34] did not know [11:02:44] :o [11:06:32] well, I still have the same problem: "Server unexpectedly closed network connection." I will try to reupload the ssh key [11:06:43] DixonD1 try bastion2 [11:06:46] !bastion [11:06:46] http://en.wikipedia.org/wiki/Bastion_host; lab's specific bastion host is: bastion.wmflabs.org (208.80.153.194) bastion2.wmflabs.org (208.80.153.202) see !access [11:07:27] the same with bastion2 [11:08:15] ok, can you try using ssh with -vvvv option [11:08:34] then pastebin it somewhere with 10 minutes expiration or that (it may contain some IP's and such) [11:09:37] @notify Ryan_Lane [11:09:39] This user is now online in #wikimedia-dev so I will let you know when they show some activity (talk etc) [11:13:07] DixonD1 sorry [11:13:13] I did a mistake :o [11:13:38] DixonD1 I gave you access to bots instead of bastion [11:13:40] fixed now [11:16:03] hoorray, I've got to bastion2) [11:16:24] actually, I need access to bots as well [11:16:39] that was my initial intent [11:18:34] ok you already got it :D [11:19:12] thanks a lot! [11:20:28] !botsdocs [11:20:28] https://wikitech.wikimedia.org/wiki/Nova_Resource:Bots/Documentation [13:06:43] petan: one more question. I don't have sufficient rights to create a directory in /data/project on bots-{3,4} [13:08:08] DixonD1 first of all, you shouldn't use bots-1, -3 or -4 [13:08:19] you can create that folder in /data/project/userdata [13:08:37] which instance should I use then? [13:08:51] !botsdocs [13:08:51] https://wikitech.wikimedia.org/wiki/Nova_Resource:Bots/Documentation [13:08:55] it is recommended to use grid [13:09:24] that means, you will store your bot on a network storage and use qsub or similar tool in order to submit your job [13:09:34] it will then run on a random execution node based on load [13:11:51] in case you needed a help setting it up let me know... or in worst case you could just manually run it on some execution node (bnrX) but that is a bad idea [13:15:53] let's start from the beginning. After I connected to bastion, to which instance should I connect? I'm just following https://wikitech.wikimedia.org/wiki/Help:Move_your_bot_to_Labs, I'm not really experienced with linux [13:16:30] bots-gs [13:18:00] Ah, sorry, I see it now from !botsdocs [13:18:31] np [13:18:40] just tell me what you need to do and I will tell you how [13:29:52] well, right now I just want to clone code from the github repository and schedule it to run once per day, for instance [13:30:21] okay, you can git clone on bots-gs to some shared folder like your home [13:30:27] or /data/project/userdata/somefolder [13:32:13] then, you should once start your thing to check it works - by doing "qsub -q main.q [13:32:31] for example qsub -q main.q blah.csh [13:35:34] Change on 12mediawiki a page Wikimedia Labs/Tool Labs/Help was modified, changed by MPelletier (WMF) link https://www.mediawiki.org/w/index.php?diff=661799 edit summary: [+726] /* How */ wip [13:37:58] !log bots petrb: installing some basic and useful packages on master server [13:38:03] Logged the message, Master [13:42:17] DixonD1 once you find out it works you can insert this to cron on -gs [13:42:43] Change on 12mediawiki a page Wikimedia Labs/Tool Labs/Help was modified, changed by MPelletier (WMF) link https://www.mediawiki.org/w/index.php?diff=661802 edit summary: [+731] /* How */ mo' [13:51:07] Change on 12mediawiki a page Wikimedia Labs/Tool Labs/Help was modified, changed by MPelletier (WMF) link https://www.mediawiki.org/w/index.php?diff=661804 edit summary: [+845] /* Simple, one-off job */ doc [13:57:49] Change on 12mediawiki a page Wikimedia Labs/Tool Labs/Help was modified, changed by MPelletier (WMF) link https://www.mediawiki.org/w/index.php?diff=661805 edit summary: [+426] /* Continuous tasks (such as bots) */ documentation is good [14:12:55] Change on 12mediawiki a page Wikimedia Labs/Tool Labs/Help was modified, changed by MPelletier (WMF) link https://www.mediawiki.org/w/index.php?diff=661808 edit summary: [+886] /* Web services */ mo' doc [14:14:37] @notify Ryan_Lane [14:14:37] This user is now online in #wikimedia-dev so I will let you know when they show some activity (talk etc) [14:24:57] Coren: Does https://www.mediawiki.org/w/index.php?diff=661804 mean, "jsub -once" will croak when the job already exists? qcronsub is completely silent if no errors occur, and I find that much saner :-). [14:25:39] if you are going to make it silent, then implement -v option that is not silent [14:25:44] because I hate silent commands [14:26:08] btw I tried your jsub - it was silent and nothing happened [14:26:22] so I fallen back to working qsub :P [14:27:27] scfc_de: It will croak, with a message in your .err file by default so you have a trace of "oops, the previous job lasted too long" [14:29:30] Coren: Okay, if continuous jobs aren't started by cron, that may be feasible. [14:30:02] You'd normally don't want to start a continuous job with cron. eeew. [14:30:10] Coren: Tradition :-). [14:30:11] Use jstart [14:30:12] :-) [14:30:32] jstart foobot [14:30:36] jstop foobot [14:30:39] Yeay simple! [14:31:22] Coren: Will jsub have no option to have different files for stdout/stderr per invocation à la qsub's $PREFIX$JOBID.out (or something like that)? [14:32:28] scfc_de: I'd rather not, that's actually a major point of user confusion. If the user is "advanced" enough that they want jobids in output file, I'm sure they can work with qsub. :_) [14:33:07] Personally, if you care about output files, you probably want to use tempfiles you name yourself (foo.err.$$ or whatever) [14:33:28] That's what I'd do if I wanted to use the output imediately after a -sync y for instance. [14:33:43] Coren: Yep, but otherwise you have race conditions: The user truncates/deletes the file, and a parallel invocation's output goes to /dev/null. [14:34:01] I'm not so much concerned that *I* may not be able to cope with that :-). [14:34:20] I would expect that this is a much /less/ frequent scenarion from newbie users. [14:41:10] Coren: Well, I associate newbies with "being very fast at deleting files and clicking on buttons they don't understand" :-). Bots operators must then just note that they need to stop and start their bot if the log gets too large. Otherwise, your setup looks really nice. [14:42:45] scfc_de: Truncating should be fine, actually, given that sge opens the output files with O_APPEND [14:44:57] I've never tested what happens when process A is appending at position X, and process B truncates the file. Does A continue at 0 or at X + 1 with probably sparse in front? [14:45:55] Unix filesystem semantics say that if you have O_APPEND then any write() will write at the current file length (atomically, natch) [14:46:58] What gluster might do.... :-) [14:47:04] Gluster is going away anyways. [14:47:42] Ah: http://pubs.opengroup.org/onlinepubs/7908799/xsh/open.html. Okay, that solves that :-). [14:50:52] petan: thanks for help, finally I've managed to configure and schedule my bot [14:51:00] DixonD1 yay [14:51:06] did you use qsub? [14:51:12] yep [14:51:22] okay cool, don't forget to cron it :P [14:51:30] is your bot using sql? [14:53:13] this one isn't [14:53:36] ok [14:54:02] is there any hope to have database replication set up anytime soon? [14:54:18] that's a question for Ryan_Lane [14:54:30] ok [14:55:35] of course personal sql databases are available [15:04:16] Change on 12mediawiki a page Wikimedia Labs/Tool Labs was modified, changed by MPelletier (WMF) link https://www.mediawiki.org/w/index.php?diff=661814 edit summary: [-450] /* Getting started */ point to new doc instead [15:21:49] DixonD1: Being worked on I believe [15:45:12] Coren, we still don't have an actual schema for service groups, do we? [15:51:16] andrewbogott: Not that I know of. We'll just leap on Ryan the moment he comes on. :-) [15:51:37] * Coren ponders. [15:52:02] andrewbogott: Do we actually need schema changes? [15:52:47] i don't know. I thought that 'We'll be adding two OUs' implied schema changes, but maybe not. [16:01:42] I don't think it does; AFAIK, adding an OU just requires normal write permission to LDAP to, well, just add the OUs. :-) [16:02:36] Hm. Though you'd want /OpenStack/ to do that automagically when you create the project; unless you do so when trying to add the group (OU missing? Add OU) [16:20:36] paravoid: Can I ask you a question about ceph/rados? [16:20:42] oh hi [16:21:14] hi :-) [16:21:27] of course, shoot :) [16:21:40] currently mod_tile does three round trips through rados to serve a single tile [16:22:06] one for stating the tile, one to retrieve the header of the meta tile, and one for the actual tile data [16:22:15] ok [16:22:45] given the latency, I presume it is probably best to try and do some local caching [16:22:59] the network latency? [16:23:07] it's all going to be on the local DC [16:23:17] we're not going to do ceph over wan [16:23:38] but I guess my actual question is how best to store some of the meta data [16:23:55] like what? [16:24:05] currently I store file access time and extra meta data as a header in the actual data blob, rather than using rados attributes [16:24:29] Is there a performance difference to doing it either way? [16:24:44] or other reasons why one would be preferable to another? [16:25:40] the rados attributes should be faster to look up [16:26:02] and it looks cleaner to me [16:26:08] but it doesn't actually seem that important [16:26:20] from a quick glance, I'm no ceph performance expert :) [16:27:09] OK, I'll need to look into it some more to see why I am only getting about 1000 tiles/s out of the rados backend [16:27:37] I saw your mail [16:27:43] we use bobtail, not argonaut [16:27:47] (0.56.x, not 0.48.x) [16:28:08] it shouldn't be a huge difference [16:28:42] installing 0.48.x was as simple as aptitude install ceph.... [16:28:58] same with 0.56 :) [16:29:04] the ceph people provide an apt repo [16:29:05] but if it makes a big difference, I'll need to try and install 0.56 locally [16:29:10] for all debian + ubuntu LTS [16:29:35] 0.56 has cephx (authentication) which makes it a bit more complicated, but you can easily disable that [16:29:39] and basically be exactly like 0.48 [16:30:26] also, 1k vs. 15k is a problem, but if it's the gap is smaller but ceph is still slower it's still going to be a benefit [16:30:38] due to being able to read from hundreds of spindles with ceph [16:30:42] Oh, I just noticed after updating to ubuntu raring a few days ago, I actually now have 0.56 [16:30:48] haha [16:30:49] so I need to redo the performance tests [16:31:45] Yeah, that was all from ram, so disk performance didn't matter. [16:41:47] A wild Ryan_Lane appears! [16:42:07] no schema changes needed [16:42:19] posixuser for users [16:42:27] posixgroup + groupofnames for groups [16:42:46] I think inetorgperson for users is unnecessary in this case [16:43:55] Ryan_Lane: Do you want to create the OU themselves upon project creation or check-and-create as the first service group is added? [16:44:11] with ceph 0.56 I get about 1600 req/s. So definately better, but still not great [16:44:49] It looks like it spend 200% CPU time in ceph-osd and about 150% CPU in apache plus some in ab [16:45:04] Coren: nah, we should create the OUs on project creation [16:45:11] and should make an ldif for the others [16:45:23] So I guess the thing is CPU limited in ceph-osd. Which if you split it accross multiple servers might not matter as much [16:47:03] Change on 12mediawiki a page Wikimedia Labs/Tool Labs/Help was modified, changed by MPelletier (WMF) link https://www.mediawiki.org/w/index.php?diff=661840 edit summary: [+750] db [17:00:12] how are SSH keys handled within Labs? are they in LDAP? [17:01:02] stwalkerster: They live in ldap, but instances get the keys via a read-only shared volume which is periodically updated from ldap [17:01:31] ah, cool. thanks :) [17:18:44] Coren: Where is the config of SGE saved? I'm interested to look into them [17:19:53] Jan_Luca: It's all reachable from qconf, usually through the options starting '-s*'. There's a lot of them, and they are a little baroque, but 'man qconf' gives a complete (if opaque) overview. [17:20:49] Coren: Oh, I can use qconf as normal user, I don't know that :-) [17:21:08] Jan_Luca: You just normally cannot change most things. [17:21:18] OK [17:22:51] iirc you can even use qmon [17:31:06] Coren: Yes, qmon works, but X via two SSH tunnels and half way around the world isn't necessarily a pleasure :-). The splash screen stayed on for several minutes :-). [17:31:38] scfc_de: Splash screens are teh evulz. Unnecessary pixmaps. [17:32:52] What are "tickets", BTW? [17:34:44] scfc_de: They're another way to allocate resources, when you want to distribute them unevenly. (Say, dept X has the right to use 60% of the cluster, and dept Z only 40%) [17:35:08] scfc_de: And you can give ticket overrides for important jobs that need more resrouces, etc. [17:36:35] Coren: Okay. [17:38:56] Coren: The leftmost tab of the main qmon window is obscured by the "GRID ENGINE" icon for me, but I apparently have the possibility to "Add" there (whereas this button is grayed on "Queue Instances" and "Hosts"). Should that be this way? [17:39:54] Leftmost tab is 'pending jobs'. Yes, you can add jobs. :-) [17:41:28] Sure? It looks like http://gridengine.info/files/qmon-hosts_default.png which would suggest "Cluster Queues" :-). [17:45:02] Coren: http://toolserver.org/~timl/qmon-1.png [17:45:30] Huh. It'd let you add a queue? How.... pointless. [17:46:11] I don't know. Could it be that the real magic is in "Queue Instances", and the "Queues" are just blueprints? [17:46:34] * Coren will check once the meeting is over. [17:47:02] Oh, wait, that might just be "add a job on the selected queue" [17:48:35] I'm tempted to try "Modify" :-), but I'll wait till your meeting's over. [17:59:47] Wikinaut: please don't send us emails about random things [17:59:51] Wikinaut: use the wikitech-l list [18:00:25] Wikinaut: we are likely not the right people to answer your question [18:03:13] od -x -N 16 /dev/random | mailx -s "Random thing" labs-l@lists.wikimedia.org [18:05:29] Ryan_Lane: OpenID question: can we map the UID back to a user? [18:06:30] uid? [18:06:49] not sure what you mean [18:07:06] you mean the id we're using for the delegation? [18:07:08] Right, the current scheme uses the user id of the user as part of the identity (and not the username). Can we map that back to a username at need? [18:07:15] yeah [18:07:21] I believe so [18:07:29] for sure inside of mediawiki [18:07:44] Because I'm thinking there are some tools that will want to be able to query 'Is that identified user a checkuser on enwp' for instance. [18:07:47] it's the user's id in the dayabase [18:07:56] oh [18:08:13] my typing is really bad lately :( [18:08:28] well, it would really be nice to give back a list of groups [18:08:35] but that's not going to work so well [18:08:54] since we can really only give back global groups [18:10:20] I'd have to look through special pages and the api to know for sure about your question, though [18:10:49] I'm looking through the API now. [18:11:11] list=users could work, if we could give a global userid rather than just ususers [18:11:27] https://www.mediawiki.org/wiki/API:Users [18:11:38] ususers wants usernames [18:12:19] yeah [18:12:34] most things want a username [18:12:57] Shouldn't be too hard to allow usgid also [18:13:16] And that'd give the local username too which would be perfect for use with everything else. [18:13:19] well, openid comes back with info, including the username [18:13:53] Ryan_Lane: Hm. Does it? [18:13:53] the user is asked if they are willing to share that info when they log in [18:14:03] * Coren needs to test that. [18:29:43] scfc_de: Let's see about qmon [18:38:59] Coren: Okay. Shall I start qmon again? [18:47:25] scfc_de: Yeah, tell me what it looks like it's letting you change. [18:47:53] scfc_de: I'm betting that it doesn't actually check preemptively whether you're allowed to change something and will just fail, but it's worth checking. :-) [18:48:58] Coren: Okay, starting now, will take a few minutes :-). [18:53:54] Coren: qmon is up, now looking at "Cluster Configuration". Should I try setting, let's say, "mailer" from "/usr/bin/mail" to "/bin/false"? (I think mail isn't working anyhow ATM :-).) [18:56:37] scfc_de: That seems safe enough [18:57:19] Coren: And you're right: 'denied: "scfc" must be manager for this operation' :-) So everything's fine. [18:57:32] It's effing crappy UI design, though. [18:59:35] Yep. [19:00:10] (Especially, since on "Queue Instances" some buttons are grayed out, so it's the inconsistency that sucks.) [19:04:16] Ryan_Lane, for entries like ou=groups,,ou=projects,dc=wikimedia,dc=org, is the cn "groups" or "groups,"? [19:14:20] andrewbogott: it needs to be a valid dn ;) [19:14:40] it's cn=groups,cn=testlabs,ou=projects,dc=wikimedia,dc=org [19:14:44] for instance [19:17:52] Do you mean ou=groups,cn=testlabs,ou=projects,dc=wikimedia,dc=org? [19:18:03] * andrewbogott is ever more confused [19:20:13] Oh, wait... [19:22:18] Well, ok, my actual question is: what do I use as my pattern to match the groups entry for a given project? '(&(&(cn=groups)(cn=))(objectclass=groupofnames))' ? [19:24:34] andrewbogott: ooohhh [19:25:08] better would be: objectalsss=groupofnames [19:25:21] with a base of "cn=groups,cn=,ou=projects,dc=wikimedia,dc=org" [19:25:29] *objectclass [19:26:07] Oh, ok! That's like what I tried originally but it wasn't working… maybe just a typo or something. [19:26:24] Well, not exactly what I tried. Hm. [19:30:10] Ah, much better [19:31:26] (setf (get 'actual-usefulness 'lisp) nil) [19:31:43] Erm. [19:32:24] * Coren will never get used to ^W closing windows instead of WERASE [19:32:52] I mean, of course (setf (get 'lisp 'actual-usefulness) nil) [19:38:37] andrewbogott: I probably just broke nova-precise2 [19:38:42] it'll be momentary [19:40:23] Ryan_Lane are you planning on ability to hot plug more storage to an instance? [19:40:34] like /dev/vdc [19:40:44] etc [19:40:44] not currently, why? [19:40:51] it would be cool for some things :> [19:41:04] like if we ran out of space on sql we could online extend it etc [19:41:41] it would likely come from slower storage, if we did [19:41:53] ah [19:42:11] you'd really want to put it fully on that storage, rather than using local storage, if we did that [19:42:38] but that wouldn't give me ability to create own fs [19:42:54] mhm [19:43:01] own fs? [19:43:06] why wouldn't it? [19:43:09] it would be block storage [19:43:13] if I used shared storage? [19:43:16] like gluster [19:43:27] that wasn't what you were asking about ;) [19:43:33] ah [19:43:37] then I didn't understand you [19:43:39] and that wasn't the answer I gave [19:43:50] Ryan_Lane: ATM, documentation I'm writing for the tool labs are on mw.org. Better on wikitech, you think? [19:44:07] Coren: until all project documentation is moved, let's keep it on mediawiki.org [19:44:15] Coren: guillom wants to move the project docs [19:44:15] Ryan_Lane: kk. [19:44:58] Ryan_Lane is there any progress on setting up that dedicated sql server? do you think it could be possible for me to get access there in future given that I would sign these papers you talked about? I am really interested in doing dba stuff :P [19:45:36] hm. I'm not sure how we're segregating that from production [19:45:37] petan: I think that was me telling you about said papers. :-) [19:45:50] Coren ryan actually did some time ago as well :P [19:46:16] and for dba related work, you need to talk to notpeter and binasher [19:46:23] well I meant dba work on that labs dedicated sql [19:46:38] I thought these 2 are working only on production things? [19:46:48] they are our dbas [19:47:18] ah, so it's not going to be set up by you? [19:49:16] probably not [19:49:23] not the dbs themselves [19:49:27] access to them, yes [20:10:44] Ryan_Lane, so I checked the 'redis' class, but it doesn't work since the default directory is /a/redis. [20:10:58] Is there a way I can override that through the web interface, or do I have go full self-hosted? [20:11:27] I see the variables section at https://wikitech.wikimedia.org/wiki/Special:NovaPuppetGroup, but I don't know if it's relevant. [20:13:33] superm401: you can use /dev/vdb and mount it on /a/ [20:14:13] hashar, I guess... Why is the default that anyway? [20:14:24] superm401: I did that in manifests/role/lucene.pp [20:14:31] look for 'vdb' [20:14:54] the relevant snippet http://dpaste.com/1029329/ [20:15:40] superm401: the puppet class are usually written for production [20:15:49] it is up to us to adapt them for other uses :-] [20:16:24] ahh redis has a $dir = '/a/redis' [20:17:03] superm401: so you could add a new role class for labs in manifests/role/redisdb.pp [20:17:34] that would set the $dir to something else ( like /mnt/redis ) or simply mount vdb on /a and leave $dir [20:18:24] labs does not let you pass parameter to a parameterized class, so you need a role class [20:19:18] hashar, what if I add a labsredis that subclasses the main one with sane defaults? [20:19:40] There is already labsmediawiki, etc. [20:20:02] ah or that yeah [20:20:05] maybe :-] [20:21:02] ah I see manifests/role/labsmediawiki.pp [20:21:22] sounds like the same need to be done for manifests/role/redisdb.pp [20:21:36] Yep, I'm gonna go for it. [20:21:41] something like role::db::redis::labs [21:17:31] <^demon> Ryan_Lane: Got a couple easy puppet changes for gerrit: https://gerrit.wikimedia.org/r/#/c/54678/, https://gerrit.wikimedia.org/r/#/c/54677/, https://gerrit.wikimedia.org/r/#/c/54514/ [21:18:10] <^demon> Oh, and https://gerrit.wikimedia.org/r/#/c/53173/ [21:26:16] ^demon: that's way more than a couple ;) [21:26:24] <^demon> Well, < 5 ;-) [21:27:33] Change on 12mediawiki a page Wikimedia Labs/Account creation improvement project was modified, changed by Ryan lane link https://www.mediawiki.org/w/index.php?diff=661913 edit summary: [+39] /* Current account creation process */ [21:38:37] ^demon: you're asking me to merge something you say needs an upstream change [21:38:44] <^demon> That's been merged. [21:38:47] ok [21:51:13] ^demon: what happened to gitblit? [21:51:33] <^demon> It's slow. [21:51:33] <^demon> On big repos. [21:51:46] slower than gitweb? [21:51:55] <^demon> On mediawiki/core, yes. Most other repos, no. [21:52:06] <^demon> Plus it doesn't like uppercase letters in repo names. Which is stupid. [21:52:11] :( [21:52:34] I don't think anything will like mediawiki/core it's huge [21:52:40] I hate gitweb [21:52:53] <^demon> I need to repack mediawiki/core. [21:52:58] <^demon> And probably operations/puppet. [21:53:03] <^demon> Most other repos are fine. [22:04:17] hi ceradon I wanted to ask you something but I forgot what it was :P [22:04:26] always this bot remind me :D [22:05:31] petan, Well I asked you if you could give me access to the db a couple days ago but you were idle. [22:05:43] I can't create databases. [22:05:52] ah that [22:05:56] you should be able to do that [22:06:04] what does it say when you try to make one? [22:06:14] did you use that system.create_db function [22:06:28] call system.create_db("ceradon_data"); [22:06:29] etc [22:06:41] everyone can use this to create new db's [22:07:05] i could of course give you all grant to create new db's but that wouldn't sort out grants [22:07:07] this function does [22:07:27] uh. I killed gerrit ! https://gerrit.wikimedia.org/r/#/c/54986/ [22:07:45] ceradon does it work now? [22:07:55] Service Temporarily Unavailable The server is temporarily unable to service your request due to maintenance downtime or capacity problems. Please try again later. [22:08:01] ah. Back again [22:08:24] Wikinaut how many times we need to tell you don't store ur porn in gerrit :P [22:08:26] it's too small [22:08:49] you can't commit 2gb files :P [22:08:49]