[00:00:14] heh [00:00:20] nimish_g: 'boolean' is also new :) [00:00:24] use tinyint [00:01:16] [worked around locally for the moment] [00:01:42] aww, no booleans either...mysql 4's killing me [00:01:43] MySQL 4.0 doesn't yet support SELECT, you have to just guess if the data is there [00:03:25] *werdna --> bed [00:04:53] hehehe [00:05:19] nimish_g: ok ready to test? :DDD [00:06:38] nothing on test yet with 1/1000 sampling, not too surprising :) [00:06:48] let's crank it up to full [00:06:54] nkomura wanted to hold off till tomorrow morning to set it to "on", in case it produced tons of data, but we could turn it on for like 5 minutes or something at 100% and then see [00:07:00] brion I like the way you think [00:07:35] http://dpaste.org/HRpm/ [00:07:37] woot [00:07:40] *nkomura hides under the desk [00:07:57] event_id 0? [00:08:35] couple more pokes: http://dpaste.org/Vp5L/ [00:08:52] nimish_g: ah i bet you're reading the event_id out of lagged slave connection or something [00:09:03] you probably should use the insert id from the insert [00:10:20] yep that's it [00:11:00] try doing the read from $dbw first (in case you just read from a lagged slave) -- then if it's still not on the master, do the insert, then grab the insertId from $dbw and return that [00:11:48] ok [00:13:20] what happens if they click it 2 times in a row, with not enough time for the slave to sync... the first time you read the slave, decide the event needs a new id, then create it and use the id. the second time you ask the slave the same question but it still doesn't know about the newly inserted row yet cause it's out of sync [00:13:44] thus resulting in double inserting on the master of the same row [00:14:01] well, the same "event_name" [00:14:09] TrevorParscal: right that's what my second suggestion was for :D [00:14:16] is the event_name field a unique key? [00:14:18] sorry [00:14:22] also should do that too prolly [00:14:35] yyyes it's unique [00:14:39] k [00:14:55] good, so in the worst case (two simultaneous write transactions attempting to add it) one would fail [00:15:08] yeah [00:15:10] it would fail out cleanly and just not record that second click [00:15:38] and we can survive lossing 1 click more than having a bunch of rows with the same event_name and different IDs [00:16:28] brion, ok, now it returns the insert ID and all that fun stuff [00:16:50] woo [00:18:08] ok lookin good [00:19:47] schweet [00:22:01] ok update deployed... [00:23:00] yay! I think nkomura wants to set the throttle tomorrow [00:23:51] congrats!!!!! [00:25:01] http://dpaste.org/eeKs/ <- ok that's looking good [00:25:08] yeah since this is our very first time, we don't want to leave it on unattended as the day wraps up [00:25:30] *TrevorParscal agrees [00:26:09] brion, k, I have to run like now, but TrevorParscal's got it under control if anything needs further tweaking...doesn't look like it does tho...yay! [00:26:26] *TrevorParscal takes control [00:27:11] supa [00:53:27] brion: when do you want to head to twitter [00:54:45] when's it start? [00:54:59] start time.... 6:30 [00:55:19] so 6:20ish? [00:56:26] at the corner of 3rd and bryant? [01:04:49] sure, see ya there then [01:04:57] k [15:05:55] RobH: I'm not sure if you had gotten the previous emails on the selenium server, but one outstanding issue is that we need mac support. There are some unused macs in the office that were being offered for this task, but another solution proposed was to use an xserve and use vmware server or parallels server for virtualization [15:07:15] RobH: I'm reluctant to use host-based virtualization as it is terribly inefficient and OS X doesn't allow virtualization of their client OS's only, their server ones. The xserve is likely to be slightly more expensive too [15:08:16] RobH: If we re-use the mac hardware in office, then we need somewhere to put them. it was proposed that we use the server room in the new office. [15:08:19] RobH: ideas? [15:08:42] This was for the quote you sent me a week or so ago right? [15:09:03] yes [15:09:17] yeah, I just didnt forward it up for approval, now i cannot find the damned thing [15:09:23] heh [15:09:24] *RobH is looking for it now [15:09:27] want me to re-send? [15:09:33] if you have it yes pls [15:09:36] ok [15:09:48] I will send over to the Dell rep asap for him to requote with our discounts [15:09:56] (sorry about letting it slip) [15:10:49] it's ok. what's your email again? [15:11:01] rhalsell@wikimedia.org [15:11:02] it isn't saved in my contacts since it sent through dell :) [15:11:59] sent. it is just a link to the cart. that's ok, right? [15:13:33] thats fine [15:15:26] Ryan_Lane: i sent on the information to our dell reps to quote [15:15:35] cool. thanks [15:15:37] they are usually really fast, within an hour half the time [15:15:47] wow. your vendor is really that quick? [15:15:57] its why i love our dell reps. [15:16:04] i can see why [15:16:13] we usually wait a couple days to hear back from them [15:16:29] government services support must also move at the speed of government [15:16:37] Sun takes a day or two, or three, or more if I do not follow up, to get me quotes [15:16:51] it may just be our one rep [15:16:54] do you quote directly from sun? [15:16:57] but we have had a few reps from them like that [15:16:59] yea [15:17:01] ah [15:17:11] our direct dell quotes are fast, cheap too. [15:17:13] we have a vendor who usually gets back within the day [15:17:15] i just like dell servers. [15:17:19] heh [15:17:30] It sucks they are never cutting edge like sun or ibm or hp [15:17:34] we keep having raid controllers fail lately [15:17:35] they are always the baseline [15:17:46] on dells? [15:17:48] yep [15:17:59] on 1950's and 2950's [15:18:07] the perc controllers, thats odd [15:18:12] they are pretty old (~4-5 years) [15:18:23] but shit happens, you may have gotten an unlucky batch that took a long time to reveal the issue [15:18:26] might have been a batch [15:18:28] yeah [15:18:36] they were all ordered around the same time [15:18:36] still stinks [15:19:06] it could be worse. we could have the vendors spec out equipment, and ship us stuff that is incompatible :) [15:19:09] like sun... [15:19:16] *Ryan_Lane shakes fist in air [15:19:31] i like that sun has some neat stuff [15:19:36] like the 45xx series we use [15:19:42] 48 disk 4 u server [15:19:45] its pretty cool. [15:19:57] but all of their stuff is really expensive, and their support sucks. [15:19:57] is that the update to the thumper? [15:20:09] thumper is the 4500, and thor is the 4540 [15:20:18] we have both in the DC now, but thors are what we order now. [15:20:29] ah ok. I visited TACC in austin. they were using those too [15:20:37] they seemed to like them a lot [15:20:39] the odd part is thor is cheaper than thumper [15:20:41] and its better. [15:20:47] heh [15:20:52] so i guess its because of quantity sales they can do that. [15:21:00] probably [15:21:09] replacing disks in those is a PITA right? [15:21:12] we wanted to put our own 1tb disks in the thor [15:21:26] instead of paying sun three times the entire server price for them [15:21:34] but it completely voids the entire support contract [15:21:42] wow. lame [15:21:47] you put in your own disks and the eithernet port goes bad [15:21:53] you are not under support =P [15:21:55] indeed. [15:22:06] well, i dont use cable management arms on any servers [15:22:10] except the 45xx [15:22:23] everything else hot swaps from either the very front or very back you know? [15:22:23] They are finicky about the drives in that unit because of heat [15:22:29] yeah [15:22:32] the 45xx has to pull completely OUT of the rack [15:22:35] fully extended [15:22:38] to get to those disks [15:22:50] you have to replace the drive within a certain amount of time too right? or the unit shuts off? [15:22:58] and if i recall, you have to do it within a few minutes [15:22:59] yep [15:23:14] yeah. the people at TACC were telling me that :) [15:23:37] I always do it pretty quickly, because we have had 5 disk failures in them so far [15:23:38] considering how long we have had them, thats pretty bad. [15:23:55] yeah [15:24:04] so yea, they get the full cable mangement rig in the back, which is a pain and very ugly [15:24:11] heh [15:24:13] but otherwise the cables would snag on something and unplug [15:24:26] i throw away the dell cable mgmt stuff [15:24:30] we don't use cable management either. [15:24:43] I have the mgmt in the rack enclosure, I dont want it on the servers [15:25:04] i think they make it harder to work on them yea [15:25:32] i've visited a few really large data centers. not a single one was using cable management arms [15:26:00] yep =] [15:26:48] heh, interesting note, we are getting this with esxi and all [15:26:49] most of these datacenters were HPCs though, where if a server dies, they replace the entire server with a spare, then fix it :) [15:26:58] and just this week someone fro vmware reached out to us [15:27:03] to see if we have interest in stuff [15:27:16] heh [15:27:31] I'm a fan of vmware [15:27:44] i am, i want enough vmthings for all devs to be able to spin up whatever they need on the fly. [15:28:02] I manage a 24 node cluster. makes things easy [15:28:20] i am not sold on them for anything nearing production level use for our cluster (apache and squid would have little to no benefit running in virtual, and a lot of drawback) [15:28:29] but production for other stuff maybe [15:28:50] though memory deduping for squid may be interesting. [15:29:06] yeah. if a server is being highly utilized it isn't a good candidate for virtualization [15:29:25] the apaches are cpu bound, so yea [15:29:28] there is a decent CPU hit, and an even worse IO hit [15:29:36] squids are disk bound [15:29:44] the deduping may be interesting for some of that [15:29:56] but i dont think it would fix it enough to outweigh the drawback of it [15:30:23] but for things like our blogs, other php based softwares, and such it would be nice [15:30:29] things that are not high load. [15:30:32] yeah. and the HA stuff doesn't help much since you guys are doing everything through load balancers with health checks [15:30:55] HA stuff? [15:31:05] high availability [15:31:09] ah [15:31:20] yea, the stuff that isnt in load balacers would be nice though [15:31:30] single server failure points in a nice esx group. [15:31:36] hardware failure would not crash them [15:31:51] well. hardware failure still does crash them [15:32:01] unless you use fault tolerance [15:32:19] otherwise another node restarts the VM [15:32:21] multiple esx servers hosting the items in the front with shared storage with redundacy in the back [15:32:28] yea [15:32:46] so yea the hardware crashes, but the vm just moves to a new host and runs [15:32:50] yeah. [15:32:57] my experience with it is very very very very limited. [15:33:03] but it seems very neat ^_^ [15:33:07] btw, if you plan on doing an ESX cluster. I recommend NFS highly [15:33:47] unfortunately NFS is a bad word for most of us. Our apaches use NFS for pulling information which leads to entire site outages when NFS server goes down. [15:33:51] the vmware/netapp combo is even nicer. but it would likely work well with ZFS too [15:33:55] So it will be uphill battle. [15:34:10] well, you can choose FCP, iSCSI, or NFS [15:34:16] we are moving away from nfs for that so yea [15:34:22] I would not mind iSCSI. [15:34:22] FCP is expensive as hell [15:34:33] iSCSI and NFS have similar performance [15:34:43] but if you use iSCSI, you have to use VMFS [15:35:11] for the host machines, not the vms within them [15:35:11] right? [15:35:28] yeah. the host machines access the vmdk's over NFS, not the clients [15:35:47] although I use NFS for clients too [15:35:56] this will be a small topic of discussion during the tech meetings this month i am sure [15:36:00] but after that when i hopefully get to tinker with proof of concept [15:36:06] but I have a clustered netapp [15:36:10] you just unwittingly volunteered yourself to help me ;] [15:36:17] \o/ [15:36:18] heh [15:36:23] thats fine. i'm on contract ;) [15:36:36] everybody wins ^_^ [15:36:39] lol [15:37:51] your ZFS server isn't clustered right? [16:52:45] morning nkomura [16:52:57] morning werdna [16:58:20] nkomura: the link you gave me to the feedback data. is that statistics? Is there somewhere I can get an example of a full feedback dump? [16:59:41] Ryan_Lane: the full dump should be there too [17:00:41] all.csv should be the complete data [17:01:11] nkomura: also stats. I think I found it though [17:01:29] cool [17:01:32] nkomura: is it this directory: http://toolserver.org/~catrope/survey/ ? [17:01:51] i think so [17:01:59] ok. cool [17:02:06] but better to confirm with roan [17:02:08] it looks like the userids are hashed [17:02:42] good to hear, i wanted to make sure the most recent dump was treated as well [17:04:31] Ryan_Lane: if you need an account in translatewiki.net, GerardM- can help you [17:04:54] no problem [17:05:52] I was able to create one. I might need some permissions. not totally sure yet. I need to talk to Nike and Siebrand about how we'll be able to use it [17:05:52] just create a profile and ask for translator rights here .. http://translatewiki.net/wiki/Project:Translator [17:06:14] ryan lane what is your user ? [17:06:28] ryan lane [17:08:43] brion, good morning. We were thinking of setting clicktracking to 1%, so setting $wgClickTrackThrottle to 100 [17:08:45] There is no Ryan lane of Ryan Lane profike [17:09:19] on translatewiki.net? [17:09:32] no [17:10:44] http://translatewiki.net/wiki/User:Ryan_lane [17:12:56] Ok you have translator rights ... meaning that you can localised [17:13:09] what is your Babel info ? [17:13:26] I can only read and write english [17:14:05] not terribly useful as a translator :) [17:14:57] We are also looking for people who want to help out with making messages understood ... and people who want to help with our development efforts [17:15:12] so you are really welcome :) [17:15:23] heh [17:15:43] did nkomura mention to you what we'd like to do? [17:16:55] Ryan_Lane: i haven't shared my idea to anyone else yet ;) [17:17:11] heh. ok [17:17:29] feel free to discuss the potential of using translatewiki.net with GerardM- or/and siebrand [17:18:12] When Naoko recommends someone ... I am happy to introduce people into translatewiki [17:18:36] Ryan.. talking at #mediawiki-i18n is best :) [17:18:41] ok [17:18:50] that is where Siebrand and Nikerabbit also hang out [17:18:58] just joined [17:19:02] :) [17:29:21] nimish_g: ok... all good so far w/ test? [17:34:40] brion: yep, looking at testwiki data and it looks good =) [17:36:00] hey TrevorParscal [17:36:11] hi [17:36:15] how's it going? [17:42:37] you guys feel ready to roll it out? you want just enwiki or a variety :D [17:43:58] yep, I think we want a variety. nkomura, any thoughts on which one(s) we want? We have historical data in all the wikis [17:44:12] let's do for all [17:44:25] threshold should be low enough (I hope) [17:45:18] if you all feel we should start lower we can start more conservatively [17:45:28] but since less than 200,000 people have enabled new toolabar so far [17:45:47] the impact should not be too bad [17:50:19] your thoughts TrevorParscal and brion? [17:51:03] the code, as he's saying, only affects users of the new toolbar [17:51:57] a ratio of like 1:1000 or something would be pretty safe [17:52:01] we could throttle up from there [17:53:49] I think 1:100 even might be safe [17:54:04] maybe 1:500? [17:54:39] happy medium [17:55:06] i'd say start at 1:1000 and make sure nothin falls down :D [17:55:33] let's start with 1:1000 [17:56:14] we do not want to melt down the database ;) [17:57:29] :D [17:57:35] ok gotta install the db tables... [18:55:51] TrevorParscal, nimish_g, adam_miller, let's head over to stillman [18:56:08] k [19:57:26] *nkomura_away nkomura [20:01:14] That would be /nick nkomura [20:04:59] haha [20:05:08] absent minded [20:06:57] Hmph [20:07:02] Today is Brion's last day at the offcie [20:07:11] And I still have a pile of babacofix revisions for him [20:08:06] brion-lunch: I don't assume you'll have time to review and deploy 16 babacofix revs during your last 4 hours at the office? [20:10:46] brion-lunch: who's doing review/deployment when you're at statusnet? tim? [20:12:26] I was under the impression that would be one of the topics at the tech meeting (more generally, how we're gonna distribute responsibilities now Brion's gone) [20:13:56] Hm is Trevor in today? [20:14:06] Trevor is in [20:14:46] he is still at stillman [20:15:43] Ah good [20:17:43] brion-lunch will be still be with us once a week [20:17:59] i'm sure he wile packed-jam busy on that day [20:18:16] Yeah I was afraid of that [20:18:41] Isn't that day also a Friday? [20:18:54] but let's ask brion-lunch when he is back when we can ask the review and deployment for ongoing bug fixes [20:19:17] i think the day is not set yet, but i heard he will be in next Wednesday [20:19:33] Yeah Wednesdays make lots more sense [20:27:44] RoanKattouw: I have trips planned for multimedia meeting, christmas, my sister's birthday (but that's a surprise for the day, shhhh), and then on the way back to London [20:28:29] Your sister lives in Paris? [20:28:37] no, she's visiting me in London [20:28:48] Oh so you're taking her [20:29:11] on her birthday, we're going to take her to the train station, telling her we're going somewhere else, and then walk into the eurostar part of the station [20:29:16] hehe [20:29:33] Don't they have like border controls there? [20:29:43] yeah, but that's about when she'll figure it out [20:29:54] Yeah; too bad UK isn't in Schengen yet [20:30:08] that would be nice [20:31:22] they can adopt the euro while they're at it [20:32:48] to be fair, I think the border controls only serve to make this one more exciting [20:34:11] That's true [20:34:19] She's probably never seen a land border yet [20:35:16] she crossed the border from France to Switzerland [20:35:20] but that's in the Schengen Area [20:36:50] Yeah I meant a real border [20:37:07] That one's just a sign saying "Suisse" or something [20:57:10] *RoanKattouw jumps Trevor with some questions [20:57:42] In particular https://bugzilla.wikimedia.org/show_bug.cgi?id=20922#c9 and downwards [20:58:27] brion, is clicktracking on? the tables are on but there haven't been any clicks on enwiki (which is also possible if it's on) [20:58:42] nimish_g: not yet. gimme a few minutes to finish some invoices [20:58:45] Yeah it's /possible/ [20:59:01] okie, paperwork on your last day, funfun [21:00:01] my *last* paperwork ;) [21:10:43] ok where was i nimish_g [21:11:32] I think we were going to set $wgClickTrackThrottle to 1000 [21:12:01] boo-yah [21:14:42] nothing recorded yet for enwiki :) [21:14:48] *brion awaits anxiously :D [21:17:13] *crickets chirp* [21:17:35] maybe let's crank it up to 1:100? [21:18:13] click tracking is on, i gathre [21:18:17] wow!!!! [21:19:36] yeah, I think 100 would work fine [21:20:24] well if there's only 200,000 users using the toolbar, 1:1000 means there's only 200 people out there total who we might clicktrack [21:21:17] bumped to 1:100 [21:28:03] Ooh JavaScript has a statement similar to PHP's list(): var [foo, bar] = functionThatReturnsAnArray(); [21:28:57] huh [21:29:08] i've never seen that before... reliably in all JS engines or is that new? [21:29:26] It says it's in the ECMAScript spec, which presumably means all browsers have it [21:29:55] still not seeing anything in the clicktracking table... it could just be that people don't use our toolbar that much [21:30:07] that's so sad... [21:32:33] nimish_g: Have you verified that the sampling thingy actually works? [21:33:29] yeah, I set it to like 50% on my local machine and would get clicktracking enabled about half the time [21:34:53] :) [21:34:58] want to crank it up farther and see what we get? [21:38:51] brion: I am all for it [21:39:45] 1:10? [21:39:51] i'm putting it at 1:1 [21:39:55] wooo [21:39:56] just to make sure things actually get stored :D [21:40:05] there we go! someone hid the help ;) [21:40:11] yay! [21:41:00] i suspect we may not need the throttle just yet as long as we're only working on a minority of a minority of edits :) [21:41:33] Apparently we are [21:43:28] a user with zero edits used the insert.gallery button =) this is fun [21:44:52] nkomura, should I write a blog post about it? (and relatedly, how do I go about doing that?) [22:01:02] nimish_g: first step, request a tech blog access to brion [22:01:33] step 2, write up a blog (brion insists a picture on blog) [22:01:48] step 3, send it for brion's review [22:01:53] step 4, publish [22:10:24] nimish_g, nkomura: click counts by event on enwiki so far :D http://dpaste.org/Bmma/ [22:12:06] awesome =) very very soon we'll have a very pretty web interface to visualize all this [22:12:19] nice [22:12:24] thanks brion! [22:13:52] whee [22:14:41] bah, roan left? [22:14:46] what was he going to ask me to do? [22:15:24] roan was asking if you can push the babacofixes [22:15:33] but was concerned that you may not have time today [22:19:10] ok we'll poke em later if they're not supervital [22:20:49] one text change was vital (trying to find a good argument why the text change is vital) [22:21:02] the description of dialogues was very confusing and we missed it out [22:22:35] got a reference? [22:22:54] TrevorParscal is pulling it up right now [22:23:43] http://svn.wikimedia.org/viewvc/mediawiki/trunk/extensions/UsabilityInitiative/EditToolbar/EditToolbar.i18n.php?view=log [22:23:52] the message changed in r57250 [22:24:13] but lots of localization, and some extra messages have been added since [22:24:58] well if it's a message change LocalisationUpdate should pick it up right? [22:25:16] or... no that won't pick up English changes [23:10:39] brion: can you give me access to the tech blog? I asked Jay, but remembered he's out of the office today [23:17:29] sure, sec [23:22:31] nimish_g: ok lemme pokoe it [23:24:22] nimish_g: ok i think it set you up [23:25:09] cool, thanks =) I think TrevorParscal can show me how to operate it [23:27:16] whee [23:27:25] pssh it's wysiwig! must be the most usable thing in the world [23:36:01] brion: both TrevorParscal and roan have bug fixes for babaco. whom shall we ask for review? we are fine waiting till next wednesday [23:36:32] but i'm afraid you will be slammed with meetings and requests from everyone [23:44:14] I could probably look over them if necessary [23:44:17] or you could poke tim [23:44:26] I have a feeling we need to delegate the review from Brion :P [23:44:54] nkomura: me or werdna for the moment would be fine :) [23:45:10] i don't have time to push anything today, i gotta run soon -- but happy to poke things later [23:45:11] thanks! [23:45:29] I'm not pushing stuff at 12:45, it'll have to wait a day or two [23:45:43] my weekends are chock until next year mostly [23:45:57] next week would be great [23:46:09] hehe [23:46:13] *nkomura weeps over brion's transition email [23:46:17] yeah i can definitely poke at it monday [23:46:18] awwwwww [23:46:22] I didn't see it [23:46:33] staff-l [23:46:34] ? [23:46:53] yeah [23:47:23]