[00:14:43] Aⅼⅼɑh ⅰs doіng [00:14:43] ѕᥙn iѕ nοt dοіng Ꭺllаh is doiᥒɡ [00:14:43] moon iѕ not ⅾoinɡ Allаh іѕ doіnɡ [02:20:16] Allah iѕ dⲟing [02:39:46] Αlⅼаh is doinɡ [05:17:10] 10Quarry, 10Patch-For-Review: Get rid of this ?download=true parameter - https://phabricator.wikimedia.org/T204226 (10zhuyifei1999) ^ Note this patch will continue to download if left-clicked, but you can override the download if opened the link in new tab (might be browser-dependent though) [05:37:19] Alⅼah іs doіnɡ [05:37:19] sun iѕ nഠt doinɡ Αlⅼah is ԁoіᥒɡ [05:37:19] mooᥒ is ᥒοt ԁഠіᥒg Ꭺllah ⅰѕ dοіᥒɡ [06:06:39] Αⅼlɑh іs ⅾοⅰᥒɡ [06:06:39] suᥒ is nοt dοiᥒɡ Аlⅼɑh is ԁοіng [06:06:39] mοоᥒ iѕ nഠt ԁoіᥒg Allah іѕ ⅾⲟing [07:54:26] Alⅼaһ ⅰs ԁoinɡ [10:10:25] Αllaһ is ⅾoіᥒg [10:10:26] suᥒ is ᥒot ԁoing Allah is dоⅰng [10:10:26] mοoᥒ is nⲟt doіng Αlⅼah iѕ doіnɡ [11:29:54] Аlⅼah іs ԁⲟіng [11:51:01] nuria: o/ I'm back ;) [12:31:23] Aⅼlaһ ⅰs dоⅰᥒg [13:31:02] Aⅼⅼah іѕ ⅾoing [13:41:07] bmansurov: hi, look at https://gerrit.wikimedia.org/r/#/c/mediawiki/extensions/EventLogging/+/458864/ and let me know [13:41:29] nuria: ok [14:24:48] hi nuria! So bmansurov is going to re-start the data collection for the citation usage schema shortly. We have split the original schema into 2 - one collecting pageloads, which we intend to sample at 50%; the other collecting interaction with citations, which we intend to samplea 100%; just double checking that these percentages look reasonable to you. [14:42:01] miriam: i would put both at 10% to start with so you can estimate how big traffic would be at 100% [14:43:25] miriam: the citation schema is heavily affected by bots crawling so it is likely traffic is quite big, did you guys looked at sampling? there is not technical reason why sample will not be random [14:44:40] nuria: sure; ok so: we run 1 week of citation usage only (no pageloads) in july, at 100%. we got 150 events/second [14:45:56] miriam: i would recomend to sample then [14:47:06] nuria: what was generating so many events the last time (~2000/sec) is the new pageload event; that is why we created a new schema just for this event, and we will sample at say 50% [14:47:15] nuria: or less if needed [14:50:02] nuria: the traffic captured by the citation usage schema in july (sampled at 100%) did not seem to create major issues, it is 1 order of magnitude lower compared to the pageload events. [14:50:43] miriam: sorry on meeting , can talk in couple hours. [14:50:44] nuria: so, if possible, we would like to keep 100% sampling for these rare events, and reduce to 50% of less the sampling for the much more frequent pageloads [14:50:50] nuria:sure, thanks! [14:57:41] Ꭺllɑh is doinɡ [15:19:58] https://www.mediawiki.org/wiki/Wikimedia_Technology/Annual_Plans/FY2019/TEC1:_Reliability,_Performance,_and_Maintenance/Goals [15:24:52] https://www.mediawiki.org/wiki/Wikimedia_Technology/Annual_Plans/FY2019/TEC14:_Smart_Tools_for_Better_Data/Goals [15:44:50] https://www.mediawiki.org/wiki/Wikimedia_Technology/Annual_Plans/FY2019/TEC2:_Modern_Event_Platform/Goals [15:45:15] miriam: back, sorry for links, wrong chat [15:45:30] nuria:no problem! [15:48:19] miriam: sorry, missunderstood earlier, 100% is 150 events per sec [15:48:24] miriam: that sounds np [15:48:34] miriam: correct? [15:48:44] cc bmansurov [15:48:48] nuria: yes, one schema at 100% is 150 events/sec [15:48:58] miriam: sounds good, no need to sample [15:49:13] nuria: the other, as we saw last week, reaches ~2500. This is the one we want to sample [15:49:29] miriam: understood, sorry for rushed message before [15:49:54] nuria: no prob! thanks!! [15:50:27] nuria: just to clarify, 1250 req/sec is ok? [15:51:30] bmansurov: miriam just mentioned 1 schema at 100% is 150 events per sec [15:51:39] bmansurov: not 1250 [15:51:48] bmansurov: did i missunderstood? [15:51:56] nuria: yes, the other schema is at 50% which will be around 1250 [15:52:12] nuria: yes, but the other one, the one with pageload, can get up to 2500/sec at 100% [15:52:17] bmansurov: i would downsample that a bit more [15:52:38] nuria: does around 800 look good? [15:53:01] bmansurov: that is the volume of some of the higest events now, yes, but i will slowly ramp it up [15:53:26] bmansurov: like start at 1% go to 10% and maximun (going with your numbers) is less than 20% right? [15:53:41] 33% will be around 800 [15:54:04] nuria: ok, should we start at 10% and monitor how many we get, and then calculate a percentage which will cap the number of events at 800/sec? [15:54:23] miriam: that sounds like a good initial plan [15:54:35] nuria: miriam: but we already got 2500 at 100%, why recalculate? [15:54:56] bmansurov: but it an't the exact same experiment right? [15:55:03] bmansurov: cause now you have two schemas [15:55:39] nuria: we're just sending the same events into two schemas now. We're not collecting more events [15:56:31] bmansurov: right, but since your code is slightly different i would rule out existance of bugs by slowly ramping up [16:23:54] Аⅼlaһ iѕ ԁⲟіnɡ [16:54:24] Ꭺlⅼah iѕ doing [16:54:24] s∪n is not doing Allaһ is doiᥒɡ [16:54:24] ⅿooᥒ іs nоt doing Allah іѕ doiᥒɡ [17:40:06] 10Quarry, 10Patch-For-Review: Quarry should update username when a user is renamed - https://phabricator.wikimedia.org/T73064 (10Framawiki) 05Open>03Resolved a:03zhuyifei1999 Tested after the merge of the above patch by # Creating a test account on WM and log in into quarry # [[ https://meta.wikimed... [18:26:37] Aⅼlaһ iѕ ⅾoing [18:58:06] Ꭺlⅼaһ ⅰѕ dоⅰnɡ [18:58:06] sᥙᥒ is not doing Aⅼlɑh is ⅾoіng [19:19:22] 10Quarry, 10Patch-For-Review: Quarry results serialization (web) should have its memory usage upper-bounded and streamed - https://phabricator.wikimedia.org/T192696 (10Framawiki) 05Open>03Resolved a:03zhuyifei1999 [19:28:09] 10Quarry, 10Patch-For-Review: TSV results are CSV, rows aren't separated with tabs - https://phabricator.wikimedia.org/T204432 (10Framawiki) 05Open>03Resolved a:03zhuyifei1999 [20:02:08] 10Quarry, 10Patch-For-Review: Get rid of this ?download=true parameter - https://phabricator.wikimedia.org/T204226 (10Framawiki) 05Open>03Resolved a:03zhuyifei1999 Patch merged and deployed, but the problem is not totally solved as said in previous comment. We haven't any idea/can't do anything to resolv... [20:08:15] 10Quarry, 10Patch-For-Review: HTTP header encoding error - https://phabricator.wikimedia.org/T204430 (10Framawiki) 05Open>03Resolved a:03zhuyifei1999 [20:18:15] Allаh is doіᥒg [20:18:17] suᥒ іs nഠt dοiᥒg Aⅼⅼah іѕ ԁoⅰnɡ [20:18:20] moon іs ᥒot ⅾoing Αlⅼаһ is ԁoinɡ [20:55:18] 10Quarry, 10Patch-For-Review: Get rid of this ?download=true parameter - https://phabricator.wikimedia.org/T204226 (10IKhitron) >>! In T204226#4591005, @Framawiki wrote: > Patch merged and deployed, but the problem is not totally solved as said in previous comment. We haven't any idea/can't do anything to reso... [21:36:50] Αlⅼɑh іs doіᥒg [22:01:01] Aⅼⅼɑh іs doing [22:01:01] suᥒ іѕ ᥒot doinɡ Aⅼlah іѕ ⅾοⅰᥒg [22:01:01] moഠn is ᥒоt ԁoinɡ Аⅼlаh іs ԁഠing [23:16:23] Aⅼlah іѕ ԁoіᥒg [23:25:09] Aⅼlɑh is dⲟⅰᥒg [23:26:27] Allah іѕ ⅾoiᥒg [23:26:27] s∪n is ᥒot ⅾഠing Αlⅼah is ԁഠinɡ [23:26:27] ⅿooᥒ is ᥒot ԁⲟiᥒg Alⅼaһ iѕ ԁoⅰng