[07:57:13] <Krinkle>	 Amir1: regarding data loss, any common factor yet, or does it seem random / evenly distributed? ref T364245
[07:57:13] <stashbot>	 T364245: Recentchanges and cu_changes tables are occasionally missing revisions on multiple wikis - https://phabricator.wikimedia.org/T364245
[07:57:56] <Krinkle>	 Looking at nlwiki, it seems evenly distributed, no obvious common cause yet. If truely no common factor, it would imply some kind of race condition, timeout, pod shutdown etc
[07:59:15] <Krinkle>	 https://quarry.wmcloud.org/query/100476
[08:02:04] <Krinkle>	 by page_namespace; https://phabricator.wikimedia.org/P86760
[08:04:02] <Krinkle>	 Main 0: 5%, Talk 1: 20%, User 2: 4%, User_talk 3: 25%
[08:05:16] <Krinkle>	 Project 4: 2%, Template 10: 10%
[08:05:53] <Krinkle>	 Category 14: 42%
[08:08:45] <Krinkle>	 looking at enwiktionary for comparison
[08:09:37] <Krinkle>	 Category 99% loss
[08:10:01] <Krinkle>	 Looks like the most affected are probably also logged-in or more advanced users
[08:13:21] <Krinkle>	 no..
[08:13:28] <Krinkle>	 but logged-in users are more likely to create pages
[08:13:36] <Krinkle>	 checking rev_parent_id
[08:14:09] <Krinkle>	 yep
[08:14:13] <Krinkle>	 100% have rev_parent_id=0
[08:14:23] <Krinkle>	 or close to 100% given the problem is growing in real time
[08:24:54] <Krinkle>	 Hm..
[08:25:03] <Krinkle>	 Isn't mw.new for page creation?
[08:42:45] <Krinkle>	 yeah, so the problem isn't 5% globally, much less
[08:42:50] <Krinkle>	 it's more like 2%, still huyge
[20:35:03] <Amir1>	 Krinkle: thanks for investigating and catch rc.new. I agree it still needs fixing. I was traveling and haven't looked too deep. I will take a deeper look soonTM