[01:02:44] <Galahad>	 After I installing the StructuredDiscussions gives me the error: [Fri Jun 12 00:59:40.816789 2020] [php7:error] [pid 13872] [client xxx:19106] PHP Fatal error:  Class 'Pimple\\Container' not found in /var/www/wikisp/extensions/Flow/includes/Container.php on line 5,
[01:12:29] <Reedy>	 Galahad: You need to install dependancies using composer
[01:31:15] <Galahad>	 Reedy, done, but after I login shown this error: Fatal exception of type "Wikimedia\Rdbms\DBQueryError"
[01:31:23] <Galahad>	 run update.php?
[01:31:46] <Reedy>	 yup, flow adds database tables
[01:35:49] <Galahad>	 If I put images with www-data solves the problem at the moment of aprove accounts?
[01:36:28] <Reedy>	 probably
[02:45:57] <Krinkle>	 Looks like Parsoid should/could do with using PSR-4
[02:45:58] <Krinkle>	 https://gerrit.wikimedia.org/r/#/c/mediawiki/vendor/+/603571/1/composer/autoload_classmap.php
[02:47:43] <Krinkle>	 or maybe composer generates this regardless?
[02:47:50] <Krinkle>	 Yeah, stuff that's using psr4 is also expanded here. Interesting
[02:49:24] <Krinkle>	 https://gerrit.wikimedia.org/r/plugins/gitiles/mediawiki/vendor/+/refs/changes/71/603571/1/composer/autoload_real.php#33
[02:49:37] <Krinkle>	 and this suggests the static is what we use on PHP 7
[02:55:09] <Krinkle>	 Ah, right, PSR4 is allowed to overlap, so it needs run-time disk I/O (stat) to know whether something exists before trying another, that's something we'd want to avoid indeed. 
[02:55:14] <Krinkle>	 So this just goes straight into opcache. Cool
[14:02:23] <justinl>	 Hey all, I have a question regarding PHP-FPM as I'm using it with MediaWiki and a recent performance change I made to it, so while this isn't a MW-specific question, I hope it's reasonable enough to ask here. 
[14:04:59] <justinl>	 Yesterday morning, based on a convincing argument on a blog post, I changed my PHP-FPM from dynamic to static and set its max_children to 32 (@ 256 MB max each = ~8 GB) and max_requests to 10000. This is working fine, and it's load-balanced so well that, given our PHP request rate, all of the children hit their 10k max within a couple of minutes
[14:05:00] <justinl>	 after running for several hours (depending on time of day).
[14:05:26] <justinl>	 Here's a graph of our 4 web servers over the last 2 days. https://imgur.com/HNnYOXw
[14:06:07] <justinl>	 The middle set of lines is the total memory used by PHP-FPM processes. The periodic dips are the PHP processes all restarting at about the exact same time on each of the 4 servers.
[14:07:29] <justinl>	 Just trying to understand if this could be a problem from a caching or other perspective that I may not be considering. My opcache hit rates are over 99.9% consistently, so I don't see an issue there, but this complete restart at the same time just got me wondering if there's anything I'm not considering, or even if it's a concern at all.