[05:54:12] <marostegui>	 arnaudb: can you silence db2194's prometheus alerts? it's been alerting for a while about  wmf_auto_restart_prometheus-mysqld-exporter@s6.service Failed as it is not provisioned yet
[08:20:25] <arnaudb>	 yes!
[08:20:34] <marostegui>	 thanks <3
[10:10:49] <btullis>	 I have a quick question; do we use compression in MariaDB anywhere in production? i.e. Page compression, compressed row format, native file-system compression? Thanks.
[10:17:07] <marostegui>	 btullis: we use compressed row format in some places
[10:17:40] <marostegui>	 Although there are bugs with row=compressed and innodb buffers, so we've been not doing it much lately
[10:17:52] <marostegui>	 as we have plenty of space on most places
[10:21:21] <btullis>	 marostegui: Thanks. I would like to propose using ZFS with compression for the clouddb1021 refresh, but I'm conscious that I might get flamed. :-)
[10:21:49] <marostegui>	 btullis: up to you, we don't handle that host. But making a host special isn't something I am happy with. But again, we don't own clouddb*
[10:22:30] <marostegui>	 btullis: I know where this is coming from, and I think a better solution is to prepare the service to be able to scale horizontally, but again, not my call
[10:31:10] <btullis>	 marostegui: OK, well I certainly don't like making you unhappy, but I've had nothing but good experience with ZFS/MySQL in the past. My feeling is that this could still be a good use case, but I wanted to find out if there was any technical reason why it wouldn't be. Thanks again, I'll give it some thought.
[10:31:35] <marostegui>	 btullis: As long as you are responsible for that snowflake, I don't have any objections
[10:31:46] <btullis>	 Understood, thanks <3
[12:13:22] <Amir1>	 And generally reads on the host should be low so decompression should be less needed or intensive. OTOH, it's getting db writes from every wiki and compression might spike the CPU. Keep an eye on that.
[13:23:58] <btullis>	 Thanks Amir. As it happened, I didn't go for the ZFS recommendation, mainly due to licensing and out-of-tree issues. Briefly researched whether btrfs is any good for database workloads yet, but it isn't so I gues we will still with uncompressed xfs.
[13:24:15] <btullis>	 *skick*