Friday, 2025-09-05

*** ralonsoh__ is now known as ralonsoh13:36
fungigouthamr: heads up that https://lists.openstack.org/archives/list/openstack-discuss@lists.openstack.org/message/VR7XUI7V7QGFNHKFOXLFJJP6P4GGKOM3/ is requesting an addition to the agenda for the next tc meeting17:18
fungithe poster didn't include [tc] in the subject so just making sure you don't miss it17:18
gouthamrah thanks for the heads up fungi 17:19
fungiany time17:19
gouthamrit was on the next week's agenda - noonedeadpunk brought it up during open discussion last week17:19
* gouthamr should hit save on the agenda edit!17:20
fungioh, right i recall seeing something related to it in the notes. so i guess it's worth a reply that the tc is already aware and actively discussing options17:20
clarkbthat email talks a lot about utf8mb3. IMO part of any changes should involve moving to utf8mb4 if there is the possibility for 4 byte utf8 characters19:51
fungiit claimed ironic could not, due to (unspecified) internal design choices19:55
clarkbfungi: if I had to guess that is probably due to ironic having indexes on text columns and mariadb/mysql limit the total byte length of indexes to something like 76819:56
JayFIronic already made a fix19:56
JayFthat sorta was around "make sure we do the right thing no matter what the default"19:57
clarkbI just don't want us to do a big effort to change the collation then 6 months later realize we need to change the type too19:57
fungii mean it was claimed ironic could not use 4-byte utf8 in its databases19:57
clarkbit is likely to be less headache to do it together19:57
funginot that ironic was unable to solve the general mariadb default change challenges19:58
fungibut why ironic is incompatible with 4-byte character encoding (if that's even true), i have no idea19:59
clarkbya my hunch is the index byte width limit19:59
clarkbwe've hit that with other systems migratiing to utf8mb4 and so far we've been able to update the indexes too in order to address it20:00
JayFhttps://docs.openstack.org/ironic/latest/install/install.html#set-up-the-database-for-bare-metal20:00
JayFFirst time I've ever read that lol20:00
fungiheh20:00
clarkbhttps://dev.mysql.com/doc/refman/8.4/en/innodb-limits.html20:02
clarkbapparently its 767 or 3072 bytes depending on the row format20:02
JayFhttps://opendev.org/openstack/ironic/commit/1435a15ce3013da0a3138e1c5ab6ac523c382bb820:03
JayFhas a bit of detail20:03
fungineat20:03
clarkbDYNAMIC is the default row format as of mysql 5.7.9 so 3072 is probably common20:03
clarkbI don't undersatnd that commit message. The field width is independent of byte size20:04
JayFclarkb: I just found the commit, I barely remembered it existing :) 20:05
clarkbI suspect that varchar(255) * 3 = 765 < 767 and its actually the index limit I'm talking about not a limit to less than 255 4 byte chars20:05
clarkbyou can varchar(255) all you want with utf8mb420:05
clarkbyou can't put an index on that column if it is utf8mb4 and you are not using dynamic rows on innodb20:05

Generated by irclog2html.py 4.0.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!