| opendevreview | melanie witt proposed openstack/nova master: WIP trait, filter, request spec https://review.opendev.org/c/openstack/nova/+/932153 | 00:20 |
|---|---|---|
| *** __ministry is now known as Guest5998 | 03:16 | |
| opendevreview | OpenStack Proposal Bot proposed openstack/nova master: Imported Translations from Zanata https://review.opendev.org/c/openstack/nova/+/929719 | 04:51 |
| sean-k-mooney | melwitt: have you tought about a new api to deiscover which iamge backend are supproted on a give cloud, also i woudl expect to want to be able to also set ad default image backed in the falvor or image | 12:03 |
| sean-k-mooney | doing it via the bdms is fine for advanced users but im not sure how good the ux is in general | 12:04 |
| sean-k-mooney | i have a few other tought on "nice to have" capablity that could be part of this work like adding "on_shared_storage=True/false" to server show or maping strage backends to placment aggreates to ensure we can schdule to the same storage pool when an isntnace is on shared storage | 12:07 |
| sean-k-mooney | im not sure if you have tought about any of those elements in https://review.opendev.org/q/topic:%22nova-imagebackend-poc%22 | 12:07 |
| *** ykarel_ is now known as ykarel | 13:04 | |
| *** haleyb_ is now known as haleyb | 13:07 | |
| dansmith | sean-k-mooney: yeah I don't love using BDMs for that either, but it's probably the vehicle we need for being able to express that we want to be root-on-shared and ephemeral-on-local | 13:57 |
| dansmith | I think probably the reason I don't like it is because I don't like BDMs in the first place | 13:57 |
| dansmith | they're just obscure and opaque and I have to look up how to use them every time I need to | 13:57 |
| sean-k-mooney | well having this in the bdm makes sense. havign that as the only wya ot express it is another matter | 13:57 |
| sean-k-mooney | i was more or less thinking of having the ablity to default it via the flavor or image and if you want diffent backends per disk then overried with bdm | 13:58 |
| sean-k-mooney | i.e. hw:root_image_backed=rbd hw:ephmeral_image_backend=lvm | 13:58 |
| sean-k-mooney | and if you need something more complex only then go to bdm | 13:59 |
| dansmith | yeah I read what you said.. and maybe, but the user really shouldn't even know that lvm or rbd are in use | 13:59 |
| sean-k-mooney | well it may be approaty if we have an api for listing the types. | 13:59 |
| sean-k-mooney | but without that i agree | 14:00 |
| sean-k-mooney | we could also have an abstration i.e. hw:root_image_backed=local|shared_storage | 14:00 |
| sean-k-mooney | that would be more cloudy | 14:00 |
| sean-k-mooney | and offten that asll you care about | 14:00 |
| dansmith | yeah, I think enumerated types (like volume types) is the thing we need.. shared/local isn't enough, IMHO | 14:01 |
| dansmith | because if someone wants local cheap slow file-on-disk, faster lvm, shared rbd, shared iscsi box, etc | 14:01 |
| jlejeune | hello, do you know where can I find the mysql/mariadb compatible versions depending on the openstack releases ? is there a compatibility matrix somewhere ? | 14:46 |
| jlejeune | because I have issue when I want to upgrade from bobcat to caracal, when 'nova-manage db sync' tries to create nova_cell0.share_mapping table on mariadb 11.5.2 | 14:50 |
| jlejeune | it seems to be related to the charset | 15:30 |
| jlejeune | the utf-8 charset has been removed on mariadb > 10.6 | 15:38 |
| sean-k-mooney | dansmith: ya i like the idea of having storage_classes as an opace api constrcut defiend by the admin that a user then just selct form like volume types | 15:40 |
| sean-k-mooney | jlejeune: for better or worse nova does not define the charater type to use in our schema | 15:40 |
| sean-k-mooney | but yes there were chagnes to the charset and utf-8 specifically there is a replacement you can use | 15:41 |
| sean-k-mooney | i think its utf8mb4 | 15:41 |
| jlejeune | indeed it's utf8mb3 | 15:42 |
| sean-k-mooney | utf8mb4 is the replacement | 15:42 |
| jlejeune | but the charset utf8 is hardcoded in migrations scripts | 15:42 |
| sean-k-mooney | i dont think its is in nova | 15:43 |
| sean-k-mooney | but it may be in your installer | 15:43 |
| sean-k-mooney | oh https://github.com/openstack/nova/blob/master/nova/db/main/migrations/versions/8f2f1571d55b_initial_version.py#L274-L275 | 15:44 |
| sean-k-mooney | that a bug | 15:44 |
| jlejeune | hé hé | 15:45 |
| sean-k-mooney | that should be utf8mb4 | 15:45 |
| jlejeune | :) | 15:45 |
| amorin | there are some latin1 as well | 15:46 |
| sean-k-mooney | i guess depending on your release there may be an alias for that | 15:46 |
| sean-k-mooney | yes so we had a discussion about if nova can move to perscifbeing a charset to use a long time ago and i tought we agrreed no | 15:46 |
| sean-k-mooney | this was incontext to maigin to a case sensitive coalation type | 15:47 |
| sean-k-mooney | which is a slictly diffent topic | 15:47 |
| sean-k-mooney | looking at https://dev.mysql.com/doc/refman/5.7/en/charset-charsets.html and 9.0 | 15:47 |
| sean-k-mooney | utf8mb4 shoudl be viable for a wide range of release | 15:48 |
| sean-k-mooney | utf8 exists in 5.7 but not in 8.0+ | 15:48 |
| sean-k-mooney | im not sure when it was removed | 15:48 |
| sean-k-mooney | jlejeune: can you file a but for this. | 15:49 |
| jlejeune | I can for sure, but it will wait for monday :) | 15:49 |
| jlejeune | I have to go | 15:50 |
| sean-k-mooney | no worries | 15:50 |
| jlejeune | see you | 15:50 |
| sean-k-mooney | o/ | 15:50 |
| sean-k-mooney | looks like we have 3 migration that are affected. 2 in the main/cell db an 1 in the api db | 15:51 |
| sean-k-mooney | https://dev.mysql.com/doc/refman/8.4/en/charset-unicode-utf8.html | 15:52 |
| sean-k-mooney | that is the deprecation notice | 15:52 |
| sean-k-mooney | so its an alias for utf8mb3 and it was remvoed in 8.0 | 15:52 |
| sean-k-mooney | utf8mb3 is also deprecated | 15:52 |
| sean-k-mooney | so if we should move to utf8mb4 for all tables if we are goign to specify one | 15:53 |
| sean-k-mooney | idealaly using utf8mb4_unicode_ci or utf8mb4_bin coralation type | 15:54 |
| sean-k-mooney | i would prefer utf8mb4_bin but there are some upgrade concerns with that | 15:54 |
| sean-k-mooney | bauzas: i know your done for the weekend but ^ might need to be a ptg topic again | 15:56 |
| sean-k-mooney | gibi: pci in placement still allow sriov networkign today right | 16:55 |
| sean-k-mooney | the neutron sriov vf are just not tracked in placment | 16:56 |
| sean-k-mooney | i tore down my home cloud a few weeks ago to do some hardware mantance and im debating if im going to redpeloy it at the weekend | 16:56 |
| sean-k-mooney | if i do im condiering going sriov only or mainly sriov for my netorking to keep ti simple. and fast | 16:57 |
| *** haleyb is now known as haleyb|out | 19:13 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!