Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2021-08-05
12:20:15 sean-k-mooney information included only the key value and key name.
12:21:46 bauzas gibi: sorry was out for lunch
12:22:05 sean-k-mooney songwenping: im surpised that affect https://github.com/openstack/cyborg/blob/1052efe93b5e7aa351b1f50cfe80f504dcf48b72/cyborg/db/sqlalchemy/api.py#L511-L516
12:22:15 bauzas gibi: about the VGPU healed allocations, well, we already verify the VGPU RC for the audit command but we don't do this for the heal_allocations
12:22:55 bauzas gibi: also, given we would now use other custom RCs, maybe we could also modify both the audit and heal_allocs commands to have a new attribute for telling which RCs to look at
12:25:15 songwenping sean-k-mooney: yes, the mysql version is update.
12:25:18 sean-k-mooney songwenping: without the oslo.db patch teh colum will be named "device_profiles.uniq_device_profiles0name" with it it will be "name"
12:25:20 sean-k-mooney oh i see
12:25:43 sean-k-mooney ok i get what is happening now so e.columns is a dict
12:26:01 sean-k-mooney if 'name' in e.columns: is checkking if there is a key that exactly match
12:26:36 sean-k-mooney the keey has chaned to now have the table name prefixed
12:26:54 sean-k-mooney so the fix you can do in cyborg is to add an elif
12:27:47 songwenping right, the e.columns isnot ['name'] any more.
12:29:34 sean-k-mooney ya so we just need to make the comparisone a little more robost
12:30:44 songwenping if i make the name and id all conflict, the e.columns is ['id'] and isnot ['id','name']
12:31:32 sean-k-mooney https://paste.opendev.org/show/807904/
12:31:47 sean-k-mooney i think this will work ^
12:32:33 songwenping this is good for now.
12:32:59 sean-k-mooney https://paste.opendev.org/show/807905/
12:33:03 sean-k-mooney or maybe that
12:33:38 sean-k-mooney add an else just in case we have a conflicat that is not on name or uuid although it would be treated as a uuid conflict today
12:34:36 sean-k-mooney songwenping: but ya the other way to do this is to preporcess the columns dict and stip the table prefix
12:35:57 songwenping does other projects have the same problems?
12:36:32 songwenping i see only cyborg distingush the uuid and name conflict.
12:36:41 sean-k-mooney i think we might define unique constraints differently then cyborg does
12:39:20 sean-k-mooney songwenping: so this is the other way to fix it https://paste.opendev.org/show/807908/
12:39:57 sean-k-mooney all that has changed here is instead of using e.colums directly i have generated a new columns dict and then the exsitng if else just uses that
12:39:59 songwenping this is same as the oslo.db does.
12:40:07 sean-k-mooney yep more or less
12:40:32 songwenping but whether it depends on mysql version
12:40:44 sean-k-mooney so on brances that cant use the new version fo oslo db you can backport that in cyborg
12:41:01 sean-k-mooney songwenping: this will work for any mysql verion
12:41:24 sean-k-mooney if you have old mysql it will be a noop as none of the column names will have 0 in them
12:41:35 songwenping ok, this is a good idea, thanks.
12:41:42 sean-k-mooney so columns and e.columns will be the same
12:42:03 songwenping right
12:42:26 sean-k-mooney that is proably the minimal change let me see quickly what is different between how nova defines unique constraints an cyborg
12:44:40 songwenping nova doesnot distingush the conflict types
12:46:02 sean-k-mooney ah ok that wold make sense then
12:46:42 sean-k-mooney i guess if any of them fail we dont realy care why we know the request is invlaid
12:47:27 songwenping yes, i also wonder if we should distinguish them
12:48:02 sean-k-mooney you proably do it today to have a better error message but you likely could do tha tdifferently
12:48:39 songwenping ack
12:49:54 sean-k-mooney oh https://github.com/openstack/nova/blob/35ddf1ad40207dee681a3c92cc9e86b061234edd/nova/db/sqlalchemy/api.py#L545-L550
12:50:18 sean-k-mooney so we do have that patteren
12:50:52 sean-k-mooney that would have changed form ServiceBinaryExists to ServiceTopicExists silently
12:52:03 songwenping so this also have problem
12:52:49 songwenping the tempest doesnot check the ServiceBinaryExists exception?
12:53:00 sean-k-mooney kind of becaue we use .get it wont fail
12:53:12 sean-k-mooney ya we likely dont have tempest coverage for this
12:53:23 sean-k-mooney although we should have and api funct tests for thsi
12:53:42 songwenping +1
12:53:47 gibi bauzas: OK, so you have plans to amend the audit support for MDEV. then that is really a good time to add some support for heal if possible.
12:54:19 bauzas gibi: well, I have around 12 hours for doing this until 3 weeks :p
12:54:27 songwenping i will commit one patch to coverage it.
12:54:34 bauzas 8 hours for tomorrow and 4 jours this afternoon :)
12:54:38 bauzas hours*
12:54:39 gibi bauzas: I'm not saying that you have to do it now. :D
12:54:56 gibi bauzas: it is totally ok to do it in Yoga
12:55:04 gibi bauzas: I just noted that we have a gap in heal
12:55:16 bauzas gibi: create a bug report and I'll own it
12:55:19 gibi I will make sure that it is now documented
12:55:27 gibi bauzas: OK I can file a bug too
12:55:30 bauzas or a blueprint maybe
12:55:31 gibi that is a good idea
12:55:45 gibi I will create a bug and later on we can transform that to a bp
12:55:50 bauzas this is a tech debt fix
12:55:51 gibi if we ever get to it
13:24:22 opendevreview Pierre Libeau proposed openstack/nova master: Reset state during service nova-compute init https://review.opendev.org/c/openstack/nova/+/803600
13:37:54 opendevreview Merged openstack/nova master: trivial: Cleanup a comment about a now removed libvirt version check https://review.opendev.org/c/openstack/nova/+/802617
13:51:14 gibi bauzas: filed the bug about heal_allocation https://bugs.launchpad.net/nova/+bug/1939020
13:51:58 bauzas gibi: and triaged
13:53:34 opendevreview Balazs Gibizer proposed openstack/nova master: Block servers wit vGPU and device profile in heal_allocations https://review.opendev.org/c/openstack/nova/+/803603
13:53:42 gibi bauzas: thanks, and here is the doc and block ^^
13:57:01 bauzas gibi: looking
15:08:27 melwitt bauzas: I'm gonna respond on the review too but the "unknown" is not a default consumer type, it's a cosmetic-only label for friendliness to the user. (same with the "all" label as a query parameter). a consumer with no type has NULL in the column. the reproposal of the spec has removed mention of the original proposed "UNKNOWN" official default type. during review reviewers preferred to leave NULL to represent the absence of a consumer
15:08:29 melwitt type
15:08:46 bauzas melwitt: oh ok
15:08:58 bauzas I only looked at the train spec
15:09:08 bauzas when this was something as a gerrit spec :p
15:09:09 melwitt ah gotcha
15:09:25 bauzas melwitt: will then remove my -1
15:09:41 melwitt cool thanks :)
15:10:08 bauzas (and yeah sorry I cheated :p )
15:11:13 melwitt haha no worry
15:14:08 lyarwood https://review.opendev.org/c/openstack/nova/+/803585 btw gibi / stephenfin / melwitt ; moving the LM jobs to NV for the time being while someone looks at the underlying QEMU failure(s)
15:14:48 melwitt /sad trombone sound
15:18:41 melwitt kashyap: I see you have commented on https://launchpad.net/bugs/1912310 before ^ but no leads yet right?
15:18:54 kashyap melwitt: Hi, /me looks
15:19:13 kashyap (The bot doesn't seem to higlight the summary on IRC anymore)
15:20:29 melwitt hm yeah not sure what happened to the bot
15:20:51 kashyap melwitt: Yeah, no leads yet
15:22:07 kashyap melwitt: Is it hitting more often? Looks like so from gibi's #7 on 31Jul
15:22:46 melwitt kashyap: yeah, lyarwood is proposing above making the live migration job non-voting because of it
15:23:03 kashyap melwitt: Oh, yeah - just saw that
15:55:48 bauzas lyarwood: unfortunately +Wd the live-mig job be non-votign
15:56:02 lyarwood ack thanks
16:06:35 gibi about the bug resolver bot I asked infra they said that https://meetings.opendev.org/irclogs/%23opendev/%23opendev.2021-08-03.log.html#t2021-08-03T16:31:32
16:11:16 melwitt thanks gibi
16:14:20 gibi so probably we lost the both the topic handling and the bug resolving capability of the bot during that transformatin

Earlier   Later