Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2023-04-06
17:15:10 sean-k-mooney *gate bug
17:15:26 dansmith no, I haven't noticed it before this, I just went looking to see if it was really new here
17:15:41 dansmith I'm not sure why we'd see it for scheduler and not conductor, which already added the singleton infra
17:15:44 sean-k-mooney so i think i have seen this before but not consitently
17:16:07 sean-k-mooney well the base class is also alredy resting the singelton
17:16:17 sean-k-mooney all you changed is doing it via the property
17:16:28 sean-k-mooney which is even more strange
17:16:48 sean-k-mooney like i dont know why calling the singelton later would chagne the behavior
17:16:52 dansmith ...right
17:19:44 sean-k-mooney so the hits you found seam to be for 878763, 875621, 879500 and now 878238
17:30:00 sean-k-mooney dansmith: so other then the fact that two of these were submited by you id dont see anything quicly that woudl make me think this is specific to your change
17:30:19 sean-k-mooney so i think this is an exting issue so we can proceed with the lazy loading
17:31:02 sean-k-mooney the first hit in the lst 30 days was Mar 28, 2023 @ 12:58:01.786
17:31:11 sean-k-mooney so maybe we could see what merged just before that
17:31:49 dansmith but we're also not hitting them very often, so I'm not sure looking at temporal merges is likely to be very fruitful
17:32:02 opendevreview Dan Smith proposed openstack/nova master: Add compute_id to Instance object https://review.opendev.org/c/openstack/nova/+/879500
17:32:34 dansmith sean-k-mooney: based on this I'm assuming you didn't see my ping about it above, which means you likely didn't see my ping before that about the spec?
17:32:34 sean-k-mooney fair i guess we can just keep an eye out and see if this starts happening more often
17:33:11 sean-k-mooney correct but scrolling up i see them
17:33:22 dansmith no rush, so don't worry about it today
17:33:37 dansmith just fyi for tomorrow or next week
17:35:09 sean-k-mooney ya proably next week
17:35:24 sean-k-mooney skiming it i did not think about migration being affect by this durign the ptg
17:35:39 dansmith me either, but we update host and node from migration a LOT
17:35:39 sean-k-mooney but ya it would make sesne if we are addign the srive_id to the instace
17:36:04 sean-k-mooney before the ptg i was not expectign to update the instnace in this cycle
17:36:16 dansmith my patch actually makes us do that less, updating it to be "this node here" instead of "whatever the migration said" in several places
17:36:23 sean-k-mooney i assumed it was just going ot be the compute but there is value in doing both
17:36:39 sean-k-mooney ack
17:37:18 sean-k-mooney ya so jira has eaten all my brin power so im not digesting the spec reading it
17:37:34 sean-k-mooney so ya ill take a look on tueday and just assume your right :)
17:38:01 sean-k-mooney i did see you had started the work on it
17:38:04 dansmith yep, np
17:38:21 dansmith as noted in the ptg,
17:38:36 dansmith I think we should do more developing the code and spec in parallel, after initial sign-off on the idea
17:38:45 dansmith so I'm trying to keep my spec updated with the wrinkles I'm finding
17:38:59 dansmith and just hope that bauzas will approve me before the spec lands :)
17:42:20 sean-k-mooney ill reply on artom patch
17:42:57 sean-k-mooney but if i recall corectly its because this is invoked in the perodic task asn we are trying to avoid loading all of the instnace numa toplogy blob form the data base
17:43:28 dansmith but we're iterating it already no?
17:43:30 sean-k-mooney this is invoked in the update_avaiable_resouces periodic
17:43:51 dansmith it seems like we're just creating a new topology by iterating the old one, then doing the math
17:44:01 dansmith like, instead of obj_clone() (if we need to be doing it on a copy)
17:45:04 sean-k-mooney i also dont have this fully loaded in cache but i think we are actully iterating over the domains intially rather then having a fully set of instance objects
17:45:30 dansmith we're iterating over host topology, AFAICT
17:45:41 dansmith but not important right now, I was just asking
17:45:45 dansmith (as noted in my reply)
17:47:05 sean-k-mooney so im not saying this cant be recactored but if i recall correctly its becasue of this https://github.com/openstack/nova/blob/49aa40394a4857a06191b05ea3b15913f328a8d0/nova/compute/resource_tracker.py#L1334-L1371 code path
17:48:33 dansmith right, it's making a copy, modifying it and returning
17:48:52 dansmith obj_clone() is the way to do that to avoid silly omissions like this bug
17:49:22 sean-k-mooney ya
17:49:25 dansmith anyway, as I said, I'm not saying it needs to be refactored right now, I was just wondering why we were doing it the most laborious and error-prone way possible
17:55:47 sean-k-mooney so this comes form the initall commit it seams
17:55:49 sean-k-mooney https://github.com/openstack/nova/commit/b11dbfa4902cdd74bad3745db177d80b1c8b07c6
17:57:32 sean-k-mooney https://github.com/openstack/nova/blob/b11dbfa4902cdd74bad3745db177d80b1c8b07c6/nova/virt/hardware.py#L1121-L1151
17:58:14 sean-k-mooney that predates OVOs being split out of nova i think i wonder if it just the clone did not exist then
17:59:08 dansmith no, it did
17:59:49 dansmith perhaps just an omission
17:59:58 sean-k-mooney https://github.com/openstack/nova/blob/b11dbfa4902cdd74bad3745db177d80b1c8b07c6/nova/objects/base.py#L370-L372
18:00:04 sean-k-mooney so ya it did
18:00:11 sean-k-mooney and since it worked we just didnt clean it up
18:00:38 sean-k-mooney anyway i woudl prefer to proceed with the small fix and maybe swap to that on master in a followup
18:01:05 sean-k-mooney we would ahve to see if there are any field that should not be cloned
18:01:27 sean-k-mooney liek we want to reset the usage filed
18:01:28 dansmith ...that's what I said
18:02:11 sean-k-mooney oh i know im just confirming that looking at the git history its not obvious why that approch was nto taken orginally
18:02:34 dansmith okay
18:03:44 sean-k-mooney ok im going to finish for today. if i dont get to your spec on tuesday feel free to remind me
18:03:51 sean-k-mooney enjoy your weekend o/
18:04:05 opendevreview Merged openstack/nova stable/xena: Add debug log for scheduler weight calculation https://review.opendev.org/c/openstack/nova/+/879404
18:04:18 dansmith ack, thanks, later
#openstack-nova - 2023-04-07
07:39:28 Uggla bauzas, can we merge https://review.opendev.org/c/openstack/nova/+/868089, as all fixes from Amit have been merged ?
07:40:16 bauzas Uggla: I can take a look
07:40:58 Uggla bauzas, sure last time you told me that it was ok. But you would like to merge Amit patches first.
07:41:31 bauzas -ETOOMANYBUGFIXESINMYBRAIN
07:43:21 bauzas I wish I could just change some internal SDcard
07:43:32 Uggla bauzas, no hurries. I just want to not forget about it.
07:44:10 Uggla I have another one for review but I will submit it to you next week. I don't want to ruin your Friday. ;)
07:44:48 bauzas well
07:44:55 bauzas my Friday is just a regular Friday
07:44:58 bauzas paperworks and so.
07:45:15 bauzas Gerrit was updated, yet again UI \o/
08:49:37 kashyap bauzas: "Bugs increase to fill the available time" ;-)
08:50:14 kashyap Original quote: "Work expands to fill the available time"
11:12:30 opendevreview Alexey Stupnikov proposed openstack/nova stable/victoria: Cleanup old resize instances dir before resize https://review.opendev.org/c/openstack/nova/+/864730
11:15:37 opendevreview Balazs Gibizer proposed openstack/nova master: Run the lower constraints jobs as periodics https://review.opendev.org/c/openstack/nova/+/879886
11:15:37 opendevreview Balazs Gibizer proposed openstack/nova master: Add lower-* unit and functional jobs https://review.opendev.org/c/openstack/nova/+/879885
11:15:38 opendevreview Balazs Gibizer proposed openstack/nova master: DNM: show that lower-* jobs has no false positives https://review.opendev.org/c/openstack/nova/+/879888
11:15:38 opendevreview Balazs Gibizer proposed openstack/nova master: DNM: example bug that caught by lower-* jobs https://review.opendev.org/c/openstack/nova/+/879887
12:03:45 songwenping_ bauzas, sean-k-mooney: hi guys, i have attach gpu to the vm with 'virsh attach-device --persistent --live' , i can find the gpu in the vm with lspci, but nvidia-smi cnanot find the gpu, with the error 'NVRM: GPU 0000:05:00.0: RmInitAdapter failed! ', is there any advices?
12:46:03 dansmith kashyap: I assume you saw this that tobias-urdin cc'd you on? https://review.opendev.org/c/openstack/nova/+/879021
13:08:48 kashyap dansmith: Looking; no, I didn't see it. Thanks for notifying me :)
13:09:31 kashyap Okay, yes, it is indeed a long-term TODO that I left to remove it that Tobias has done, nice -- https://review.opendev.org/c/openstack/nova/+/879021/2/nova/virt/libvirt/driver.py
13:21:34 kashyap tobias-urdin: Thanks for the above code! It looks good to me, I just like a double-check from another reviewer about the code-removal in this method, check_can_live_migrate_source()
15:08:05 opendevreview Balazs Gibizer proposed openstack/nova master: DNM: show that lower-* jobs has no false positives https://review.opendev.org/c/openstack/nova/+/879888
15:08:05 opendevreview Balazs Gibizer proposed openstack/nova master: DNM: example bug that caught by lower-* jobs https://review.opendev.org/c/openstack/nova/+/879887
15:57:31 opendevreview Balazs Gibizer proposed openstack/nova master: Add lower-* unit and functional jobs https://review.opendev.org/c/openstack/nova/+/879885
15:57:32 opendevreview Balazs Gibizer proposed openstack/nova master: DNM: example bug that caught by lower-* jobs https://review.opendev.org/c/openstack/nova/+/879887
15:57:32 opendevreview Balazs Gibizer proposed openstack/nova master: Run the lower constraints jobs as periodics https://review.opendev.org/c/openstack/nova/+/879886
15:57:33 opendevreview Balazs Gibizer proposed openstack/nova master: DNM: show that lower-* jobs has no false positives https://review.opendev.org/c/openstack/nova/+/879888

Earlier   Later