Wednesday, 2022-05-04

opendevreviewRico Lin proposed openstack/cinder master: Add disable_image_conversion config
ricolinrosmaita: jbernard 04:22
ricolinthanks  for the review04:23
ricolinjust update the patch accordingly04:23
ricolinlet me know if that update works for you, thanks04:24
opendevreviewTobias Urdin proposed openstack/cinder master: backup/swift: Add support sending service user token
*** dviroel|out is now known as dviroel11:20
opendevreviewArun KV proposed openstack/cinder master: Reintroduce DataCore driver
whoami-rajatCinder meeting in #openstack-meeting-alt at 1400 UTC13:59
whoami-rajatjungleboyj rosmaita smcginnis tosky whoami-rajat m5z e0ne geguileo eharney walshh_ jbernard sfernand enriquetaso hemna fabiooliveira yuval13:59
rosmaitawhoami-rajat: ty13:59
geguileowhoami-rajat: thanks!13:59
opendevreviewEric Harney proposed openstack/cinder master: Groups: remove unneeded "status" variable
simondodsleyCan I get a +W on please?14:09
opendevreviewSimon Dodsley proposed openstack/cinder stable/xena: PureStorage FlashArray: Add active/active replication
enriquetaso#startmeeting cinder_bs15:01
opendevmeetMeeting started Wed May  4 15:01:31 2022 UTC and is due to finish in 60 minutes.  The chair is enriquetaso. Information about MeetBot at
opendevmeetUseful Commands: #action #agreed #help #info #idea #link #topic #startvote.15:01
opendevmeetThe meeting name has been set to 'cinder_bs'15:01
enriquetasoHello, 5 new bugs were reported this period. 15:01
enriquetaso#topic  Temporary volume accepts deletion while it is used 15:02
enriquetasoA temporary volume is created when an attached volume is backing up. This temporary volume can be deleted by DELETE API because its status is 'available'.15:02
enriquetasoThe fix is already merged on master. 15:02
enriquetasoMoving on with a related bug15:02
enriquetaso#topic Temporary volume could be deleted with force15:02
enriquetasoFix proposed to master15:02
enriquetaso#link 15:03
enriquetasoMoving on15:06
enriquetaso#topic  Volume reset-state API validation state checking is incorrect 15:06
enriquetasoeharney, Code from I0a53dfee "Reset state robustification for volume os-reset_status" aims to reject volume state updates to "error_deleting" and "detaching" but fails to do so due to a typo.15:06
enriquetasoOriginal fix:15:07
enriquetasoFix proposed to master:15:07
eharneyyeah i'm going to rework the patch for this based on rosmaita's input15:07
eharney773985 isn't an original fix, it's where the bug was introduced15:07
enriquetasooh, my bad15:07
enriquetasosounds good, please cinder team review :)15:08
enriquetasomoving on15:08
enriquetaso#topic rbd_store_chunk_size in megabytes is an unwanted limitation15:08
enriquetasoThe report requests some changes regarding rbd_store_chunk_size. The reporter proposed two alternatives. 15:09
enriquetasoI need help to understand if this makes sense or if I should ask more questions about the problem. 15:09
eharneywell there are restrictions on what we can set it to, since glance and cinder in many configurations need to agree on the chunk size15:09
eharneywe should look into this area, there are some deficiencies we still need to address w/ RBD as far as sector sizes too  (512 vs 4k)15:10
eharneybut i don't think the suggestion to just add a new config value that lets deployers set whatever is necessarily the right answer15:10
eharneyas usual, we don't have much concrete performance data to analyze, so more info on the actual problem would be helpful15:11
enriquetasoOK, so this bug is not trivial then. 15:12
enriquetasoI'll ask for concrete data and point out all this 15:13
eharneyright, it's a good idea to improve this, but it needs a lot of thought15:13
enriquetasomakes sense 15:13
rosmaitai think his question #2 is a good one15:14
eharneyit is15:14
rosmaitaall i can think is it's leftover from the days before there were dedicated pools?15:15
eharneywell, specifying the chunk size helps prevent situations where you end up with images that can't be moved between pools during migration, cinder<->glance, etc15:17
eharneybut i don't have a nice clear answer at the moment15:17
*** dviroel is now known as dviroel|lunch15:18
enriquetasoOK, so this looks like something interested to look for in the future. I can bring it back later, as Eric pointed out, it needs more thought.  15:20
enriquetasoLast but not least15:20
enriquetaso#topic Could not find any paths for the volume15:20
enriquetasoIn an environment using NVMeoF with SPDK, when an instance is shut off or hard restarted, it is not able to find the volume again. The volume is visible on the node with "nvme list" but nova reports: "Could not find any paths for the volume."15:21
enriquetasogeguileo, I'm not familiar with NVMeoF. The reporter mentioned that this could be an os-brick problem. I think the bug report makes sense. 15:21
geguileoit's an os-brick problem15:21
geguileoit's already reported, and I will be fixing it shortly15:22
enriquetaso\o/ 15:22
geguileo(there's a patch proposed now, but I'm making changes)15:22
enriquetasosure, thanks Gorka15:22
geguileoI'm saying it without looking at the bug report15:22
geguileoI mean, it can be one of 2 cases...15:23
geguileothat I'm fixing...15:23
geguileolet's not forget that I'm fixing close to 12 nvmeof bugs...15:23
geguileothis is most likely the one where you cannot call os-brick with an already existing subsystem15:23
geguileomixed with the not disconnecting a subsystem on volume_disconnect15:24
eharneywe should probably just reevaluate any nvmeof bugs after geguileo's current stuff lands15:24
geguileoif they wait for 10 minutes it probably works   lol15:24
enriquetasoof course, well, i can bring it back later when you update the patches, sounds like the right thing to do15:24
geguileobecause the nvmeof timeout will kick in and remove it...15:24
rosmaitathose bugs are great, by the time you get help from support, there is no problem15:24
enriquetasoOK, that's all i have for today's meeting15:25
rosmaitai have 2 quick questons15:25
rosmaitai came across this when i was looking into eharney's patch that he mentioned earlier15:26
rosmaitai didn't file a bug, but wonder if i should15:26
eharneydoes it have any observable behavior impacts?15:27
rosmaitanot now, my concern is that if someone actually uses this and then backports the usage, it will break without this patch15:27
rosmaita(does that make sense?)15:27
eharneyi think so, but i wouldn't be inclined to file a bug for it myself15:28
rosmaitaok, works for me15:28
enriquetasono bug then 15:28
rosmaitaok, my other issue is test-only, so i won't file a bug for it either15:29
rosmaitaok, sorry i wasted our time15:29
enriquetasodon't worry 15:29
enriquetasoOK, running out of time15:30
opendevmeetMeeting ended Wed May  4 15:30:15 2022 UTC.  Information about MeetBot at . (v 0.1.4)15:30
opendevmeetMinutes (text):
rosmaitathanks enriquetaso!15:30
eharney   was +Workflow'd by rosmaita, whoami-rajat__, and jungleboyj in April 2021.  maybe it could use another look?15:31
rosmaitaoh no15:33
rosmaitawhat did we do?15:33
eharneyit went into merge conflict or something and then just got lost15:33
rosmaitaoh, i thought we let something bad slip into the code base15:33
eharneyrosmaita: i'm almost certain that this new test was written based on a discussion we had in here (also needs review):
rosmaitai would've sworn that i had reviewed that, but i guess not15:42
*** dviroel|lunch is now known as dviroel16:15
opendevreviewPrajakta Swapnil Belapurkar proposed openstack/cinder master: Volume Deletion rejection message  does not include "awaiting-transfer" in invalid state
opendevreviewPrajakta Swapnil Belapurkar proposed openstack/cinder master: Volume Deletion rejection message  does not include "awaiting-transfer" in invalid volume state
opendevreviewPrajakta Swapnil Belapurkar proposed openstack/cinder master: Volume Deletion rejection message  does not include "awaiting-transfer" in invalid volume state
opendevreviewBrian Rosmaita proposed openstack/cinder master: Remove use of mock CONF object
opendevreviewRico Lin proposed openstack/cinder master: Add image_conversion_disable config
opendevreviewRico Lin proposed openstack/cinder master: Add image_conversion_disable config
ricolinrosmaita: ^^^ updated and depends on your patch, thanks for the mock fix :)16:45
opendevreviewDavid White proposed openstack/cinder master: Seagate/Lenovo drivers: Update get_driver_options
opendevreviewPrajakta Swapnil Belapurkar proposed openstack/cinder master: Volume Deletion rejection message  does not include "awaiting-transfer" in invalid volume state
opendevreviewMerged openstack/cinder stable/yoga: Handle the case when tempest fails
opendevreviewMerged openstack/cinder master: mypy:
*** whoami-rajat__ is now known as whoami-rajat17:25
hemnasorry to interrupt the meeting earlier 17:53
hemnabut we should start thinking about how to add a volume action transaction tracking in cinder17:54
hemnaright now, there is really no way to see if cinder is 'busy' working on a volume or not, which is problematic for bouncing the service.17:54
hemnaalso would be helpful in debugging stuck volumes17:54
opendevreviewGorka Eguileor proposed openstack/cinder master: TOX: Document install_command usage
hemnaI'd be able to write another script to do a 'top' like action against cinder to have it show the live actions being taken against volumes17:55
geguileorosmaita: whoami-rajat__ ^ as agreed in todays bug scrub meeting17:55
geguileohemna: we have it for some operations (those that are cleanable)17:56
geguileoin the workers table17:56
geguileobut I'm not sure if that's the right table for what you want to do17:56
hemnaworkers table?17:58
hemnahrmm my workers table is basically empty18:00
hemnathere are a few creating and deleting rows, but that's it.18:03
hemnawhat is the workers table used for?18:03
rosmaitageguileo: ty18:04
geguileohemna: the workers table is used to track ongoing operations that are cleanable18:04
geguileoit's used because if cinder-volume is deployed in a cluster (active-active) a different node may need to clean up the operation18:05
hemnaok, so it's kinda useful18:05
hemnabut in my case not really.18:05
geguileoso while the operation is ongoing it is kept there saying which node is currently working on it (first the API, then the scheduler, finally the volume)18:05
geguileoyou cannot see the history, but you can see those ongoing operations18:06
opendevreviewMerged openstack/cinder stable/xena: Handle the case when tempest fails
opendevreviewMerged openstack/cinder stable/wallaby: Handle the case when tempest fails
opendevreviewMerged openstack/cinder stable/victoria: Handle the case when tempest fails
opendevreviewWalt proposed openstack/cinder master: Fix tracking allocated_capacity_gb
hemnafound another problem with tracking allocated_capacity_gb20:41
hemnafor volume migration from one host to another that's running in a separate c-vol service.20:41
hemnarequired adding a new rpcapi entry to call the remote host to update it's allocated_capacity_gb, since the source host gets the migrate_volume call.  the destination host was never notified and hence had out of date allocated_capacity_gb20:42
hemna has a fix for it20:42
opendevreviewMerged openstack/cinder-tempest-plugin master: Replace the CentOS Stream 8 lvm job with a CS9 one
*** dviroel is now known as dviroel|out21:50

Generated by 2.17.3 by Marius Gedminas - find it at!