15:06:24 <enriquetaso> #startmeeting cinder_bs
15:06:25 <openstack> Meeting started Wed Mar 17 15:06:24 2021 UTC and is due to finish in 60 minutes.  The chair is enriquetaso. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:06:26 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:06:28 <openstack> The meeting name has been set to 'cinder_bs'
15:06:33 <enriquetaso> #topic roll call
15:06:38 <enriquetaso> hi :P
15:06:38 <kinpaa12389> hi
15:06:40 <rosmaita> o/
15:06:56 <enriquetaso> Etherpad:
15:06:56 <enriquetaso> #link https://etherpad.opendev.org/p/cinder-bug-squad-meeting
15:07:20 <enriquetaso> Full list of bugs:
15:07:22 <enriquetaso> #link http://lists.openstack.org/pipermail/openstack-discuss/2021-March/021135.html
15:07:56 <enriquetaso> #topic bug_1:"SADeprecationWarning: The joinedload_all() function is deprecated, and will be removed in a future release. Please use method chaining with joinedload() instead"
15:08:02 <enriquetaso> #link https://bugs.launchpad.net/cinder/+bug/1832164
15:08:04 <openstack> Launchpad bug 1832164 in Cinder "SADeprecationWarning: The joinedload_all() function is deprecated, and will be removed in a future release. Please use method chaining with joinedload() instead" [Critical,In progress] - Assigned to Gorka Eguileor (gorka)
15:08:21 <enriquetaso> wanted to name the bug because it is in critical condition but it is already handled by Gorka.
15:08:23 <zenkuro> rosmaita: thanks for your detailed comments on CI, Im applying fixes right now
15:08:32 <rosmaita> that one illustrates a point eric made last week
15:08:45 <rosmaita> namely, we need to review some old bugs
15:09:04 <enriquetaso> rosmaita++
15:09:04 <rosmaita> enriquetaso: i can make it un-critical
15:09:06 <enriquetaso> My idea was to do an update of the old bugs but I could not do it for this week, I will prepare it for the next and maybe some for the PTG if they are important.
15:09:23 <rosmaita> yeah, i am not saying that you personally need to do it
15:09:27 <enriquetaso> Fix proposed:
15:09:29 <rosmaita> just that "we" need to do it
15:09:32 <enriquetaso> #link https://review.opendev.org/c/openstack/cinder/+/780755
15:09:45 <enriquetaso> rosmaita sure
15:10:05 <rosmaita> so the story is that sqlalchemy 1.4.0 drops that joinedload_all call
15:10:14 <rosmaita> and 1.4.0 was released on monday
15:10:24 <rosmaita> and our docs job was using it and breaking
15:10:39 <rosmaita> in the mean time, 1.4.0 is not included in upper-constraints
15:10:59 <rosmaita> our doc job was ignoring upper constraints (sort of)
15:11:10 <rosmaita> so, we have fixed our jobs to respect upper constraints
15:11:31 <rosmaita> what i am saying is that it is important that we fix this, because we will be broken in xena
15:11:42 <rosmaita> but it is not as critical as i thought yesterday
15:11:51 <rosmaita> and now i will shut up
15:11:52 <enriquetaso> ohh
15:11:57 <eharney> well, it's a pretty simple change
15:12:15 <enriquetaso> thanks rosmaita for the details \o/
15:12:36 <enriquetaso> The following bug is already assigned and is being worked on, but I wanted to highlight it in case someone is using quotas.
15:12:51 <enriquetaso> #topic bug_2: "Automatic quota refresh counting temporary volumes"
15:12:59 <enriquetaso> #link https://bugs.launchpad.net/cinder/+bug/1919161
15:13:00 <openstack> Launchpad bug 1919161 in Cinder "Automatic quota refresh counting temporary volumes" [High,New] - Assigned to Gorka Eguileor (gorka)
15:13:08 <enriquetaso> 'When using the automatic quota refresh via `until_refresh` and `max_age` configuration options the calculated quota usage by the refresh will not be correct if there are temporary volumes (such as those created from snapshots for backups)'
15:13:39 <enriquetaso> I couldn't find a patch for ^ so I think it doesn't have one yet
15:14:11 <enriquetaso> Moving on... I need an opinion on next one:
15:14:23 <enriquetaso> #topic bug_3: "id of encryption for volume type not verified"
15:14:28 <enriquetaso> #link https://bugs.launchpad.net/cinder/+bug/1918879
15:14:29 <openstack> Launchpad bug 1918879 in Cinder "id of encryption for volume type not verified" [Undecided,New]
15:14:37 <enriquetaso> The description of the bug just says "when try to update encryption for a given volume type, this api received one parameter is id, but it is not used,therefore, when it is a wrong encryption id, the api layer not check".
15:14:38 <enriquetaso> I guess the bug is incomplete, I'd like to ask about steps of how to reproduce the problem and probably if It's using LUKS or something else. However, I've never updated a volume type before so I'd like to be double sure about what to ask.
15:14:55 <eharney> proably https://review.opendev.org/c/openstack/cinder/+/779436
15:15:53 <enriquetaso> great
15:16:36 <rosmaita> enriquetaso: i think start by asking exactly what api call, what they are passing in the request, and also what release they are talking about
15:18:30 <enriquetaso> #action enriquetaso: ask for more detailed information
15:18:46 <enriquetaso> next one:
15:19:03 <enriquetaso> #topic bug_4: "ensure_export model_update ignored"
15:19:08 <enriquetaso> #link https://bugs.launchpad.net/cinder/+bug/1918449
15:19:09 <openstack> Launchpad bug 1918449 in Cinder "ensure_export model_update ignored" [Wishlist,Incomplete]
15:19:20 <enriquetaso> "The volume manager calls the driver's ensure_export() at startup time. There are several volume drivers that return a model_update from ensure_export, but the volume driver ignores that model_update."
15:20:02 <enriquetaso> I set the importance to wishlist, should I change it to medium?
15:20:17 <eharney> this still needs some justification that it is really a bug, and we need a discussion/decision about whether drivers should really be returning updates from ensure_export calls
15:21:05 <eharney> likely to end up as a ptg discussion i guess
15:21:10 <enriquetaso> Yes, please read Eric's comment
15:21:57 <eharney> before we resolve this bug, we really need to resolve: "how exactly does the ensure_export driver API behave?" which seems to be up in the air a bit
15:22:26 <rosmaita> https://github.com/openstack/cinder/blob/master/cinder/volume/driver.py#L1430
15:22:49 <rosmaita> so the other ones explicitly say " can optionally return a dict of changes", this one does not
15:23:10 <eharney> right, and we have had code for ages that does not accept a return dict from it
15:23:29 <eharney> so it seems reasonable to start at the idea that it probably shouldn't be updating things that require it to return a dict, etc
15:23:46 <rosmaita> i agree
15:23:58 <eharney> which goes back to the questions i was asking before: why exactly should we _start_ doing that?  still unclear
15:24:25 <enriquetaso> so, it's a good topic for the next PTG
15:25:27 <enriquetaso> OK..
15:25:30 <enriquetaso> #topic Open Discussion
15:25:36 <kinpaa12389> https://bugs.launchpad.net/cinder/+bug/1298135
15:25:37 <openstack> Launchpad bug 1298135 in Cinder "Cinder should handle token expiration for long ops" [Medium,Confirmed]
15:25:49 <kinpaa12389> can we discuss this ?
15:25:55 <enriquetaso> sure
15:26:09 <enriquetaso> #link https://bugs.launchpad.net/cinder/+bug/1298135
15:26:17 <enriquetaso> "Cinder should handle token expiration for long ops"
15:26:29 <kinpaa12389> so cinder <-> nova and cinder <-> glance support service_token
15:26:36 <kinpaa12389> cinder <->swift does not..
15:27:07 <kinpaa12389> so I added that support and sent service_token to swift, modified swiftclient to consider session to talk with keystone middleware.
15:27:24 <kinpaa12389> however this does not work. and token expiry trigger 401 Error.
15:27:25 <eharney> i'd like to understand why it does not
15:28:38 <kinpaa12389> the swift proxy-server validates token for each backup chunk and it talks with keystone middleware to validate token.. This report failure as invalid token because its expired.
15:28:40 * enriquetaso likes the tag 'bugsmash' :P
15:29:08 <kinpaa12389> So, I see two ways to solve it:
15:29:09 <kinpaa12389> 1) Either patch both Cinder and Swift to support service_auth and its tokens.
15:29:09 <kinpaa12389> 2) Or use Keystone trust and refresh tokens like Glance already does for long running image uploads.
15:30:14 <eharney> my general sense is that this shouldn't require patching Cinder and Swift
15:30:47 <kinpaa12389> eharney can you try this experiment. may be I am doing incorrect steps.
15:30:50 <rosmaita> i'm not sure what the current status of keystone trusts is
15:30:58 <eharney> i'm not sure that keystone trusts are needed for this
15:31:27 <eharney> i don't think we use trusts for service tokens used to talk to nova from cinder, do we?
15:31:45 <kinpaa12389> no trusts only used for glance <->swift
15:31:48 <kinpaa12389> its not in cinder
15:32:08 <kinpaa12389> ok for option 1.. add expiration in keystone.conf e.g. 100 seconds and try to upload backup of 2GB in swift.
15:32:28 <eharney> i think there's more to it than that
15:32:29 <rosmaita> ok, so that's a problem right there
15:32:36 <eharney> i.e. the details in  https://docs.openstack.org/cinder/latest/configuration/block-storage/service-token.html
15:32:38 <rosmaita> the service tokens also expire
15:32:42 <kinpaa12389> right.
15:32:47 <kinpaa12389> forgot to mention
15:33:06 <kinpaa12389> u need to add [service_user] section and respective support in cinder/backup/drivers/swift.py
15:33:42 <kinpaa12389> please update bug if it works for anyone whoever can try.
15:34:20 <enriquetaso> OK, i need to end the meeting
15:35:44 <enriquetaso> the last update was in 2019
15:36:07 <rosmaita> oh meant to mention this one: https://bugs.launchpad.net/oslo.db/+bug/1814199
15:36:09 <openstack> Launchpad bug 1814199 in Cinder "soft_delete is wrong" [High,In progress] - Assigned to Gorka Eguileor (gorka)
15:36:24 <enriquetaso> sure
15:36:45 <rosmaita> this is another old one that gorka has a patch up for
15:36:56 <enriquetaso> #link https://review.opendev.org/c/openstack/cinder/+/776974
15:37:03 <rosmaita> so there's probably not much to say about it
15:37:04 <enriquetaso> Fix proposed ^
15:37:15 <rosmaita> other than i promised gorka that i would review it
15:37:32 <enriquetaso> #action review https://review.opendev.org/c/openstack/cinder/+/776974
15:37:37 <rosmaita> thanks!
15:37:56 <rosmaita> it's a bit more complicated than that other sqlalchemy patch
15:38:34 <enriquetaso> I'll try to reproduce the bug kinpaa12389, but i'm not familiar with the topic
15:39:16 <enriquetaso> rosmaita++
15:39:18 <rosmaita> kinpaa12389: you may not see much action on that until after RC-time
15:39:36 <rosmaita> we kinda have our hands full at the moment
15:39:47 <enriquetaso> true..
15:39:59 <enriquetaso> OK, thank you all!
15:40:00 <enriquetaso> #endmeeting