Wednesday, 2021-09-15

*** bhagyashris is now known as bhagyashris|off05:35
opendevreviewAbhishek Kekane proposed openstack/glance master: Xena RC-1 release notes  https://review.opendev.org/c/openstack/glance/+/80912907:39
opendevreviewjinyuanliu proposed openstack/python-glanceclient master: Clean up extra spaces  https://review.opendev.org/c/openstack/python-glanceclient/+/80915508:33
opendevreviewPranali Deore proposed openstack/glance-tempest-plugin master: Add protection testing for metadef namespaces  https://review.opendev.org/c/openstack/glance-tempest-plugin/+/80090211:11
opendevreviewPranali Deore proposed openstack/glance-tempest-plugin master: Add protection testing for metadef namespaces  https://review.opendev.org/c/openstack/glance-tempest-plugin/+/80090211:58
abhishekkRC1 reno patch up for review, https://review.opendev.org/c/openstack/glance/+/80912913:42
* dansmith is reviewing already13:42
abhishekkcool, thank you13:42
dansmithleft you a question13:43
abhishekkack, ++13:44
opendevreviewAbhishek Kekane proposed openstack/glance master: Xena RC-1 release notes  https://review.opendev.org/c/openstack/glance/+/80912913:47
abhishekkI think we are good for Xena now13:54
abhishekkexcept for metadef tempest-plugin patches, I will give ack to release of tempest-plugin patch and then we can backport it to stable branch13:55
dansmithoh we have to release the tempest plugin?13:56
dansmithI had assumed not13:57
abhishekkhmm, we need to13:57
dansmithokay13:58
abhishekkwait, we don't do releases for tempest-plugin, you are right13:59
abhishekkgmann, I think we don't create stable branch but do the release, right ?14:04
abhishekkcroelandt, last review request for this cycle, https://review.opendev.org/c/openstack/glance/+/809129 14:05
opendevreviewPranali Deore proposed openstack/glance-tempest-plugin master: Implement API protection testing for metadef objects  https://review.opendev.org/c/openstack/glance-tempest-plugin/+/80279314:10
opendevreviewPranali Deore proposed openstack/glance-tempest-plugin master: Implement API protection testing for metadef resource types  https://review.opendev.org/c/openstack/glance-tempest-plugin/+/80279214:10
opendevreviewPranali Deore proposed openstack/glance-tempest-plugin master: Implement API protection testing for metadef properties  https://review.opendev.org/c/openstack/glance-tempest-plugin/+/80279414:10
opendevreviewPranali Deore proposed openstack/glance-tempest-plugin master: Implement API protection testing for metadef tags  https://review.opendev.org/c/openstack/glance-tempest-plugin/+/80279514:10
gmannabhishekk: we do not create stable branch but plugins needs to release a tag for cycle support/end-of-support etc14:10
abhishekkgmann, ack, so last date for release is 17th ?14:11
gmannabhishekk: https://docs.openstack.org/tempest/latest/tempest_and_plugins_compatible_version_policy.html14:11
gmannabhishekk: as they are with cycle-with-intermediary release model I think 13th was deadline? but you can check in release channel I think other plugins are also not tagged yet (tempest tag is also not merged in release side)14:12
croelandtabhishekk: is it just me, or does this one seem easier than the previous ones? :)14:13
abhishekkcroelandt, :D14:13
abhishekkgmann, https://review.opendev.org/c/openstack/releases/+/80878214:14
abhishekkthis commit mentions 17 as a dead line14:14
gmannabhishekk: perfect14:14
abhishekkok, that means we have couple of days to work on pending changes and get those merged14:15
gmannabhishekk: yeah, 17th is in case PTL does not respond on release patches. 14:15
abhishekkso is there a backdoor to push it further >14:15
gmannabhishekk: if anything very urgent then we can wait but as things are tested with plugins master branch so there is no missing testing for xena cycle things. 14:15
abhishekkack14:16
abhishekksounds good then14:16
croelandtIs there any reason why os_hash_algo and os_hash_value may not be set on an image? Is there an option to disable the multihash feature or something?14:17
gmannabhishekk: main idea behind release wise tag for tempest & plugins was to have a known compatible tags between tempest and plugins so that any production cloud can install them easily.  but as you know tempest and plugins master support many stable branch so other version also works fine as we test in upstream too14:17
gmannabhishekk: so if you release now or after few patches that is no so much difference as such14:17
abhishekkyep, got it 14:18
abhishekkwill wait for today or else give the ack on the patch14:18
abhishekkthank you for explaining in detail14:18
abhishekkcroelandt, I think those are set automatically14:18
croelandtI got a user downstream who has no os_hash_algo nor os_hash_value on his images14:19
croelandtI cannot reproduce it, a Horizon dev tried to reproduce it, no luck14:20
croelandtThis seems to happen only when they create a snapshot of an instance14:20
croelandtit's driving me crazy14:20
abhishekkack, brb14:20
dansmithmaybe direct snapshot?14:21
dansmithi.e. nova creates the snapshot in ceph directly and just tells glance about it?14:21
croelandtoh that is a thing?14:22
gmannpdeore: abhishekk dansmith lbragstad on rbac test. one suggestion. I think making create_namespace() taking client object or project_id for which namespace needs to be created will be easy code to read/maintain and test can pass their required client, project_id they want to test on - https://review.opendev.org/c/openstack/glance-tempest-plugin/+/800902/27/glance_tempest_plugin/tests/rbac/v2/base.py#9114:22
dansmithcroelandt: kindof an important one :)14:22
croelandtdansmith: so Nova creates the snapshot, and then creates a "pointer" in Glance to the rbd location?14:22
gmannI think that is what lbragstad points on his comment ?14:23
dansmithcroelandt: location yeah14:23
dansmithgmann: I dunno, it only ever needs to use the namespace client to do that14:23
gmanndansmith: client i mean admin, system, reader etc14:24
dansmithgmann: I think the thing that would help clean all that up is not keep re-writing self.client to be "the one client I'm mostly testing here" and just create them all like self.images_client, self.namespace_client, self.objects_client, etc14:24
croelandtdansmith: so in that case we don't hash the data?14:24
dansmithcroelandt: wouldn't surprise me.. we have t read it all in order to calculate the hash right? and in those cases, we never do (I assume)14:24
gmanndansmith: yeah that is true, self.client are bad things and many in tempest too. if we want to test what all persona can create/access namespace then passing that persona in create_namespace() is easy.14:25
dansmithgmann: I also think the getattr(cls, f'os_{cls.credentials[0]}') pattern is pretty dangerous14:25
dansmithgmann: because I changed the ordering of cls.credentials at one point and that broke things and I couldn't figure out why14:25
gmannyeah that is why I had hard time to understand the things in those tests14:26
lbragstadyeah - that's obtuse 14:26
dansmithI dunno if lbragstad copied that from someone, but I know he *loves* f-strings14:26
dansmithbut in that case, it seems pretty not great to me14:26
lbragstadyeah - i used the keystone tests as a template14:27
lbragstadand i kept it consistent for the time being14:27
dansmithalright, you can only keep using that excuse so many times :D14:27
gmann:)14:27
dansmithbut yeah, especially in these metadef cases,14:27
lbragstadf-strings aside, the ordering part is probably the more fragile part, yeah?14:27
dansmithwhere we have a bunch of clients all kinda stacked up, and lots of creds, I just think namespacing them and making them explicit would be better14:28
dansmithlbragstad: yeah it's just the creds[0] thing14:28
lbragstadright - ok 14:28
dansmithlbragstad: f-strings make that less likely for a reader to realize there's code in that string, which is why I hate them, but yeah14:28
gmann+1. explicit is much correct/easy. 14:28
dansmithnormally I would kinda skim over strings when looking for conditional code, and ... f-strings break that14:29
* dansmith stops ranting about f-strings14:29
lbragstadyeah - some sort of method that iterates the credentials and exposes them without an implied order enforced is a good idea14:29
lbragstadi feel like that should be in tempest proper14:29
dansmithyeah, maybe self.${cred}.${thing}_client pattern would be nice14:30
gmannlbragstad: you mean self.os_system_admin etc ?14:30
lbragstadwell - guess we could just do that14:31
lbragstadand not alias anything, right?14:31
gmannthat is we have in tempest. self.system_admin|reader.<x_client>14:31
lbragstadright now we do things like,14:31
lbragstadself.admin = self.os_system_admin.namespaces_client14:32
gmannyeah alias are mistake in tempest and as they were so many we never got rid of those completely especially self.client  14:32
lbragstadhmm 14:32
lbragstadok14:32
dansmithgmann: ++14:32
lbragstadwell - if we don't think aliasing and shortening the names is important, then we could just define the credentials list and we don't have to rely on an order in setup_clients()14:32
lbragstadand it removes the "wait, what is self.client() again?"14:33
lbragstadwhen you're reading the tests14:33
dansmithright14:33
gmannyes14:33
dansmithI do like short names, no doubt,14:33
dansmithbut it would also be easy to, at the top of a test, do "client = self.admin.namespaces_client"14:33
gmannreading test is more imp i think as we donot have great test docuemntation 14:33
gmannat least within that test file14:33
lbragstadsame, but i'm fine reading a bit more if it means the tests are more robust 14:34
gmannand if we have in test file then it is easy to know test requested those creds and using them in their tests14:35
gmann*same test file14:35
croelandtdansmith: so the whole point of direct snapshots is to be quicker, which explains why it would not make sense to compute the hash?14:48
dansmithwell, the whole point of direct snapshots is to avoid sending the bits from ceph->nova->glance->ceph when ceph can just CoW the disk and return a new reference (or whatever)14:49
dansmithbut the side-effect is that we don't actually handle the bits, we just handle the reference14:49
dansmithand as such, I'm not sure how we *could* know the hash, unless ceph was able to tell us (and I expect it doesn't/cant')14:50
croelandtok14:52
croelandtso, fun fact14:52
croelandtIf you want to edit an image that does not have os_hash_* set, you can using the CLI14:52
croelandtbut Horizon is like "NOPE"14:53
croelandtThen people ask me why I hate GUI programs, and why I don't like the Web14:53
dansmith"edit" meaning what? just update some other properties?14:54
dansmithI'm guessing that's either a bug, or an ill-placed safeguard on horizon's part14:54
croelandtyeah I think they're trying to edit some properties14:55
croelandtand the "edit" button is greyed out14:55
croelandtI'll check with them that it is possible to do itthrough the CLI14:55
croelandtbut yeah it feels like it's a check from Horizon, something like "if any of the properties is NULL, then something must be wrong"14:56
dansmithyeah14:57
* abhishekk scrolling back14:59
abhishekkso it looks like glance-tempest-plugin needs some more refactoring15:01
opendevreviewMerged openstack/glance master: Xena RC-1 release notes  https://review.opendev.org/c/openstack/glance/+/80912917:23
abhishekkRC1 release patch - https://review.opendev.org/c/openstack/releases/+/80862517:43
* abhishekk signing out for the day18:15
*** timburke_ is now known as timburke20:59
*** prometheanfire is now known as Guest121:20
*** promethe- is now known as prometheanfire21:53

Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!