Earlier  
Later
New Search
Posted Nick Remark
#openstack-sdks - 2018-10-01
15:26:49 mordred dtantsur: zomg. the timeouts are killing me
15:27:36 dtantsur yep :(
15:28:00 mordred dtantsur: it seems like the last several have been in the v1/object_store functional tests, yeah?
15:28:36 dtantsur I did not check all of them, to be honest
15:28:49 mordred yeah - me either - I just started and there's a network one :(
15:38:18 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Bump the per-test timeout for functional tests https://review.openstack.org/606980
15:38:40 mordred dtantsur: trying a sledge hammer
15:39:52 dtantsur sigh
15:54:32 mordred yeah
15:59:36 mordred I very much want to fix the tests for real - like -looking at this: http://logs.openstack.org/14/604414/7/check/openstacksdk-functional-devstack-tips/828e753/job-output.txt.gz#_2018-10-01_11_21_49_493288
15:59:40 mordred something went away for 4 minutes
16:18:10 Shrews can we just remove the tests????
16:23:02 dtantsur ++ we have unit tests, they're enough
22:31:14 openstackgerrit Duc Truong proposed openstack/openstacksdk master: Add functional tests for clustering https://review.openstack.org/607061
#openstack-sdks - 2018-10-02
00:35:32 openstackgerrit Duc Truong proposed openstack/openstacksdk master: Add functional tests for clustering https://review.openstack.org/607061
04:46:54 openstackgerrit Merged openstack/openstacksdk master: Clarify error message is from nova https://review.openstack.org/605544
10:05:44 ml is someone so kind to take an additional look at https://review.openstack.org/#/c/592303/
12:12:37 openstackgerrit Merged openstack/api-sig master: fix tox python3 overrides https://review.openstack.org/606618
14:05:55 mordred ml: looks great - and also, I think from reviewing it I just realized why something unrelated is having issues -so thanks!
14:08:15 ml cool. thanks for looking at it
14:10:54 mordred dtantsur: the normal devstack-functional job for sdk has octavia and designate enabled - and now fwaas- I'm thinking maybe we should split those out into a different job and do what you did for ironic limiting which tests we run - maybe running the extra services is putting us just over the line of what the 8G nodes can handle
14:11:46 dtantsur mordred: ++ I was thinking of asking you the same
14:11:58 dtantsur I think more jobs is better than huge jobs
14:12:12 mordred yah - cause otherwise these timouts are killing me
14:14:01 dtantsur right, we may even do some less frequently used stuff non-voting like ironic
14:23:03 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Bump the per-test timeout for functional tests https://review.openstack.org/606980
14:23:03 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Clean up python3 test and remove duplicate jobs https://review.openstack.org/604517
14:23:04 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Split advanced networking services into a job https://review.openstack.org/607199
14:24:28 mordred dtantsur: ok. there's a stab at splitting networking - let's see how that does. if it works, I think maybe squashing all three of those into a single patch
14:25:47 dtantsur something weird with these patches - the two top ones are "Cannot merge"
14:45:09 dtantsur mordred, edleafe, elmiko, cdent, I cannot remember from our discussions (if we discussed that at all): if somebody commits the crime against humanity and raises the minimum microversion, what will the default be?
14:45:15 dtantsur I mean, when a microversion is not provided?
14:45:38 cdent dtantsur: presumably the new min?
14:46:04 dtantsur yeah, I guess my question is: new min or HTTP NOT ACCEPTABLE?
14:46:13 cdent how could it be anything else?
14:46:42 dtantsur it can be an error
14:46:56 dtantsur i.e. we could treat absent version as requesting the old min
14:47:49 cdent why would you want to do that? the goal of absent version is to have at least _something_ happen
14:50:08 edleafe Agreed: it should work the same as it does now when there is no microversion supplied - the minimum is assumed
14:51:30 dtantsur okie
17:51:45 mordred dtantsur|afk: the default will be that my face comes out of the computer and mocks someone
17:59:16 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Split advanced networking services into a job https://review.openstack.org/607199
18:08:46 openstackgerrit Merged openstack/python-openstackclient master: fix tox python3 overrides https://review.openstack.org/606702
18:22:48 mordred dtantsur|afk: http://logs.openstack.org/14/604414/7/gate/openstacksdk-functional-devstack-tips/a3fc28f/controller/logs//syslog.txt.gz#_Sep_28_18_56_40 <-- timburke found the issue because he's smart
#openstack-sdks - 2018-10-03
01:30:12 openstackgerrit Merged openstack/python-openstackclient master: Add metavar for name parameter in subnet create https://review.openstack.org/602062
02:32:45 openstackgerrit Andriy Shevchenko proposed openstack/python-openstackclient master: fix tox python3 overrides https://review.openstack.org/607453
12:11:25 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Split advanced networking services into a job https://review.openstack.org/607199
12:13:06 ml mordred: https://review.openstack.org/#/c/592303/
12:13:27 ml am I able to trigger a recheck myself or do you have to reset the workflow?
12:13:52 mordred ml: you can - but hold off for now- the issue that's happening is the test node running out of memory
12:14:03 ml alright
12:14:26 mordred ml: I'm trying to fix with https://review.openstack.org/607199 - if it fixes it, I'll probably just squash it into your patch and merge it that way
12:15:40 ml alright, I'll keep an eye on this change. thanks for your work
13:31:04 rosmaita cdent: have an api question for you when you have a few minutes
13:33:51 cdent rosmaita: i've got time, for you, right now
13:34:32 rosmaita cdent: thanks! can you take a look at https://review.openstack.org/#/c/597648/7 /
13:34:44 rosmaita (meant that to be a ? at the end)
13:41:54 cdent rosmaita: okay, I think I'm caught up on the text of the spec and the comments
13:42:12 cdent what's the main issue you're concerned with?
13:42:50 rosmaita cdent: whether i'm right about us having to introduce a new json-patch mime type for this
13:43:03 rosmaita and whether that's a good idea
13:43:11 smcginnis Mimes are never a good idea.
13:43:25 rosmaita smcginnis: i can't hear you
13:43:31 rosmaita :)
13:43:39 smcginnis :)
13:44:57 cdent you both need to go sit on the naughty step
13:47:13 cdent rosmaita: I'd say the issue here is that the validation_data is being marked out as a separate in the representation. If checksum, os_hash_algo, and os_hash_value are legit members of the representation of the 'locations' attribute, then just pass those and _in_code_ be aware that they are validation bits
13:47:25 cdent otherwise you are not truly doing a replace operation (from a json-patch standpoint)
13:47:53 rosmaita cdent: tjat
13:48:09 rosmaita that's the key issue, we aren't doing a replace
13:48:25 cdent it sure looks like you are :)
13:49:43 rosmaita the locations attribute has a 'metadata' property, which is a json object iirc, but it doesn't have individual checksum, os_hash_* properties
13:50:28 rosmaita and we don't really want to leave 'checksum' as a location metadata item, we want there to be only one true checksum, as an image property
13:51:53 rosmaita cdent: but that's probably too much detail. i guess the key thing is that a json-patch replace operation should actually replace stuff or fail (and not do weird side-effect stuff like we're proposing)
13:52:03 cdent you say in one of your initial comments " because the "validation_data" object does not exist in the image/location schema; it's sort of an instruction, really, that produces a side effect (the population of other properties on that image)". Can you just populate those properties?
13:52:35 cdent yes: if you're using json-patch replace, it should do json-patch replace, not some special glance-only magic
13:52:54 rosmaita cdent: yes, that's the alternative (but those properties are currently read-only in the schema, which is a sticking point)
13:53:37 rosmaita the issue is that we only want these properties to be settable under one specific circumstance, and never at any other time
13:53:40 cdent making them writable during certain states is probably OK
13:53:46 cdent jinxish
13:58:37 cdent rosmaita: Have I helped or just made it worse?
13:58:48 rosmaita cdent: thinking!
13:59:27 rosmaita cdent: last question ... if we make these writable under one tiny circumstance, then we have to remove readOnly: true from the schema for them, right?
14:01:21 cdent I think that depends on how you are using and publishing the schema. In my mind schema are a validation tool used indepdently for each method:resource pair, not a contract in an of themselves
14:03:09 rosmaita cdent: thanks, that is helpful and will give us something to think about
14:03:28 cdent you're welcome
16:37:04 Luzi mordred are you there?
16:37:23 mordred Luzi: *sortof* - I'm at a conference so I'm in and out
16:38:15 Luzi okay, short question: it is planned to use openstacksdk in more projects right?
16:38:23 Luzi like nova and cinder maybe?
16:39:38 smcginnis Nothing active that I've seen, but not out of the question.
16:42:27 mordred Luzi: yes - it's already in use in heat, senlin and masakari
16:42:48 mordred Luzi: once the 1.0 is cut, I'll start poking at making patches for folks to make that possible
16:43:15 mordred Luzi: I don't know that the various services have agreed that they want such a migration yet - but from my/sdk point of view, once 1.0 is out it's a great idea
16:43:43 smcginnis mordred: Forum topic?
16:44:08 Luzi mordred, thank you for now :)
16:44:20 smcginnis Too late now to be proposed, but we may be able to find a time for it if there is interest/need.
16:50:50 mordred smcginnis: yah. I was supposed to talk to nova about it in denver - but I missed the irc ping
18:45:45 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Split advanced networking services into a job https://review.openstack.org/607199
19:35:34 kmalloc mordred: how far is SDK away from being ... usable say, in lieu of keystoneclient?

Earlier   Later