| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-sdks - 2018-08-03 | |||
| 21:17:03 | mnaser | mordred: do you know anything why the tests are failing in `ansible-role-cloud-launcher-functional` ? it looks like it's refusing to create a router based on policy? | |
| 21:17:10 | mnaser | i dunno if that's something you've ran into before | |
| 21:18:34 | openstackgerrit | Sean McGinnis proposed openstack/python-openstackclient master: Fix broken gate jobs https://review.openstack.org/587005 | |
| 22:24:28 | openstackgerrit | James E. Blair proposed openstack/openstacksdk master: Don't wait for task in submit_task https://review.openstack.org/588656 | |
| #openstack-sdks - 2018-08-04 | |||
| 14:28:27 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: Don't wait for task in submit_task https://review.openstack.org/588656 | |
| 15:08:46 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: Don't wait for task in submit_task https://review.openstack.org/588656 | |
| 15:10:20 | mordred | corvus: ok, that ^^ fixes the tests - the issue at hand is that for async tasks (currently only used during uploading swift segments for large objects) the tasks are wrapped in concurrent futures, so shifting the wait was causing stuff to not finish properly | |
| 15:11:03 | mordred | corvus: the change I made was basically to apply your fix to the sync path but not the async - I think we probably want to dig in further and wrap our heads around the problem for the async path too | |
| 15:12:00 | corvus | mordred: are the async tasks fire-and-forget, or does the caller care whether they eventually finish? | |
| 15:15:45 | mordred | the caller cares - but the caller uses the Future returned to get the eventual result | |
| 15:17:28 | mordred | corvus: if you grab version 2 of the change and run the unittests you can see the 5 failures that come from the async side of the house | |
| 15:18:37 | mordred | corvus: I can't quite wrpap my head around why version 2 of the change breaks in the way that it does | |
| 15:19:39 | mordred | corvus: oh - wait ... I'm wrong | |
| 15:20:06 | mordred | corvus: they are, from a task manager /caller perspective fire and forget | |
| 15:20:49 | mordred | corvus: we only task-manager/rate-limit the start of the task - the concurrent.future that is returned handles exception propogation | |
| 15:21:13 | mordred | so in this case I think the new version of the patch is actualy correct - it just needs more comments | |
| 15:23:13 | mordred | (it's also ugly, so it's entirely possible it could be refactored to not have quite so many if async: in it) | |
| 15:29:07 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: Don't wait for task in submit_task https://review.openstack.org/588656 | |
| 15:29:20 | mordred | corvus: ^^ added some comments and collapsed one of the copies of the code | |
| 15:50:22 | openstackgerrit | Merged openstack/openstacksdk master: Fix filter style consistency for keystone assignment API https://review.openstack.org/569924 | |
| #openstack-sdks - 2018-08-05 | |||
| 14:33:18 | openstackgerrit | Merged openstack/openstacksdk master: Update storyboard links to use name https://review.openstack.org/588596 | |
| 15:44:20 | openstackgerrit | Merged openstack/openstacksdk master: Send disk_over_commit if nova api < 2.25 https://review.openstack.org/585579 | |
| #openstack-sdks - 2018-08-06 | |||
| 02:17:13 | openstackgerrit | Merged openstack/openstacksdk master: Pass microversion info through from Profile https://review.openstack.org/587435 | |
| 03:57:35 | openstackgerrit | Chen proposed openstack/python-openstackclient master: Add info to --live flag for 'openstack server migrate' https://review.openstack.org/589012 | |
| 08:14:49 | openstackgerrit | Jens Harbott (frickler) proposed openstack/python-openstackclient master: Add DNS support to floating IP commands https://review.openstack.org/558824 | |
| 11:38:14 | cdent | edleafe, elmiko, dtantsur: looks like we have some bad links, maybe in zuul jobs. On http://specs.openstack.org/ the link of guidelines is /api-sig but the content is at /api-wg | |
| 11:42:16 | dtantsur | ugh | |
| 11:42:20 | dtantsur | yeah, it may be zuul jobs | |
| 11:43:20 | cdent | I'm fixing some links in the actual content right now, but though I would pawn that off on one you :) | |
| 12:01:18 | openstackgerrit | Chris Dent proposed openstack/api-sig master: Update web frontpage to link to SIG not WG https://review.openstack.org/589130 | |
| 12:01:18 | openstackgerrit | Chris Dent proposed openstack/api-sig master: Explicitly link to the HTTP RFCs https://review.openstack.org/589131 | |
| 12:01:19 | openstackgerrit | Chris Dent proposed openstack/api-sig master: Add link to Building Protocols with HTTP https://review.openstack.org/589132 | |
| 13:12:30 | openstackgerrit | Merged openstack/microversion-parse master: fix tox python3 overrides https://review.openstack.org/573880 | |
| 14:18:26 | elmiko | cdent: ack, just saw the message | |
| 14:23:37 | mordred | dtantsur|brb: I made some patches over the weekend I think you're going to like ... | |
| 15:28:31 | dtantsur | mordred: heads-up: os-service-types from master breaks openstacksdk | |
| 15:28:46 | mordred | dtantsur: really? well crap | |
| 15:28:48 | dtantsur | try ./.tox/py27/bin/pip install git+https://git.openstack.org/openstack/os-service-types && tox -epy27 | |
| 15:29:16 | mordred | k. will do in just a sec | |
| 15:31:10 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: Make resource a dict subclass usable by shade layer https://review.openstack.org/530835 | |
| 15:31:11 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: Rename Resource get and update to not clash with dict https://review.openstack.org/589193 | |
| 15:31:11 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: Add computed attribute type and location to base resource https://review.openstack.org/589194 | |
| 15:31:12 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: Remove special handling of stacks https://review.openstack.org/589195 | |
| 15:31:41 | mordred | dtantsur: ^^ so that's ALMOST working (this is making the Resource objects dict subclasses so we can have shade methods return them directly) | |
| 15:32:13 | dtantsur | yay! | |
| 15:32:44 | mordred | dtantsur: there's one weird bug in the final patch that is confusing to me - I've got an infinite recursion thing happening and I cannot see why | |
| 15:33:07 | mordred | there's 2 lines commented out in openstack/resource.py with a note above them about it | |
| 15:33:34 | mordred | looing at the os-service-types thing now | |
| 15:33:42 | mordred | oh - fun | |
| 15:34:30 | dtantsur | mordred: filed a story https://storyboard.openstack.org/#!/story/2003314 | |
| 15:35:30 | openstackgerrit | Monty Taylor proposed openstack/os-service-types master: Fix service_type normalization https://review.openstack.org/589199 | |
| 15:35:35 | mordred | dtantsur: there ya go ^^ | |
| 15:35:41 | mordred | and thanks! | |
| 15:36:49 | dtantsur | np, it failed the RDO build of metalsmith :) | |
| 15:39:37 | openstackgerrit | Merged openstack/openstacksdk master: Don't wait for task in submit_task https://review.openstack.org/588656 | |
| 15:47:54 | mordred | dtantsur: yay for CI! | |
| 15:49:42 | dtantsur | :) | |
| 15:52:56 | openstackgerrit | Monty Taylor proposed openstack/os-service-types master: Fix service_type normalization https://review.openstack.org/589199 | |
| 15:53:23 | dtantsur | mordred: I wonder if we should be calling get_all_types if service_type is None at all | |
| 15:54:44 | mordred | dtantsur: nope. I'm pretty sure we shouldn't | |
| 16:01:45 | cdent | i was gonna ask that | |
| 16:07:15 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: Stop calling get_all_types when service-type is None https://review.openstack.org/589212 | |
| 16:07:26 | mordred | dtantsur, cdent: ^^ there, fixed it on that side too | |
| 16:07:51 | dtantsur | mordred: a unit test here as well? :) | |
| 16:08:11 | mordred | dtantsur: if you get bored with your real job and feel like trying to debug the intersection between magical descriptor objects and dict subclasses ... | |
| 16:08:45 | mordred | dtantsur: hrm. just one that calls _get_config with None? | |
| 16:09:02 | dtantsur | mordred: 1. not THAT bored, but can give it a try, 2. yep | |
| 16:18:29 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: Stop calling get_all_types when service-type is None https://review.openstack.org/589212 | |
| 16:18:36 | mordred | dtantsur, cdent: ^^ how does that look? | |
| 16:20:21 | cdent | mordred: looks sane, but I'm not quite sure I grok what it means for service-type to be None at this stage, so I can mostly just take your word for it | |
| 16:20:45 | mordred | cdent: well - the times in which service-type should be none in the real world are pretty low | |
| 16:21:43 | mordred | oh - hahahahahaha | |
| 16:22:13 | mordred | I was about to make some snide comments about unit tests - but amusingly the thing I randomly chose as the config value for that unittest | |
| 16:22:24 | mordred | is the use case that caused this to fail in the first place | |
| 16:23:24 | mordred | in fact - lemme change that test .. | |
| 16:25:54 | openstackgerrit | Merged openstack/os-service-types master: Fix service_type normalization https://review.openstack.org/589199 | |
| 16:30:39 | dtantsur | heh | |
| 16:31:04 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: Stop calling get_all_types when service-type is None https://review.openstack.org/589212 | |
| 16:31:12 | mordred | k. maybe that test is clearer | |
| 16:41:50 | mordred | dtantsur: aha! the dict issue that exists in that patch exists without that patch - the patch just triggers it | |
| 16:42:20 | mordred | dtantsur: tl;dr - there are some objects that, if you call to_dict() on them - will infinite recursion | |
| 16:42:51 | dtantsur | \o/ | |
| 16:42:56 | mordred | yeah. SO exciting | |
| 16:42:58 | dtantsur | this is Monday | |
| 16:43:00 | mordred | yup | |
| 16:43:07 | mordred | it's such a monday bug isn't it? | |
| 16:43:21 | dtantsur | truly | |
| 16:43:31 | mordred | otoh - it makes debugging it easier - because I was going CRAZY trying to figure out how the patch was breaking it | |
| 16:43:48 | mordred | turns out - it wasn't :) | |
| 16:45:27 | dtantsur | this happens and usually is quite frustrating indeed :) | |
| 18:17:59 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: Fix to_dict recursion issues with circular aliases https://review.openstack.org/589250 | |
| 18:18:19 | mordred | dtantsur|afk: ^^ that's a test that shows the breakage | |
| 19:14:31 | Leo_m | hi, where in git can I see the methods of each service of the connection object, for ex. conn.identity.users() where conn is the openstack sdk connection object | |
| 19:15:57 | mordred | Leo_m: that one is either in http://git.openstack.org/cgit/openstack/openstacksdk/tree/openstack/identity/v3/_proxy.py#n660 or http://git.openstack.org/cgit/openstack/openstacksdk/tree/openstack/identity/v2/_proxy.py#n250 | |
| 19:21:57 | Leo_m | mordred: thx! do you know how to set a timeout on those type of calls in case the cloud is unavailable? | |
| 19:23:23 | Leo_m | maybe in the clouds.yaml config, or the connection object creation or a method service call? | |
| 19:25:29 | mordred | Leo_m: you can set "api_timeout" in clouds.yaml - or you should be able to pass api_timeout to the Connection constructor | |
| 19:44:10 | Leo_m | mordred: thanks! that did it. btw, couldn't find all the params the clouds.yaml file can have, just found a couple of examples. | |
| 20:00:50 | mordred | Leo_m: yeah - sorry about that - it's an unfortunately long-lived bug that the docs for clouds.yaml are incomplete | |
| 21:15:10 | Leo_m | mordred: no worries, thanks for the feedback! | |