Earlier  
Later
New Search
Posted Nick Remark
#openstack-sdks - 2018-09-25
14:41:13 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Make RateLimitingTaskManager the TaskManager https://review.openstack.org/604520
14:41:14 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Replace TaskManager with a semaphore and lock https://review.openstack.org/604926
16:04:24 openstackgerrit Merged openstack/openstacksdk master: Allow search on containers https://review.openstack.org/600680
16:17:29 openstackgerrit Merged openstack/openstacksdk master: Normalize image when using PUT on Glance v2 https://review.openstack.org/602031
16:17:31 openstackgerrit Merged openstack/openstacksdk master: Run all tasks through the threadpool https://review.openstack.org/603739
16:40:44 samueldmq fwiw https://review.openstack.org/#/c/605081/ is a really quick review/merge
16:40:54 openstackgerrit Samuel de Medeiros Queiroz proposed openstack/openstacksdk master: Normalize network resources https://review.openstack.org/602218
16:50:31 mordred samueldmq: why yes it is!
16:50:56 samueldmq mordred: \o/
16:52:38 timburke mordred: samueldmq: fyi on the container headers sample -- looks like that's coming from radosgw; i wouldn't rely on there being a X-Container-Bytes-Used-Actual header
16:53:09 timburke up to you whether to expose it to users anyway, but it's not something swift would respond with
16:53:14 samueldmq timburke: hmm but X-Container-Bytes-Used should be there right?
16:53:19 timburke yup!
16:53:52 mordred timburke: thanks! yeah - I'd like the sdk interface be things swift sends
16:53:58 samueldmq kk we wont rely on radosgw. it is not even 100% compatible
16:54:13 samueldmq it returns me 200 OK when I do get on an object that does not exist
16:54:26 timburke hmm... :-/
16:54:32 mordred that seems like an unfortunate life choice for it
16:54:59 samueldmq I will open a bug on that project
16:55:07 timburke i've got my own beef with some of our choices of response code, but that's pretty bad
16:55:29 samueldmq but it's hard to say you're compatible with an api that you don't run that api's funcional tests
16:55:39 samueldmq especially in a big and dynamic project like opesntack :-)
16:55:48 mordred that said - some amount of compat code in sdk to work around such things on behalf of users is in-bounds for sdk I think - we work around some weird life choices in rax publiccloud, for instance - but we should definitely keep the interface to be what's in swift
16:56:20 samueldmq mordred: agreed. we should try to keep it within openstack world
16:56:28 mordred ++
16:56:37 samueldmq in a few days someone runs openstack api's with amazon underneath
16:56:46 samueldmq and we don't need to make shade work with that all
16:57:38 mordred yeah ... I don't have much personal desire for that thing to have success
16:58:17 samueldmq :-)
16:59:42 openstackgerrit Merged openstack/openstacksdk master: Fix location region field in docs https://review.openstack.org/605081
17:30:24 openstackgerrit Samuel de Medeiros Queiroz proposed openstack/openstacksdk master: Normalize subnet resources https://review.openstack.org/602228
17:42:17 samueldmq mordred: on the get container convo ...
17:42:37 samueldmq if I look up the headers and create an object including storage_policy
17:43:00 samueldmq an entity returned by get_container would be different than one returned by list_containers
17:43:19 samueldmq because the one from get would have more attrs (e.g storage_policy)
17:43:45 samueldmq which is a bit weird
17:50:46 samueldmq or is it okay if a GET returns more details than a list do?
18:16:54 openstackgerrit Samuel de Medeiros Queiroz proposed openstack/openstacksdk master: Allow returning munch on create and get container https://review.openstack.org/596546
18:18:14 openstackgerrit Samuel de Medeiros Queiroz proposed openstack/openstacksdk master: Normalize container resources https://review.openstack.org/602233
18:31:22 Shrews mordred: comment on https://review.openstack.org/#/c/604521/6/openstack/connection.py
18:34:59 openstackgerrit Samuel de Medeiros Queiroz proposed openstack/openstacksdk master: Normalize object resources https://review.openstack.org/602234
18:36:43 mordred Shrews: your comment is correct
18:37:45 Shrews well that's a first
18:41:41 openstackgerrit Merged openstack/keystoneauth master: Reformat Adapter docstring https://review.openstack.org/605042
18:49:16 openstackgerrit Monty Taylor proposed openstack-infra/shade master: Trim away the cover and py35 jobs https://review.openstack.org/605168
18:52:46 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Clean up python3 test and remove duplicate jobs https://review.openstack.org/604517
19:01:19 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Clean up python3 test and remove duplicate jobs https://review.openstack.org/604517
19:05:40 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Restore timeout_scaling_factor https://review.openstack.org/604628
19:34:35 openstackgerrit Matthias Lisin proposed openstack/openstacksdk master: Support firewall service for SDK https://review.openstack.org/592303
19:34:42 openstackgerrit Matthias Lisin proposed openstack/openstacksdk master: openstackcloud.py: Implement FWaaS wrapper methods. https://review.openstack.org/604811
22:05:38 openstackgerrit Samuel de Medeiros Queiroz proposed openstack/openstacksdk master: Update vendor support info for catalyst https://review.openstack.org/605245
22:11:22 openstackgerrit Samuel de Medeiros Queiroz proposed openstack/openstacksdk master: Update vendor support info for ecs https://review.openstack.org/605248
22:16:05 openstackgerrit Samuel de Medeiros Queiroz proposed openstack/openstacksdk master: Update vendor support info for switchengines https://review.openstack.org/605249
22:22:26 openstackgerrit Samuel de Medeiros Queiroz proposed openstack/openstacksdk master: Update vendor support info for vexxhost https://review.openstack.org/605252
22:30:46 openstackgerrit Samuel de Medeiros Queiroz proposed openstack/openstacksdk master: Add compute API info and fix provider names https://review.openstack.org/605253
22:31:48 samueldmq these 5 are very neat too ^ just updating the vendor-support info
23:08:49 openstackgerrit Merged openstack/openstacksdk master: Allow search on objects https://review.openstack.org/600683
#openstack-sdks - 2018-09-26
03:47:56 openstackgerrit LIU Yulong proposed openstack/openstacksdk master: Support firewall service for SDK https://review.openstack.org/592303
07:56:37 openstackgerrit Matthias Lisin proposed openstack/openstacksdk master: openstackcloud.py: Implement FWaaS wrapper methods. https://review.openstack.org/604811
11:13:22 openstackgerrit Chen proposed openstack/python-openstackclient master: Use 2.latest compute microversion as default https://review.openstack.org/605398
13:34:53 openstackgerrit Matthias Lisin proposed openstack/openstacksdk master: openstackcloud.py: Implement FWaaS wrapper methods. https://review.openstack.org/604811
15:07:09 openstackgerrit Merged openstack-infra/shade master: Trim away the cover and py35 jobs https://review.openstack.org/605168
15:07:10 openstackgerrit Merged openstack-infra/shade master: Remove the task manager https://review.openstack.org/605052
15:50:33 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Clean up python3 test and remove duplicate jobs https://review.openstack.org/604517
17:18:30 openstackgerrit Merged openstack/openstacksdk master: Restore timeout_scaling_factor https://review.openstack.org/604628
18:47:53 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Remove profile https://review.openstack.org/604414
18:47:54 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Use discovery instead of config to create proxies https://review.openstack.org/600292
18:47:54 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Set endpoint_override from endpoint with noauth https://review.openstack.org/604415
18:47:55 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Special case network and use it in openstack.cloud https://review.openstack.org/604645
18:47:55 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Remove api version default values https://review.openstack.org/600330
18:47:56 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Remove all the deprecated stuff https://review.openstack.org/605508
18:56:36 openstackgerrit Merged openstack/openstacksdk master: Update vendor support info for catalyst https://review.openstack.org/605245
19:05:46 openstackgerrit Merged openstack/openstacksdk master: Update vendor support info for switchengines https://review.openstack.org/605249
19:05:46 openstackgerrit Merged openstack/openstacksdk master: Update vendor support info for ecs https://review.openstack.org/605248
19:05:47 openstackgerrit Merged openstack/openstacksdk master: Update vendor support info for vexxhost https://review.openstack.org/605252
19:05:48 openstackgerrit Merged openstack/openstacksdk master: Add compute API info and fix provider names https://review.openstack.org/605253
19:08:04 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Replace TaskManager with a semaphore and lock https://review.openstack.org/604926
19:08:25 mordred timburke: thanks for the review on that ^^ I think that should fix the logic error - but I agree, I need to write a big pile of tests
19:09:23 timburke no problem. i remember an awful lot of staring and thinking while reviewing when we added a concurrency primitive to Swift (https://review.openstack.org/#/c/493636/)
19:10:20 timburke tests are going to be invaluable, but we'll still need to think a whole lot about all the possible ways things could get out of order
19:10:41 timburke (er, not "out of order" exactly... "interleaved")
19:11:16 mordred the most fun code are the classes where there are more comments and test than code :)
19:11:48 mordred ++ yeah - interleaved, blocked, etc
19:12:45 timburke fwiw, a co-worker of mine wrote https://github.com/swiftstack/python-stack-xray/blob/master/python-stack-xray that's proven invaluable when i suspect that i've hit a deadlock (particularly if eventlet is involved, too)
19:13:56 mordred all the eventlet/gevent stuff totally blows my mind
19:15:04 mordred timburke: seeing as how you have more eventlet experience than I do - my assumption with the semaphore/mutex code above is that it should work fine for eventlet callers too, right?
19:15:16 mordred (should probaby make an eventlet test case to test that out)
19:16:40 timburke probably? i'd need to think on it a bit more. i suspect it'll hurt if eventlet hasn't patched out threading and queue yet, though
19:17:54 mordred that's a good point
19:17:56 timburke (ie, one greenthread would wait on the semaphore and block the hub, so no other greenthreads in that os thread would be able to execute)
19:19:23 mordred https://eventlet.net/doc/modules/queue.html and https://eventlet.net/doc/modules/semaphore.html exist at least - but I don't know if that means they monkeypatch those or not
19:19:34 openstack Launchpad bug 1710328 in OpenStack Object Storage (swift) "object server deadlocks when a worker thread logs something" [High,Fix released] - Assigned to Samuel Merritt (torgomatic)
19:19:34 timburke oh... but if you've got a mix of greenthreads and os threads... you might get into a situation like https://bugs.launchpad.net/swift/+bug/1710328 ...
19:21:12 timburke i think the logic generally goes like this though: (1) don't mix greenthreads and os threads unless you know what you're doing. (2) no one knows what they're doing, and those that think they do are fooling themselves
19:21:29 mordred yes. I agree with both of those
19:21:58 Shrews hrm, this entire conversation leads to wonder if we should leave well enough alone with the TaskManager :)
19:22:05 mordred BUT - that bug does indicate that eventlet patches out threading.RLock
19:22:19 mordred so as long as the calling program is all threads or all mutexes life should be fine
19:23:28 mordred Shrews: I think we're fine - we're not spawning threads or eventlets ourselves - calling code _should_ be fine from either calling convention, and we're not adding threads to the mix ourselves
19:23:59 mordred Shrews: OR - we could just write the semaphore class in C++

Earlier   Later