Earlier  
Later
New Search
Posted Nick Remark
#openstack-sdks - 2018-07-19
20:18:06 mordred corvus: should I wait for his response before sending email?
20:21:06 corvus mordred: yeah, maybe a few mins
20:21:09 mordred kk
20:23:31 corvus the docs i linked are old and don't exist anymore. the closest analog is https://docs.openstack.org/swift/latest/middleware.html#staticweb
20:23:59 corvus that doesn't appear to be any less ambiguous
20:27:15 mordred corvus: agree. it mentions directory marker objects but does not indicate that they are required or why you'd want them
20:27:23 mordred corvus: maybe they ARE required in swift for listings
20:27:38 mordred corvus: but not for index
20:27:49 mordred corvus: since the doc section is in the listings section
20:35:30 corvus mordred: apparently not? at least not according to notmyname. i pasted the relevant parts of my convo with him in the etherpad
20:36:04 corvus tldr -- it seems like they are mostly leftover from earlier stuff and not needed anymore
20:36:41 mordred cool
20:37:08 mordred SO - it seems like it would be a correct action for radosgw to stop requiring them
20:37:18 corvus yep i think that's the best way forward
20:41:33 mordred corvus: hows' that look now?
20:42:20 corvus mordred: lgtm!
20:47:49 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Add flag for disabling object checksum generation https://review.openstack.org/584044
20:47:49 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Add ability to pass data to create_object https://review.openstack.org/584045
20:47:50 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Add create_directory_marker_object method https://review.openstack.org/584050
22:08:24 openstackgerrit Merged openstack-infra/shade master: Use valid filters to list floating IPs in neutron https://review.openstack.org/584076
23:04:21 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Add create_directory_marker_object method https://review.openstack.org/584050
#openstack-sdks - 2018-07-20
00:33:24 openstackgerrit Sean McGinnis proposed openstack-infra/shade master: Remove redundant target in README https://review.openstack.org/584169
00:53:19 openstackgerrit Sean McGinnis proposed openstack-infra/shade master: Remove redundant target in README https://review.openstack.org/584169
01:10:02 openstackgerrit Merged openstack-infra/shade master: Remove redundant target in README https://review.openstack.org/584169
02:17:31 openstackgerrit Merged openstack/openstacksdk master: Use valid filters to list floating IPs in neutron https://review.openstack.org/584029
06:04:44 openstackgerrit Benoît Knecht proposed openstack/python-openstackclient master: compute: host: expand kwargs in host_set() call https://review.openstack.org/583225
06:33:55 openstackgerrit Merged openstack/openstacksdk master: Add flag for disabling object checksum generation https://review.openstack.org/584044
06:33:56 openstackgerrit Merged openstack/openstacksdk master: Add ability to pass data to create_object https://review.openstack.org/584045
06:33:57 openstackgerrit Merged openstack/openstacksdk master: Add create_directory_marker_object method https://review.openstack.org/584050
08:38:17 openstackgerrit Benoît Knecht proposed openstack/python-openstackclient master: compute: host: expand kwargs in host_set() call https://review.openstack.org/583225
11:44:47 openstackgerrit Dmitry Tantsur proposed openstack/openstacksdk master: baremetal: correct the default timeout in Node.set_provision_state https://review.openstack.org/584331
11:44:53 dtantsur mordred: sorry for this ^^^
12:54:57 mordred dtantsur|brb: :)
15:11:06 mordred dtantsur: we're gonna have to stable-branch backport that change (it didn't make the freeze cutoff)
15:11:24 mordred corvus: 0.17.0 of sdk has been released which has those new methods we added
15:11:40 corvus mordred: sweet, i'll try 'em out soon
15:11:47 dtantsur ack, I'll propose it
15:12:52 mordred dtantsur: (I don't know if stable/rocky has been cut yet - I think maybe that's today?)
15:13:23 dtantsur mordred: the person releasing openstacksdk should propose that to releases repo
15:14:33 dtantsur mordred: on a related note, which deadline affects openstacksdk: the non-client yesterday or the client next week?
15:14:48 dtantsur I would expect openstacksdk to follow the same cadence as python-XXXclient stuff
15:15:19 mordred dtantsur: non-client
15:15:25 dtantsur that's weird to me
15:15:34 mordred dtantsur: it gets lumped in to that because there are python-*client libs thatuse it
15:15:41 dtantsur puts openstacksdk in a worse position wrt to catching up with features
15:15:52 dtantsur like, API additions can land till next week.
15:16:00 dtantsur but openstacksdk has no chances of exposing them
15:16:10 mordred yeah - I think it's maybe something we should talk about next cycle as we start to use sdk directly for more and more of the projects
15:16:35 mordred dtantsur: good point
15:16:46 dtantsur also for non-client you was supposed to create stable/rocky yesterday with 0.17 :)
15:16:53 dtantsur * for non-client deadline
15:21:32 mordred oh. piddle.
15:21:40 mordred I'm not used to these stable branch things ...
15:25:23 openstackgerrit Merged openstack/openstacksdk master: baremetal: correct the default timeout in Node.set_provision_state https://review.openstack.org/584331
15:27:21 dtantsur mordred: if you don't know, this is how it looks https://review.openstack.org/#/c/584021/3/deliverables/rocky/sushy.yaml
15:33:25 mordred dtantsur: yeah - I just sent one up
15:33:29 mordred and thanks!
15:33:34 dtroyer osc-lib has the same issue as sdk wrt non-client freeze, I agree it is probably time to re-think that and move them to client-lib category
15:33:40 dtroyer for stein
15:42:03 mordred ++
15:42:48 mordred dtroyer: speaking of - we're maybe 2 patches away from 1.0 - so I think targetting that for stein should be *totally* doable
15:42:53 mordred (for sdk)
15:43:33 mordred dtroyer: in fact, I'm hoping I can get those written during freeze-time ... so hopefully we can cut 1.0 very early in the stein cycle
15:43:41 dtroyer \o/ \O/ \o/
15:44:13 mordred dtroyer: wehave a few more weeks before osc itself is froze, yeah?
15:46:20 dtroyer next week I think, we've considered it a client lib due to other project dependencies
15:46:41 dtroyer so moving osc-lib and sdk would lump them all together but I think that is OK
15:47:24 dtroyer I was hoping to have more of the backlog in OSC merged but I'm planning to cut the to-be-stable release tuesday
15:48:15 mordred dtroyer: kk. I've got a fairly simple command I wanted to add - I'll try to get that done today
15:49:14 dtroyer mordred: sounds good
15:50:44 corvus mordred: is there a way to get the requests.Session() that keystoneclient/openstacksdk use?
15:51:07 mordred corvus: yes
15:51:43 mordred corvus: conn.config.get_session()
15:53:13 corvus mordred: perfect, thanks!
15:53:49 corvus (i wanted to set the max poolsize for the connection pool, since i'm multithreaded; this will allow more connection reuse)
15:53:50 mordred corvus: out of curiosity - are you wanting to use it to make a rest call against one of the openstack endpoints?
15:54:02 mordred ah!
15:54:04 corvus sess = cloud.config.get_session()
15:54:04 corvus adapter = requests.adapters.HTTPAdapter(pool_maxsize=100)
15:54:04 corvus sess.mount('https://', adapter)
15:54:23 mordred neat. that should totally work
15:54:28 corvus it does!
15:54:42 mordred corvus: do you think that's an option we should expose in openstack.config ?
15:54:45 corvus no more warnings about the pool being full
15:54:48 mordred (not today - just in general)
15:56:00 corvus mordred: perhaps -- might be a nice friendly thing -- there's some subtlety about how connection pooling works for multiple hosts that i skimmed over because i'm only talking to one host. would probably need to read up a bit more on that before implementing something.
15:56:19 mordred corvus: ++
15:57:07 corvus aiui from my quick reading, there's only one pool by default, so if you access a second host, it drops the pool for the first host. but in addition to pool_maxsize (max connections in pool), there's also pool_connections -- number of pools
15:57:36 corvus probably the thing to do would be to expose both of those. but like i said, i skipped some details so i'm not 100% sure
15:58:14 mordred yeah - especially since for openstack it's extrememly common for each service to have its own host - and for sdk to use the same session for all of them
15:58:31 corvus fwiw, changing the pool size did not greatly impact the total amount of time spent uploading objects to swift. but it still makes me feel better. :)
15:59:19 mordred corvus: we could maybe get fancy and look in the catalog, count the number of distinct hosts contained in it and use that to set a default for pool_connections
15:59:41 corvus mordred: may want more for the multi-cloud case...
15:59:53 corvus oh, this is a per-cloud session anyway
15:59:57 mordred corvus: multi-cloud case would ... yeah
15:59:58 corvus so yeah, your idea could work
16:00:10 corvus unless redirects are involved?
16:00:19 corvus (heh, just double it! :)
16:00:34 mordred well, if they are involved, we can still expose the setting :)
17:56:32 PagliaccisCloud hey guys i'm trying to fix a typo in the help menu on python-openstackclient. to create a bug, do i make a new story on storyboard?

Earlier   Later