Monday, 2019-10-28

*** nanzha has joined #openstack-swift01:22
*** rcernin has quit IRC02:10
*** ianychoi__ has joined #openstack-swift02:38
*** ianychoi_ has quit IRC02:41
*** nanzha has quit IRC03:11
*** nanzha has joined #openstack-swift03:21
*** diablo_rojo has quit IRC03:37
openstackgerritzhangboye proposed openstack/python-swiftclient master: Stop testing python2.7
*** gkadam has joined #openstack-swift07:19
*** gkadam has quit IRC07:19
*** tkajinam has quit IRC08:04
*** lifeless has joined #openstack-swift08:05
*** tesseract has joined #openstack-swift08:13
*** rdejoux has joined #openstack-swift08:28
*** rpittau|afk is now known as rpittau08:30
*** ccamacho has joined #openstack-swift08:38
*** mikecmpbll has joined #openstack-swift08:55
*** nanzha has quit IRC09:13
*** nanzha has joined #openstack-swift09:15
alecuyerbinwiederhier: thanks for sharing!09:21
*** root____1 has joined #openstack-swift09:28
*** nanzha has quit IRC11:38
*** nanzha has joined #openstack-swift11:43
*** nanzha has quit IRC12:05
rledisezgood morning, don't know who commented on but I added some notes/answers to the comments12:49
*** diablo_rojo has joined #openstack-swift13:34
*** zaitcev__ has joined #openstack-swift13:44
*** ChanServ sets mode: +v zaitcev__13:44
*** zaitcev_ has quit IRC13:48
binwiederhieralecuyer: i know it's very much the opposite of what you guys have been doing, so i hope i didn't step on your guys' toes. it's just a PoC after all, just wanted to try if it was possible.14:00
binwiederhierThat said, my thing obviously has drawbacks, such as that the application has to be packing-aware.14:01
openstackgerritThiago da Silva proposed openstack/swift master: WIP: New Object Versioning mode
timburkerledisez, that was me -- sorry, i should have left my nick15:12
timburkebinwiederhier, yeah, as i was looking at it, i realized that the name you use for GET is different from what you use for PUT -- it's not how swift usually behaves, but it definitely opens up a lot of possibilities for us15:19
timburkefwiw, i realized that we're *kind of* going to be doing that with the object versioning work -- client does a put to a name, server responds with a version id, and then the client can always get that particular data by requesting /v1/<account>/<container>/<name>?version-id=<id> (even if /v1/<account>/<container>/<name> is subsequently overwritten)15:19
binwiederhierIf this was implemented in swift or as middleware, you could keep the path-to-pack relationship in the container db, but i'm sure then it wouldn't be as simple as it is right now anymore.15:25
*** gyee has joined #openstack-swift15:25
binwiederhierLarge objects are also "different" for the client, so there is definitely precedent (not that that's good though).15:27
alecuyerbinwiederhier: don't worry about my toes ;) It's a cool idea, I was also wondering about having it directly in swift, or in the client library. (if you can handle the X-Item-Path there, with properties on the container/packed object, "transparent" small files support :) )15:54
*** zaitcev__ is now known as zaitcev15:56
zaitcevI had my concerns about the external DB, but they were mostly misgivings. CouchDB seems to work for Ceph.15:56
zaitcevFunny thing though, one day Sage decided to get away from using a filesystem and start consuming raw block devices, and oops, you have to mod the DB then.15:57
gmanntimburke: ping16:17
openstackgerritClay Gerrard proposed openstack/swift master: use request_helpers import namespace for tests
tdasilvatimburke: do you remember why we added a separate s3api func test job to the gate as opposed to just enabling s3api on the other tests (e.g., func-encryption) ?16:36
timburkegmann, hi!16:36
timburketdasilva, no i don't :-(16:37
gmanntimburke: i would like to sync up with the py2.7 drop plan for swift . L22  -
gmanntimburke:  you might have seen my mail and this etherpad.16:38
gmannbasically, there is no issue in keeping the support. swift can keep the support but we would like to know what all projects/lib/testing tool etc swift might need the support of py2.716:39
gmanncommon lib like oslo or testing tooling like devstack, tempest will drop the support from milestone-1 to milestone-2 (as mentioned in schedule section of the etherpad)16:40
timburkesounds good -- the main things we'd be interested in would be devstack (which we use to test our keystone support) and keystonemiddleware (and all of its dependencies)16:41
timburkenote that devstack doesn't need to have all services running under py2 -- we'd be content with soemthing like the long-standing previous behavior where all services *except swift* were running py316:42
timburkekeystonemiddleware could probably be addressed with a version cap and/or a py2-specific upper-constraints file16:43
timburkei don't want to keep projects from dropping py2 support, i just want to make sure we don't find ourselves unable to test swift on py2 :-)16:44
* gmann ok. I will make note for devstack. how about client, keystoneclient and openstackclient are also being used for testing - 16:44
gmann ok. I will make note for devstack. how about client, keystoneclient and openstackclient are also being used for testing -16:45
gmannthose also can be handled via version cap16:45
timburkehmmm... and castellan and barbicanclient are also integration points...16:45
gmannbut capping the version for py2.7 need g-r support which we can check with prometheanfire16:46
timburkei don't care too much about clients. i'll probably end up -2'ing but that's because we own it. if some other client wants to drop py2, i've got no objections16:47
patchbotpatch 691608 - python-swiftclient - Stop testing python2.7 - 1 patch set16:47
timburkeor rather, i don't care too much about *what language* clients *are implemented in* ;-) i care very much about not breaking clients16:47
*** rpittau is now known as rpittau|afk16:48
gmannyeah but there might be possibility of having them break on py2.7. so either we can go when it happen and cap the version for py2.716:49
gmannbut i do not know how to handle that from g-r side as it has to be done only for swift (or any other projects keeping py2.7 )16:49
gmannmay be we can discuss on #openstack-requirements channel with prometheanfire16:50
timburkeas i recall, openstackclient is just being used to configure keystone to work with swift and create users and such. i could double check that. keystoneclient comes in as a transitive dep for keystonemiddleware and even (iirc) used to provide the middleware16:56
timburkefwiw, castellan and barbicanclient are *untested* integration points -- part of how slipped by us :-/16:59
openstackLaunchpad bug 1847755 in OpenStack Object Storage (swift) "On-disk encryption is borken with Python 3" [High,Fix released]16:59
timburkei really need to find some time to get barbican up and getting tested in our dsvm jobs16:59
timburkegotta get to a meeting -- back in a bit. or we can talk more about it async17:02
gmanntimburke: ok. I will put that on ML with requirements team input also, thanks17:04
*** pcaruana has joined #openstack-swift17:31
openstackgerritThiago da Silva proposed openstack/swift master: WIP: New Object Versioning mode
*** mikecmpbll has quit IRC17:33
*** rdejoux has quit IRC17:46
*** henriqueof has quit IRC17:47
openstackgerritGilles Biannic proposed openstack/swift master: probe: Add test for syncing a delete when the remote 404s
openstackgerritGilles Biannic proposed openstack/swift master: Stop retrying every deletes in container-sync
*** pcaruana has quit IRC18:30
*** pcaruana has joined #openstack-swift18:31
*** tesseract has quit IRC18:31
*** Buggys has quit IRC18:45
*** StevenK has quit IRC18:45
claygtdasilva: you got the encryption stuff fixed up!?  Or ... we're just not running those tests in the gate?18:46
*** StevenK has joined #openstack-swift18:46
*** openstackgerrit has quit IRC18:50
tdasilvaclayg: there was an issue in `handle_post` when encryption is enabled that should be fixed in the latest patchset. I've also enabled object versioning in the encryption gate jobs since it was running before19:19
clayg!!! 😍19:20
openstackclayg: Error: "!!" is not a valid command.19:20
tdasilvaclayg, timburke: which actually brings me to a question i've been pondering about. Should the new object versioning mode be enabled by default19:20
tdasilvawe currently have the option in the middleware, but it's 'false' by default. I'm wondering if we should keep the option, but have it enabled by default19:20
claygtdasilva: good question - we should talk about that on wednesday and the PTG - preference?19:21
claygi mean s3api is *really* going to want it on - but I guess as long as the option is there we'll have to check19:21
tdasilvano preference, but I'm going to start a list of open questions on the ehterpad, maybe it can be discussed at the PTG...19:21
tdasilvaclayg: right, even more reason to have it on by default i think... /me needs to think more of the drawbacks19:22
claygan etherpad would be GREAT!  Thank you!19:23
*** renich has joined #openstack-swift19:26
*** pcaruana has quit IRC20:02
*** openstackgerrit has joined #openstack-swift20:12
openstackgerritThiago da Silva proposed openstack/swift master: WIP: Enabled s3api tests across all func tests
timburkei kinda want to have a swift-tox-func-minimal (with basically no feaures enabled, and maybe even just one user) and swift-tox-func-maximal (with basically *everything* enabled)20:23
timburkeas much as anything, the minimal would be there to ensure that we have appropriate feature checks in place against /info20:24
timburkestill gotta have separate jobs for non-client-facing things though -- EC vs replicated, encryption on/off...20:25
timburkemaybe we could do something with periodic jobs to make sure we still have some coverage to ensure that, for example, s3api doesn't *absolutely require* that you enable slo...20:26
tdasilvatimburke: yeah, defining the test matrix is tricky because it can get out of hand quickly20:34
openstackgerritMerged openstack/swift master: use request_helpers import namespace for tests
tdasilvatimburke: my current plan is to absorb s3api into the other tests as I'm not sure why we would need a separate job just for that, then maybe staticweb is the same?20:35
tdasilvaagree about separate encryption/ec jobs20:36
*** umbSublime has quit IRC21:01
*** diablo_rojo has quit IRC21:45
*** diablo_rojo has joined #openstack-swift22:00
*** timburke_ has joined #openstack-swift22:19
*** ChanServ sets mode: +v timburke_22:19
*** timburke_ has quit IRC22:20
openstackgerritClay Gerrard proposed openstack/swift master: WIP: Allow internal clients to use reserved namespace
openstackgerritClay Gerrard proposed openstack/swift master: WIP: New Object Versioning mode
openstackgerritClay Gerrard proposed openstack/swift master: WIP: s3api: Implement object versioning API
*** rcernin has joined #openstack-swift22:58
*** tkajinam has joined #openstack-swift23:00
*** mvkr has joined #openstack-swift23:13
*** diablo_rojo has quit IRC23:27
*** gyee has quit IRC23:57

Generated by 2.15.3 by Marius Gedminas - find it at!