| seongsoocho | morning~ | 00:05 |
|---|---|---|
| *** rdejoux has quit IRC | 00:46 | |
| *** rdejoux has joined #openstack-swift | 01:01 | |
| kota_ | morning~ | 02:15 |
| *** threestrands has joined #openstack-swift | 02:51 | |
| *** psachin has joined #openstack-swift | 03:31 | |
| *** evrardjp has quit IRC | 04:35 | |
| *** evrardjp has joined #openstack-swift | 04:35 | |
| f0o | morning | 07:14 |
| f0o | Is there a way to update an object while retaining it's metadata? | 07:14 |
| *** rpittau|afk is now known as rpittau | 07:22 | |
| *** ccamacho has joined #openstack-swift | 07:43 | |
| *** rcernin has quit IRC | 08:13 | |
| *** threestrands has quit IRC | 08:20 | |
| *** tkajinam has quit IRC | 08:23 | |
| *** dtantsur|afk is now known as dtantsur | 09:49 | |
| *** rpittau is now known as rpittau|bbl | 10:32 | |
| *** rledisez has quit IRC | 11:45 | |
| *** rledisez has joined #openstack-swift | 11:49 | |
| *** ChanServ sets mode: +v rledisez | 11:49 | |
| *** rledisez has quit IRC | 12:05 | |
| *** rledisez has joined #openstack-swift | 12:09 | |
| *** ChanServ sets mode: +v rledisez | 12:09 | |
| f0o | Also, I noticed that Large Files (with manifest) have a weird handling when using X-Delete-At/After Headers... I expected that the Manifest with X-Delete-At would also delete the partial objects that it references but it seems that only the Manifest is being delted. So how is the ideal way of expiring large objects? Do all partials need their X-Delete-* headers? How would those be kept | 12:16 |
| f0o | in-sync? | 12:16 |
| *** alecuyer has quit IRC | 12:25 | |
| *** alecuyer has joined #openstack-swift | 12:40 | |
| *** rpittau|bbl is now known as rpittau | 12:49 | |
| *** psachin has quit IRC | 13:35 | |
| *** tkajinam has joined #openstack-swift | 13:42 | |
| *** mikecmpbll has joined #openstack-swift | 13:52 | |
| *** irclogbot_2 has joined #openstack-swift | 14:09 | |
| *** irclogbot_2 has quit IRC | 14:13 | |
| *** irclogbot_2 has joined #openstack-swift | 14:24 | |
| *** irclogbot_2 has quit IRC | 14:25 | |
| *** irclogbot_2 has joined #openstack-swift | 14:27 | |
| *** irclogbot_2 has quit IRC | 14:29 | |
| *** irclogbot_2 has joined #openstack-swift | 14:31 | |
| *** manuvakery has joined #openstack-swift | 14:32 | |
| *** irclogbot_2 has quit IRC | 14:35 | |
| *** irclogbot_1 has joined #openstack-swift | 14:37 | |
| *** irclogbot_1 has quit IRC | 14:39 | |
| *** irclogbot_3 has joined #openstack-swift | 14:41 | |
| *** irclogbot_3 has quit IRC | 14:45 | |
| *** irclogbot_0 has joined #openstack-swift | 14:47 | |
| *** irclogbot_0 has quit IRC | 14:51 | |
| *** irclogbot_1 has joined #openstack-swift | 14:53 | |
| *** tkajinam has quit IRC | 14:54 | |
| *** irclogbot_1 has quit IRC | 14:55 | |
| *** irclogbot_0 has joined #openstack-swift | 14:58 | |
| *** irclogbot_0 has quit IRC | 14:59 | |
| *** irclogbot_2 has joined #openstack-swift | 15:10 | |
| *** irclogbot_2 has quit IRC | 15:11 | |
| *** irclogbot_2 has joined #openstack-swift | 15:18 | |
| *** irclogbot_2 has quit IRC | 15:19 | |
| *** irclogbot_2 has joined #openstack-swift | 15:38 | |
| clayg | tdasilva: do you remember anything about what we decided with ACL's on the new object versoining mode? | 15:49 |
| clayg | f0o: a metadata update to an object always replaced all metadata | 15:49 |
| clayg | f0o: yes you need to set expiration on any segments you want to expire | 15:49 |
| clayg | f0o: changing the metaata on a manifest won't change the metadata on any segements - there's no mechinism to keep the x-delete-at "in sync" between a manifest and segments | 15:50 |
| clayg | N.B. you can have a segment referenced from more than one manifests - they don't really "belong" to any multi-part object like s3 MPUs | 15:51 |
| f0o | Damn | 15:56 |
| f0o | I mean it all makes sense and it just kills my usecase right now haha | 15:57 |
| f0o | I was trying to pre-allocate object from a priviledged service and pass a tempurl to PUT data into that object to the client so it can directly upload files/content to it without needing permissions so to say | 15:58 |
| *** rpittau is now known as rpittau|afk | 16:07 | |
| *** alecuyer has quit IRC | 16:26 | |
| *** mikecmpbll has quit IRC | 16:26 | |
| *** hoonetorg has quit IRC | 16:26 | |
| *** noonedeadpunk has quit IRC | 16:26 | |
| *** tonyb has quit IRC | 16:26 | |
| *** irclogbot_2 has quit IRC | 16:28 | |
| *** irclogbot_1 has joined #openstack-swift | 16:30 | |
| *** hoonetorg has joined #openstack-swift | 16:31 | |
| *** mikecmpbll has joined #openstack-swift | 16:32 | |
| *** noonedeadpunk has joined #openstack-swift | 16:32 | |
| *** tonyb has joined #openstack-swift | 16:32 | |
| clayg | that sounds awesome! | 16:32 |
| clayg | I guess the problem with PUT tempurls is you can't set x-delete-at, or just that you can't do SLOs with PUT tempurls at all? 🤔 | 16:33 |
| DHE | I've been using tempurl PUTs with x-delete-at for a while now. let me look... | 16:34 |
| f0o | well I could do X-* but that would allow the client to override TTL and such :/ hence why I hoped the preallocation would solve that | 16:34 |
| f0o | I think I will end up writing a midleware to either re-set the TTL or allow adding data to objects without replacing them | 16:35 |
| *** evrardjp has quit IRC | 16:35 | |
| DHE | for non-slo/dlo objects my PUT works just fine with delete | 16:36 |
| f0o | I mean it all works right now that way, it just wont expire the objects haha. A super cheap way to solve it is to configue with the manifest route and have a cron checking those and delete the linked objects (semantic naming here) when they're gone - but that's unsatisfying hehe | 16:37 |
| f0o | continue* typoday... mondays... | 16:37 |
| openstackgerrit | Alexandre Lécuyer proposed openstack/swift feature/losf: New key format for objects in the index-server https://review.opendev.org/723609 | 16:39 |
| *** ChanServ has quit IRC | 16:42 | |
| *** ChanServ has joined #openstack-swift | 16:45 | |
| *** tepper.freenode.net sets mode: +o ChanServ | 16:45 | |
| *** evrardjp has joined #openstack-swift | 16:46 | |
| *** alecuyer has joined #openstack-swift | 16:46 | |
| *** dtantsur is now known as dtantsur|afk | 17:18 | |
| *** noonedeadpunk has quit IRC | 17:26 | |
| *** tonyb has quit IRC | 17:26 | |
| *** ccamel has quit IRC | 17:29 | |
| *** noonedeadpunk has joined #openstack-swift | 17:29 | |
| *** tonyb has joined #openstack-swift | 17:29 | |
| *** ChanServ has quit IRC | 17:39 | |
| *** ChanServ has joined #openstack-swift | 17:42 | |
| *** tepper.freenode.net sets mode: +o ChanServ | 17:42 | |
| *** manuvakery has quit IRC | 17:42 | |
| *** camelCaser has joined #openstack-swift | 17:42 | |
| *** ccamel has joined #openstack-swift | 17:48 | |
| *** camelCaser has quit IRC | 17:49 | |
| *** hoonetorg has quit IRC | 18:02 | |
| f0o | ok so I ended up implementing a cronjob on the service side. Abusing the fact that an object can also be a directory... I hope this never changes haha | 18:18 |
| f0o | basically I got /foo as object with x-delete-after/at and the client gets a tempurl to upload to /foo/blob and it will put files in there. the cronjob will then recursively delete /foo/ if /foo is expired or deleted | 18:19 |
| *** rdejoux has quit IRC | 18:19 | |
| f0o | at least now it still works with upstream swift and is vendor neutral as oposed to implementing my own middleware haha | 18:19 |
| f0o | So tempurl is evaluated when the whole request is done it seems... it takes me 1.3 minutes to upload 2G and the tempurl expires in 1m. This means my upload will never be accepted... This is good to know but also slightly annoying because the request was made before the TTL expires just not parsed by swift-proxy in time | 18:36 |
| DHE | are you doing 100-continue? | 18:43 |
| f0o | no just a plain PUT with 2G data | 18:43 |
| DHE | I am, and while I haven't run into this kind of issue (I have much smaller objects) the 100-continue response fails if my headers are bad | 18:43 |
| f0o | I can see that the PUT gets sent in full and then the request shows up in the proxy log with a 4xx error as if the tempurl expired | 18:44 |
| f0o | which is technically correct as by the time the log entry is generated the ttl has expired. but at the time of request init it isnt expired yet, so I was expecting it to pass regardless of the end of the data stream arriving after expiry | 18:45 |
| *** rdejoux has joined #openstack-swift | 18:45 | |
| f0o | I think we should parse the headers early and not wait for the full data to be loaded as this varies grately depending on bandwith availability from clients | 18:46 |
| f0o | I dont think signing longer TTLs (which would be up to 10+m here for 2G data) is a good solution as it would allow any bearer to just override the data if it gets leaked | 18:46 |
| *** mikecmpbll has quit IRC | 19:35 | |
| *** mikecmpbll has joined #openstack-swift | 19:37 | |
| *** ccamacho has quit IRC | 20:38 | |
| *** rcernin has joined #openstack-swift | 21:14 | |
| openstackgerrit | Tim Burke proposed openstack/swift master: Use separate name for HeaderKeyDict var vs list of response headers https://review.opendev.org/723736 | 22:20 |
| timburke | ^^^ WIP; needs tests | 22:20 |
| *** tkajinam has joined #openstack-swift | 22:49 | |
| *** tkajinam has quit IRC | 22:49 | |
| *** tkajinam has joined #openstack-swift | 22:50 | |
| mattoliverau | morning | 23:04 |
| openstackgerrit | Merged openstack/swift master: obj: _finalize_durable may succeed even when data file is missing https://review.opendev.org/722128 | 23:26 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!