| *** jistr has quit IRC | 00:11 | |
| *** jistr has joined #openstack-swift | 00:16 | |
| *** EmilienM has quit IRC | 00:30 | |
| *** tovin07_ has joined #openstack-swift | 00:43 | |
| *** gyee has quit IRC | 00:49 | |
| *** two_tired has joined #openstack-swift | 01:06 | |
| clayg | timburke: acoles: be on the lookout for torgomatic ‘s fix for op bug #1741371 | 01:23 |
|---|---|---|
| openstack | bug 1741371 in OpenStack Object Storage (swift) "Expiring objects have skewed X-Delete-At values" [Undecided,New] https://launchpad.net/bugs/1741371 | 01:23 |
| mattoliverau | did he decide which way he was going to take it? req timestamp or drop x-delete-after? | 02:06 |
| *** threestrands_ has joined #openstack-swift | 02:14 | |
| *** threestrands_ has quit IRC | 02:14 | |
| *** threestrands_ has joined #openstack-swift | 02:14 | |
| *** awkwardpenguin has joined #openstack-swift | 02:15 | |
| *** awkwardpenguin has quit IRC | 02:15 | |
| *** threestrands has quit IRC | 02:16 | |
| kota_ | good morning Swift world and a happy new year! | 02:18 |
| kota_ | I had been completely off until today. | 02:19 |
| * kota_ is booting his head | 02:19 | |
| mattoliverau | kota_: morning and happy new year! | 02:22 |
| kota_ | mattoliverau: happy new year! | 02:24 |
| kota_ | it looks like others have started new year work | 02:25 |
| mattoliverau | yeah, but still seems a little quiet. I assume everyones still getting back up to speed :) | 02:25 |
| kota_ | mattoliverau: ;) | 02:33 |
| *** JimCheung has quit IRC | 02:44 | |
| *** links has joined #openstack-swift | 03:29 | |
| *** links has quit IRC | 03:34 | |
| *** links has joined #openstack-swift | 03:56 | |
| *** psachin has joined #openstack-swift | 04:33 | |
| openstackgerrit | Samuel Merritt proposed openstack/swift master: Fix time skew when using X-Delete-After https://review.openstack.org/531290 | 04:38 |
| openstackgerrit | Samuel Merritt proposed openstack/swift master: Ignore directory .stestr https://review.openstack.org/531291 | 04:41 |
| *** two_tired has quit IRC | 04:55 | |
| *** ianychoi has quit IRC | 05:52 | |
| *** threestrands_ has quit IRC | 06:13 | |
| *** geaaru has quit IRC | 06:33 | |
| *** armaan_ has joined #openstack-swift | 06:54 | |
| *** hseipp has joined #openstack-swift | 07:39 | |
| openstackgerrit | Merged openstack/swift feature/deep: Set swiftclient log level to WARNING for sharding probe tests https://review.openstack.org/530900 | 07:52 |
| *** geaaru has joined #openstack-swift | 08:08 | |
| *** tesseract has joined #openstack-swift | 08:12 | |
| *** armaan_ has quit IRC | 08:15 | |
| *** armaan has joined #openstack-swift | 08:16 | |
| *** armaan has quit IRC | 08:16 | |
| *** rcernin has quit IRC | 08:28 | |
| *** armaan has joined #openstack-swift | 08:31 | |
| *** silor has joined #openstack-swift | 08:51 | |
| acoles | good morning | 09:05 |
| acoles | kota_: happy new year! | 09:05 |
| * acoles imagines kota booting his head and thinks kota must have very flexible joints :) | 09:06 | |
| openstackgerrit | Matthew Oliver proposed openstack/swift feature/deep: Use _root.db and _shard_db on sharded containers https://review.openstack.org/531350 | 10:09 |
| mattoliverau | acoles: morning | 10:09 |
| mattoliverau | opps already see a commit message typo.. oh well. | 10:09 |
| acoles | mattoliverau: hi, thanks for that ^ and your review comments | 10:09 |
| mattoliverau | ^ that is still WIP and I still want to look at adding some more tests. But it's getting late and wanted to push what I had so far | 10:10 |
| acoles | +1 for pushing WIP. have a good weekend | 10:10 |
| mattoliverau | acoles: you too, have a great friday too ;) | 10:14 |
| acoles | mattoliverau: I hope to have something on the shard range state idea soon. TBH, I suspect that eventually we may need both that and some timestamp based analyser to cope with potentially competing sharders. But I think the state base scheme will get to us a tool that can we can start to use in 'controlled' environments. | 10:15 |
| *** tovin07_ has quit IRC | 10:18 | |
| openstackgerrit | Alistair Coles proposed openstack/swift feature/deep: fixes for sharding/shrinking races https://review.openstack.org/529931 | 10:27 |
| openstackgerrit | Alistair Coles proposed openstack/swift feature/deep: Add includes method to ShardRange https://review.openstack.org/530901 | 10:27 |
| openstackgerrit | Alistair Coles proposed openstack/swift feature/deep: Re-introduce shard range updates to root after shard has sharded https://review.openstack.org/530902 | 10:27 |
| *** szaher has joined #openstack-swift | 10:54 | |
| *** pcaruana has joined #openstack-swift | 11:02 | |
| acoles | I'm looking at p 531290 | 11:05 |
| patchbot | https://review.openstack.org/#/c/531290/ - swift - Fix time skew when using X-Delete-After | 11:05 |
| *** kong has quit IRC | 12:23 | |
| *** hoonetorg has quit IRC | 12:57 | |
| *** silor has quit IRC | 13:08 | |
| *** psachin has quit IRC | 13:09 | |
| *** silor has joined #openstack-swift | 13:30 | |
| *** silor has quit IRC | 13:32 | |
| *** silor has joined #openstack-swift | 13:35 | |
| *** links has quit IRC | 13:47 | |
| *** hseipp has quit IRC | 13:48 | |
| *** hseipp has joined #openstack-swift | 13:48 | |
| *** tonyb has quit IRC | 13:49 | |
| *** tonyb has joined #openstack-swift | 13:50 | |
| *** jistr is now known as jistr|biab | 13:51 | |
| openstackgerrit | Alistair Coles proposed openstack/swift master: Document that x-delete-after takes precedence over x-delete-at https://review.openstack.org/531384 | 14:50 |
| *** jistr|biab is now known as jistr | 15:04 | |
| *** hoonetorg has joined #openstack-swift | 15:10 | |
| *** two_tired has joined #openstack-swift | 15:33 | |
| *** two_tired has quit IRC | 15:41 | |
| openstackgerrit | Alistair Coles proposed openstack/swift master: Tighten up test_check_delete_headers https://review.openstack.org/531403 | 15:59 |
| openstackgerrit | Alistair Coles proposed openstack/swift master: Correct 400 response message when x-delete-after is zero https://review.openstack.org/531404 | 15:59 |
| *** mtreinish has quit IRC | 16:08 | |
| *** mtreinish has joined #openstack-swift | 16:09 | |
| *** openstackstatus has quit IRC | 16:40 | |
| *** openstackstatus has joined #openstack-swift | 16:42 | |
| *** ChanServ sets mode: +v openstackstatus | 16:42 | |
| *** armaan has quit IRC | 16:57 | |
| *** armaan has joined #openstack-swift | 16:58 | |
| timburke | good morning | 17:02 |
| acoles | timburke: o/ | 17:05 |
| *** joeljwright has joined #openstack-swift | 17:10 | |
| *** ChanServ sets mode: +v joeljwright | 17:10 | |
| *** pcaruana has quit IRC | 17:21 | |
| *** hseipp has quit IRC | 17:26 | |
| *** joeljwright has quit IRC | 17:35 | |
| *** JimCheung has joined #openstack-swift | 17:36 | |
| *** early has quit IRC | 17:42 | |
| openstackgerrit | Alistair Coles proposed openstack/swift master: Use _update_x_timestamp method in object controller DELETE method https://review.openstack.org/531447 | 17:43 |
| *** early has joined #openstack-swift | 17:45 | |
| openstackgerrit | Alistair Coles proposed openstack/swift master: Document that x-delete-after takes precedence over x-delete-at https://review.openstack.org/531384 | 17:45 |
| timburke | acoles: the more i think about this _update_x_timestamp(req) business, the more i feel like it should happen up in the proxy server app (maybe as part of update_request?) | 17:50 |
| timburke | *every* request ends up needing to set it at some point, yeah? may as well centralize that | 17:51 |
| acoles | timburke: heh. I spent while looking for the place in the proxy server where x-timestamp *is* set | 17:51 |
| acoles | the current *one* place is https://github.com/openstack/swift/blob/5c564e53966b74d6d7589c3505dd33816b3c19f1/swift/proxy/controllers/base.py#L1510-L1534 | 17:51 |
| acoles | but the obj PUT path doesn't call that IIRC, so does its own thing | 17:52 |
| timburke | ...and meanwhile we've discovered a variety of places where, actually, it'd be kinda nice if it was already set... | 17:54 |
| timburke | i know i've run into that with swift3 before, too | 17:54 |
| timburke | maybe the One True Place to set X-Timestamp ought to be in gatekeeper, near https://github.com/openstack/swift/blob/2.16.0/swift/common/middleware/gatekeeper.py#L85-L90 | 17:54 |
| acoles | I suspect the history behind the special cases is to do with container sync - the comment is a big hint at that | 17:54 |
| acoles | but yeah it doesn't seem very coherent | 17:55 |
| timburke | ...except then we'd still need it in the proxy server for internal clients... | 17:55 |
| torgomatic | sounds like proxy __call__ is the place to do it, and then we can stop worrying | 17:56 |
| torgomatic | I think you guys are right, and the current weirdness is an artifact of how container sync used to be | 17:56 |
| acoles | torgomatic: let's not muddle with it in your x-delete-after patch though...that looks close to being done | 17:58 |
| torgomatic | acoles: agreed | 17:58 |
| timburke | sounds good. proxy __call__ will be good, but i think gatekeeper might also be nice -- then we wouldn't need to carry some of the legacy of https://github.com/openstack/swift3/commit/d8ffe5a into s3api | 18:02 |
| torgomatic | I think I'm going to squash in patch 531403 but leave the other two cleanups to stand on their own unless anyone objects | 18:02 |
| patchbot | https://review.openstack.org/#/c/531403/ - swift - Tighten up test_check_delete_headers | 18:02 |
| acoles | torgomatic: +1 ^^ that is the only one I felt should be included, the other stuff can follow on | 18:03 |
| acoles | timburke: both? then internal client is happy | 18:03 |
| timburke | acoles: that's what i was thinking | 18:03 |
| openstackgerrit | Samuel Merritt proposed openstack/swift master: Fix time skew when using X-Delete-After https://review.openstack.org/531290 | 18:15 |
| torgomatic | acoles: timburke: there's the patch ^^ with the test fixes included | 18:17 |
| openstackgerrit | Merged openstack/swift master: Ignore directory .stestr https://review.openstack.org/531291 | 18:23 |
| *** armaan has quit IRC | 18:44 | |
| *** armaan has joined #openstack-swift | 18:47 | |
| torgomatic | heh, the bug in patch 531290 is slightly more annoying than I thought... if you're running a big EC scheme, say 12+3, you've got 15 fragment archives, each of which could have its own X-Delete-At | 19:08 |
| patchbot | https://review.openstack.org/#/c/531290/ - swift - Fix time skew when using X-Delete-After | 19:08 |
| torgomatic | however, only 3 of them update the expirer queue, so 12 of them may end up with X-Delete-At values that aren't reflected in the expirer queue | 19:09 |
| torgomatic | so worst case, you have 3 fragment archives deleted by the expirer and 12 that can't be; they will be cleaned up by the reconstructor, but that's much slower | 19:10 |
| *** openstack has joined #openstack-swift | 21:15 | |
| *** ChanServ sets mode: +o openstack | 21:15 | |
| torgomatic | clayg: that is a very nice property to have | 21:20 |
| clayg | yup - wtg ssync! | 21:20 |
| *** tonyb has quit IRC | 21:21 | |
| *** vinsh has joined #openstack-swift | 21:25 | |
| *** mweshi has joined #openstack-swift | 21:26 | |
| *** mweshi_ has joined #openstack-swift | 21:39 | |
| openstackgerrit | Merged openstack/swift master: Document that x-delete-after takes precedence over x-delete-at https://review.openstack.org/531384 | 21:52 |
| timburke | ahahaha CORS is such a mess, especially as it interacts with swift... i started looking through the spec and noticed something about "redirect steps"... so i start messing with symlinks, the only place that we currently *can* have redirects sent to the client (as i recall) | 21:58 |
| timburke | i try a cross-domain post to a symlink... and it works! 307, browser follows, 202! | 21:58 |
| clayg | timburke: good find | 21:59 |
| timburke | but i wasn't actually setting any headers... if i try sending a x-object-meta-* header, i'm no longer on the happy "Simple Cross-Origin Request" path | 22:00 |
| timburke | so i get my preflight request... which 200s with appropriate headers... so i POST, get back a 307... and then the browser balks: CORS header ‘Access-Control-Allow-Origin’ missing | 22:03 |
| timburke | because symlink swallows the proxy-server response and sends back its own | 22:04 |
| timburke | but even if i *fix* that... | 22:04 |
| *** mweshi_ has quit IRC | 22:05 | |
| *** mweshi has quit IRC | 22:05 | |
| timburke | all i've done it prevent the warning. i still just see the preflight & POST to the symlink | 22:05 |
| timburke | and i don't get anything useful in the browser, not even something to tell me "oh, hey, i got a 307" | 22:06 |
| timburke | because in the non-simple case, you follow "If the response has an HTTP status code of 301, 302, 303, 307, or 308: Apply the cache and network error steps." | 22:07 |
| timburke | so, what can we POST and still stay on the happy path? | 22:09 |
| timburke | Content-Type. but only if it's one of application/x-www-form-urlencoded, multipart/form-data, or text/plain | 22:09 |
| timburke | (also, Accept, Accept-Language, or Content-Language -- but none of those stick to an object) | 22:11 |
| *** mweshi_ has joined #openstack-swift | 22:12 | |
| *** mweshi has joined #openstack-swift | 22:13 | |
| *** armaan_ has quit IRC | 22:14 | |
| timburke | oh yeah, and /info requests are busted -- no access-control-allow-methods header | 22:17 |
| *** mweshi has quit IRC | 22:17 | |
| timburke | no, never mind... that parts fine. i messed up my request | 22:21 |
| timburke | but if you go off the happy path, preflight fails (no allow-origin) | 22:22 |
| timburke | that's... maybe reasonable? *shrug* | 22:24 |
| *** mweshi_ has quit IRC | 22:31 | |
| timburke | still not sure what to do for symlink... i guess this is maybe a reason to bring back ?symlink=post ... but how likely is this to really come up? you'd need a pretty full-featured object-store browser that's a web app served from a separate box... but why wouldn't you just put that on your proxies so it scales out with your cluster? | 22:38 |
| timburke | and we *already* make it difficult to do this sort of thing -- see https://bugs.launchpad.net/swift/+bug/1418530 | 22:39 |
| openstack | Launchpad bug 1418530 in OpenStack Object Storage (swift) "TempAuth and CORS is not working" [Low,Confirmed] | 22:39 |
| timburke | (and think about the fact that unless you make the container you're manipulating fully public, you'll necessarily be on the preflight path because of x-auth-token) | 22:40 |
| *** mweshi has joined #openstack-swift | 22:47 | |
| *** awkwardpenguin has joined #openstack-swift | 23:02 | |
| *** mweshi has quit IRC | 23:24 | |
| *** gyee has quit IRC | 23:35 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!