| *** donnyd has quit IRC | 00:18 | |
| *** tdasilva has quit IRC | 00:18 | |
| *** tdasilva has joined #openstack-swift | 00:19 | |
| *** ChanServ sets mode: +v tdasilva | 00:19 | |
| *** donnyd has joined #openstack-swift | 00:20 | |
| *** rpittau|afk has joined #openstack-swift | 00:35 | |
| *** rpittau|afk has quit IRC | 00:42 | |
| *** tdasilva has quit IRC | 00:42 | |
| *** clayg has quit IRC | 00:42 | |
| *** donnyd has quit IRC | 00:43 | |
| *** gmann has quit IRC | 00:43 | |
| *** seongsoocho has quit IRC | 00:44 | |
| *** tdasilva has joined #openstack-swift | 00:44 | |
| *** ChanServ sets mode: +v tdasilva | 00:44 | |
| *** beisner has quit IRC | 00:45 | |
| *** abelur has quit IRC | 00:46 | |
| *** coreycb has quit IRC | 00:46 | |
| *** jrosser has quit IRC | 00:48 | |
| *** nottrobin has quit IRC | 00:48 | |
| *** openstackgerrit has joined #openstack-swift | 00:53 | |
| openstackgerrit | Tim Burke proposed openstack/swift master: py3: Fix formpost unicode filename issues https://review.opendev.org/701497 | 00:53 |
|---|---|---|
| *** nottrobin has joined #openstack-swift | 00:54 | |
| openstackgerrit | Merged openstack/python-swiftclient master: Add test for bulk-delete-attempt-counter fix https://review.opendev.org/694635 | 00:58 |
| *** nottrobin has quit IRC | 00:58 | |
| timburke | hmm... https://linux.conf.au/schedule/presentation/137/ | 00:59 |
| timburke | sadly, i'm not sure 22 is really that high for us :-/ | 01:00 |
| openstackgerrit | Tim Burke proposed openstack/swift master: Middleware that allows a user to have quoted Etags https://review.opendev.org/700056 | 01:19 |
| *** gyee has quit IRC | 01:30 | |
| timburke | :-( i don't like the logic here, seems dangerous: https://github.com/openstack/swift/blob/2.23.0/swift/container/sync.py#L406-L412 | 02:25 |
| timburke | what if the replicator rsyncs the db while container-sync has run sync_point2 ahead? | 02:25 |
| timburke | also, i feel like https://docs.openstack.org/swift/latest/overview_container_sync.html should maybe have something about one-way sync | 02:27 |
| openstackgerrit | Tim Burke proposed openstack/swift master: Clean up container-sync docs https://review.opendev.org/702782 | 02:32 |
| openstackgerrit | Tim Burke proposed openstack/swift master: container-sync: Stop continuing to update sync point 2 after failures https://review.opendev.org/702783 | 02:42 |
| *** renich has quit IRC | 03:06 | |
| *** psachin has joined #openstack-swift | 03:36 | |
| *** openstackgerrit has quit IRC | 04:12 | |
| *** openstackgerrit has joined #openstack-swift | 04:16 | |
| openstackgerrit | Merged openstack/swift master: Clean up container-sync docs https://review.opendev.org/702782 | 04:16 |
| *** tkajinam has quit IRC | 04:39 | |
| *** tkajinam has joined #openstack-swift | 04:41 | |
| *** tkajinam_ has joined #openstack-swift | 05:03 | |
| *** tkajinam has quit IRC | 05:07 | |
| *** evrardjp has quit IRC | 05:34 | |
| *** evrardjp has joined #openstack-swift | 05:34 | |
| openstackgerrit | Thiago da Silva proposed openstack/swift master: Skip container-sync when versioning enabled https://review.opendev.org/698139 | 06:23 |
| *** tdasilva has quit IRC | 06:23 | |
| *** gmann has joined #openstack-swift | 06:49 | |
| *** abelur has joined #openstack-swift | 06:49 | |
| *** clayg has joined #openstack-swift | 06:50 | |
| *** ChanServ sets mode: +v clayg | 06:50 | |
| *** donnyd has joined #openstack-swift | 06:52 | |
| *** beisner has joined #openstack-swift | 06:52 | |
| *** seongsoocho has joined #openstack-swift | 06:58 | |
| *** pcaruana has joined #openstack-swift | 07:00 | |
| *** nottrobin has joined #openstack-swift | 07:00 | |
| *** coreycb has joined #openstack-swift | 07:03 | |
| *** jrosser has joined #openstack-swift | 07:03 | |
| seongsoocho | Hi, Is X-Delete-At support millisecond unixtimestamp ?? My customer upload object with X-Delete-At: 1579753284000 (it is saved.. ) and got 503 error during download object. | 07:17 |
| seongsoocho | The error is below | 07:17 |
| seongsoocho | File "/usr/lib/python2.7/site-packages/swift/obj/diskfile.py", line 2594, in _verify_data_file | 07:17 |
| seongsoocho | if x_delete_at <= current_time and not self._open_expired: | 07:17 |
| seongsoocho | File "/usr/lib64/python2.7/functools.py", line 58, in <lambda> | 07:17 |
| seongsoocho | ('__ge__', lambda self, other: not self < other)], | 07:17 |
| seongsoocho | File "/usr/lib/python2.7/site-packages/swift/common/utils.py", line 1371, in __lt__ | 07:17 |
| seongsoocho | other = Timestamp(other) | 07:17 |
| seongsoocho | File "/usr/lib/python2.7/site-packages/swift/common/utils.py", line 1283, in __init__ | 07:17 |
| seongsoocho | raise ValueError('timestamp too large') | 07:17 |
| seongsoocho | I think normalize_delete_at_timestamp does not affect when uploading object. | 07:18 |
| seongsoocho | In ocata version, download also work. but in train version, download fail.. | 07:18 |
| *** rpittau|afk has joined #openstack-swift | 07:19 | |
| *** rcernin has quit IRC | 07:25 | |
| *** tdasilva has joined #openstack-swift | 07:47 | |
| *** ChanServ sets mode: +v tdasilva | 07:47 | |
| *** tesseract has joined #openstack-swift | 08:10 | |
| *** rdejoux has joined #openstack-swift | 08:15 | |
| *** rpittau|afk is now known as rpittau | 08:31 | |
| *** ccamacho has joined #openstack-swift | 09:29 | |
| kota_ | let me check | 09:53 |
| seongsoocho | kota_: thanks~! | 09:53 |
| kota_ | https://github.com/openstack/swift/commit/e6194113a3c81563590eabf8f761ccb988bb917c | 09:57 |
| kota_ | it looks like since 2.6.0 we have the cap. | 09:59 |
| kota_ | so it means Ocata should affect the same issue... | 10:00 |
| kota_ | exactly it should be checked at uploading tho | 10:00 |
| kota_ | ah... I should go lookup diskfile impl too. it might be changed the timestamp handling... | 10:02 |
| kota_ | seongsoocho: that's what https://github.com/openstack/swift/commit/7a7677868d4ec7693132014779909ac70d4f391d probably what you're looking for. | 10:08 |
| kota_ | since Rocky, the object expire timestamp uses swift.common.utils.Timestamp, that was capped the max value with https://github.com/openstack/swift/commit/e6194113a3c81563590eabf8f761ccb988bb917c | 10:09 |
| seongsoocho | kota_: yes. That's it. And I think proxy-server should use normalized delete at timestamp when pass a request to object-server. | 10:10 |
| seongsoocho | https://github.com/openstack/swift/blob/master/swift/proxy/controllers/obj.py#L526 | 10:10 |
| kota_ | and it looks the patch (https://github.com/openstack/swift/commit/7a7677868d4ec7693132014779909ac70) seems to touch only object server so it didn't touch the x-delete-at value assertions | 10:11 |
| kota_ | yup | 10:11 |
| seongsoocho | So, I have no idea how to recover a object which is stored with large integer of x-delete-at. | 10:12 |
| kota_ | https://github.com/openstack/swift/blob/master/swift/common/utils.py#L1496-L1516 :/ | 10:13 |
| kota_ | i think we should think a couple of things | 10:14 |
| kota_ | 1. how to fix the code, probably the float cast should be changed to Timestamp object | 10:14 |
| kota_ | 2. how to recover the existing objects | 10:14 |
| *** rpittau is now known as rpittau|bbl | 10:21 | |
| seongsoocho | :-( I will regist this on launchpad bugs. | 10:21 |
| kota_ | ok. I'm still curious, the commit says "use Timestamp" but I didn't find the logic yet. | 10:29 |
| kota_ | ah... it says just "use x-timestamp" not Timestamp object | 10:34 |
| kota_ | got it | 10:36 |
| kota_ | https://github.com/openstack/swift/commit/7a7677868d4ec7693132014779909ac70d4f391d#diff-5d16e75b6728dcfa2f94aae89b3ffd08R959-R971 | 10:37 |
| kota_ | valid_timestamp refers the Timestamp object from request object so the current_time should be Timestamp object then the `=<` triggers to translate larger x-delete-at integer to Timestamp | 10:39 |
| kota_ | and bom... | 10:39 |
| kota_ | seongsoocho: https://gist.github.com/bloodeagle40234/7dd81e1cc663a79014c539c8f3203a91 perhaps, this is one solution (sorry, not tested) like this. specialize the expire_date. honesty I don't like the specialization but save existing objects should be a most priority thing, my policy. | 10:46 |
| seongsoocho | kota_: oh !! thanks . I will test your patch in my dev env. | 10:47 |
| seongsoocho | It's been a long day ... (Because of this issue. 😢 ) | 10:52 |
| kota_ | yeah, and we woke up early for the meeting ;) | 10:56 |
| kota_ | please save yourself seongsoocho. I'm heading for home. | 10:57 |
| seongsoocho | kota_: Thanks. Take your time. | 10:58 |
| *** rpittau|bbl is now known as rpittau | 12:25 | |
| *** manuvakery has joined #openstack-swift | 12:57 | |
| manuvakery | > | 13:18 |
| *** ianychoi_ has quit IRC | 13:44 | |
| *** diablo_rojo has joined #openstack-swift | 14:04 | |
| *** diablo_rojo has quit IRC | 14:15 | |
| *** diablo_rojo has joined #openstack-swift | 14:16 | |
| *** ianychoi has joined #openstack-swift | 14:25 | |
| *** psachin has quit IRC | 14:26 | |
| *** pcaruana has quit IRC | 14:33 | |
| *** mvkr has quit IRC | 15:33 | |
| *** gyee has joined #openstack-swift | 15:34 | |
| *** mvkr has joined #openstack-swift | 15:48 | |
| openstackgerrit | David Sariel proposed openstack/swift master: Let developers/operators add watchers to object audit https://review.opendev.org/212824 | 16:02 |
| openstackgerrit | David Sariel proposed openstack/swift master: WIP: Add a reference implementation of audit watchers https://review.opendev.org/701139 | 16:02 |
| *** ianychoi_ has joined #openstack-swift | 16:05 | |
| *** ianychoi has quit IRC | 16:08 | |
| *** mvkr has quit IRC | 16:20 | |
| *** rpittau is now known as rpittau|afk | 16:27 | |
| *** mvkr has joined #openstack-swift | 16:32 | |
| *** tesseract has quit IRC | 16:39 | |
| *** psachin has joined #openstack-swift | 16:54 | |
| *** rdejoux has quit IRC | 16:56 | |
| *** evrardjp has quit IRC | 17:34 | |
| *** evrardjp has joined #openstack-swift | 17:34 | |
| timburke | sorry seongsoocho, i feel like i keep breaking things for you 😢 | 17:46 |
| openstackgerrit | Tim Burke proposed openstack/swift master: Allow Timestamp comparisons against out-of-range values https://review.opendev.org/702950 | 18:18 |
| timburke | seongsoocho, i *think* ^^^ should mostly get things readable? i've got another issue i need to dig into, but i'll see if i can repro your case before you wake up ;-) | 18:20 |
| rledisez | timburke: I was sure I did a bugreport (but apparently not), but the same issue arise in SSYNC replication. Example here: | 18:31 |
| rledisez | https://bugs.launchpad.net/swift/+bug/1821204 (look the stack, the bugreport is about something else) | 18:31 |
| openstack | Launchpad bug 1821204 in OpenStack Object Storage (swift) "SSYNC subrequests does not log the body of the error" [Undecided,New] | 18:31 |
| *** psachin has quit IRC | 18:46 | |
| *** jvisser has joined #openstack-swift | 18:50 | |
| *** openstackgerrit has quit IRC | 18:58 | |
| *** ccamacho has quit IRC | 19:18 | |
| timburke | rledisez, good news is, i'm pretty sure my patch will address that particular problem, too :) | 20:16 |
| timburke | seongsoocho, repro'ed your issue and confirmed that the Timestamp patch seems to address it! | 20:25 |
| *** rdejoux has joined #openstack-swift | 20:33 | |
| *** rdejoux has quit IRC | 20:37 | |
| *** rdejoux has joined #openstack-swift | 20:39 | |
| *** jvisser has quit IRC | 20:54 | |
| timburke | lol! we have no upper-bound on x-delete-at timestamp | 21:24 |
| timburke | when i was testing, i commented out the bound check in Timestamp, but didn't actually need to | 21:24 |
| *** jvisser has joined #openstack-swift | 21:31 | |
| *** jvisser has quit IRC | 21:33 | |
| *** jvisser has joined #openstack-swift | 21:33 | |
| *** jvisser has quit IRC | 21:34 | |
| *** jvisser has joined #openstack-swift | 21:34 | |
| *** rdejoux has quit IRC | 21:42 | |
| *** rcernin has joined #openstack-swift | 22:07 | |
| *** jvisser has quit IRC | 22:21 | |
| timburke | tdasilva, i think i'm gonna squash in https://review.opendev.org/#/c/701643/ (once i fix up the unit test failure) | 22:31 |
| patchbot | patch 701643 - swift - squash: flesh out object-versioning probe test a bit - 2 patch sets | 22:31 |
| timburke | any objections? | 22:31 |
| *** rdejoux has joined #openstack-swift | 22:40 | |
| tdasilva | timburke: nope :) | 22:41 |
| *** rdejoux has quit IRC | 22:47 | |
| timburke | hmm... we don't carry container read ACLs over between primary and archive containers, do we... | 22:50 |
| tdasilva | timburke: i´m torn on https://review.opendev.org/#/c/698139/10/swift/container/sync.py@607 I'm not sure I like the idea of symlink_override as it would seem to break the api contract. | 23:14 |
| patchbot | patch 698139 - swift - Skip container-sync when versioning enabled - 11 patch sets | 23:14 |
| tdasilva | your concern about changing our minds at a later point is that it would require a new PUT/POST to sync it, right? | 23:14 |
| timburke | yup | 23:16 |
| timburke | idk that it breaks the cotnract any more than container-sync breaks contracts for slo though | 23:16 |
| timburke | tdasilva, i'm scratching my head a little over https://review.opendev.org/#/c/698139/11/test/probe/test_container_sync.py@698 -- we wrote twice, but only to one name, right? why'd we end up with more than one version getting synced? i feel like i must be missing something... | 23:19 |
| patchbot | patch 698139 - swift - Skip container-sync when versioning enabled - 11 patch sets | 23:19 |
| tdasilva | timburke: it's because we allow enabling versioning on destination containers | 23:22 |
| tdasilva | see line 714 | 23:22 |
| timburke | but where'd the extra row come from for the source db? | 23:23 |
| tdasilva | not sure I follow you, but in _test_sync the sequence goes: write v1 > sync > write v2 > sync > confirm destination has 2 versions | 23:25 |
| timburke | 👍 turns out i can't read. missed the cycle in between PUTs | 23:28 |
| tdasilva | timburke: Users should not that not every overwrite might be versioned on the destionation container, it would depend if sync had a chance to make a pass | 23:28 |
| timburke | yup | 23:28 |
| tdasilva | s/Users shoud not/Users should note | 23:28 |
| tdasilva | timburke: we could write a v3 to call that out: v1 > sync > v2 v3 > sync > confirm destination has 2 versions | 23:29 |
| timburke | i think i maybe like http://paste.openstack.org/show/788509/ -- the override plumbing was pretty tame... | 23:44 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!