| *** gyee has quit IRC | 00:27 | |
| *** ianychoi has quit IRC | 01:17 | |
| *** ianychoi has joined #openstack-swift | 01:18 | |
| *** BjoernT has joined #openstack-swift | 02:47 | |
| *** psachin has joined #openstack-swift | 03:12 | |
| *** BjoernT has quit IRC | 03:14 | |
| *** BjoernT has joined #openstack-swift | 03:18 | |
| *** BjoernT has quit IRC | 03:29 | |
| *** spsurya has joined #openstack-swift | 03:29 | |
| *** gmann_afk is now known as gmann | 03:33 | |
| *** gkadam has joined #openstack-swift | 03:55 | |
| *** ianychoi has quit IRC | 03:59 | |
| *** ianychoi has joined #openstack-swift | 04:00 | |
| *** gkadam has quit IRC | 04:08 | |
| *** spsurya has quit IRC | 04:37 | |
| *** spsurya has joined #openstack-swift | 04:38 | |
| *** pcaruana has joined #openstack-swift | 05:01 | |
| *** ianychoi has quit IRC | 05:03 | |
| *** ianychoi has joined #openstack-swift | 05:05 | |
| *** ianychoi has quit IRC | 05:27 | |
| *** ianychoi has joined #openstack-swift | 05:29 | |
| *** ianychoi has quit IRC | 06:00 | |
| *** ianychoi has joined #openstack-swift | 06:01 | |
| *** ianychoi has quit IRC | 06:40 | |
| *** ianychoi has joined #openstack-swift | 06:42 | |
| *** tesseract has joined #openstack-swift | 07:16 | |
| *** rdejoux has joined #openstack-swift | 07:23 | |
| *** mikecmpbll has joined #openstack-swift | 07:58 | |
| *** Jeffrey4l has quit IRC | 07:59 | |
| *** Jeffrey4l has joined #openstack-swift | 08:00 | |
| *** rcernin has quit IRC | 08:02 | |
| *** rcernin has joined #openstack-swift | 08:03 | |
| *** tkajinam has quit IRC | 08:04 | |
| *** e0ne has joined #openstack-swift | 08:29 | |
| *** mikecmpbll has quit IRC | 10:04 | |
| *** mikecmpbll has joined #openstack-swift | 10:54 | |
| *** ianychoi has quit IRC | 12:10 | |
| *** ianychoi has joined #openstack-swift | 12:13 | |
| *** mikecmpbll has quit IRC | 12:19 | |
| *** mikecmpbll has joined #openstack-swift | 12:23 | |
| *** NM has joined #openstack-swift | 12:34 | |
| *** rcernin has quit IRC | 12:58 | |
| *** mikecmpbll has quit IRC | 13:00 | |
| *** mikecmpbll has joined #openstack-swift | 13:02 | |
| *** BjoernT has joined #openstack-swift | 13:26 | |
| *** NM has quit IRC | 13:50 | |
| *** NM has joined #openstack-swift | 14:07 | |
| *** BjoernT has quit IRC | 15:12 | |
| *** BjoernT has joined #openstack-swift | 15:18 | |
| *** mikecmpbll has quit IRC | 15:18 | |
| *** ianychoi has quit IRC | 15:20 | |
| *** ianychoi has joined #openstack-swift | 15:23 | |
| *** mikecmpbll has joined #openstack-swift | 15:25 | |
| *** tesseract has quit IRC | 15:33 | |
| *** gyee has joined #openstack-swift | 15:34 | |
| *** itlinux has joined #openstack-swift | 15:35 | |
| *** NM has quit IRC | 15:36 | |
| *** BjoernT has quit IRC | 15:39 | |
| *** mikecmpbll has quit IRC | 15:44 | |
| *** BjoernT has joined #openstack-swift | 15:44 | |
| *** BjoernT_ has joined #openstack-swift | 15:48 | |
| *** BjoernT has quit IRC | 15:49 | |
| *** rdejoux has quit IRC | 16:13 | |
| *** mikecmpbll has joined #openstack-swift | 16:13 | |
| *** e0ne has quit IRC | 16:18 | |
| *** BjoernT has joined #openstack-swift | 16:40 | |
| *** mikecmpbll has quit IRC | 16:41 | |
| *** BjoernT_ has quit IRC | 16:42 | |
| *** NM has joined #openstack-swift | 16:46 | |
| *** zaitcev_ has joined #openstack-swift | 17:11 | |
| *** ChanServ sets mode: +v zaitcev_ | 17:11 | |
| *** zaitcev has quit IRC | 17:15 | |
| *** e0ne has joined #openstack-swift | 17:35 | |
| *** NM has quit IRC | 18:00 | |
| *** NM has joined #openstack-swift | 18:05 | |
| *** psachin has quit IRC | 18:10 | |
| *** psachin has joined #openstack-swift | 18:13 | |
| *** e0ne has quit IRC | 18:17 | |
| *** psachin has quit IRC | 18:17 | |
| *** e0ne has joined #openstack-swift | 19:08 | |
| *** spsurya has quit IRC | 19:09 | |
| openstackgerrit | Tim Burke proposed openstack/swift master: Allow internal clients to use null namespace https://review.opendev.org/682138 | 19:20 |
|---|---|---|
| *** NM has quit IRC | 19:25 | |
| *** NM has joined #openstack-swift | 19:32 | |
| clayg | oh tim, I had at least one of those fixes with tests on my local machine - sorry I was so slow getting it pushed up 😞 | 19:47 |
| clayg | oh, you got some null tests too! well that's great! 😁 | 19:49 |
| *** NM has quit IRC | 19:50 | |
| clayg | subdir's might be tricky... if we have a null container 'foo/bar/\x00baz' but no container 'foo' - it's possible the correct prefix query params might return something that doesn't nesecarily have a null-byte in the truncated name 🤔 | 19:55 |
| *** NM has joined #openstack-swift | 19:55 | |
| timburke | clayg, yeah, i had that thought too... makes me glad we've been leaning towards putting versions in a separate container! ;-) | 19:59 |
| clayg | oh right, you can't make a container with a slash... but sure that the same might not apply to carefully selected prefix-delim tho | 20:00 |
| clayg | I was thinking the account & container servers that are stuffing in the truncated names for subdir might want to notate `internal: True` or something? | 20:01 |
| timburke | it gets tricky though -- what if the prefix represents a mix of normal and null-containing names? | 20:03 |
| clayg | well, we don't do the truncation in sql right - we could annotate it in python on a per entry basisis before sending it up to the proxy? | 20:04 |
| clayg | you've probably groked that code way more recently than i have | 20:05 |
| timburke | but what if the item that causes us to truncate has a null but there are other items with the shared prefix that are normal? or the other way around? because of the skip-jumps, we can't tell | 20:06 |
| clayg | on a side note - I recently finished "stranger in a strange land" (introduced the term "grok") - that book had way more sex and religion than you might expect for an early 60's scifi novel!? | 20:06 |
| clayg | @timburke oh shit - yeah I see - uhhhh | 20:06 |
| timburke | "don't do that" is the best answer i've got :P | 20:07 |
| clayg | yeah, idk, that kinda sucks | 20:07 |
| openstackgerrit | Tim Burke proposed openstack/swift master: Allow internal clients to use null namespace https://review.opendev.org/682138 | 20:20 |
| clayg | doesn't the sort order give us some kind of hint... `foo` would show up before `foo\x00` - annotating `subdirs` seems like it wouldn't hurt, unless it's worse than useless | 20:22 |
| clayg | maybe this doesn't matter since we don't allow null-namespace objects in public namespace containers | 20:23 |
| *** e0ne has quit IRC | 20:27 | |
| *** pcaruana has quit IRC | 20:33 | |
| clayg | @timburke so you understanding of the problem leads you to belive this doesn't work in the general sense: https://gist.github.com/clayg/1dd732407c98b27aa387322325b9b63b | 20:54 |
| clayg | because unless we mess with the markers too when we annotate the subdir it's possible that we'd filter someting that really *should* be there based on a non-null name that we'd have skipped passed | 20:55 |
| clayg | obviously the magic number enum for "is_subdir" and "is_internal_subdir" is a kludge | 20:56 |
| timburke | yes. for example, if you did a PUT to '/v1/AUTH_test/nullish' | 20:57 |
| timburke | one of the downsides of not having a proper column :-/ | 20:59 |
| clayg | well, this sucks | 21:00 |
| timburke | i'm under the impression that people don't really do delimiter queries on accounts -- i'm not gonna feel too bad if we have some rough edge there | 21:01 |
| timburke | and since nulls in *objects* are only allowed in null-containing containers.. | 21:01 |
| clayg | i will - this is supposed to be foundational - these kind of warts really stick out | 21:02 |
| clayg | what if we said you have to *start* with a null char? wasn't there some reason we wanted `foo` and `foo\x00versions` to appear near each other in listings? | 21:02 |
| timburke | i was debating about that earlier. it for sure doubles the number of listings we make, but we might be able to get away with it... | 21:03 |
| timburke | hey, i even called it out in https://etherpad.openstack.org/p/swift-object-versioning ! | 21:04 |
| timburke | > add "system containers" with a scheme like "<container name><NUL byte>versions" or "<NUL byte>versions-<container name>" | 21:05 |
| clayg | what if we never allow "subdiring" the null namespace - can we just add it to the entry and continue - i.e. never skip jump on a null entry? | 21:05 |
| *** NM has quit IRC | 21:05 | |
| clayg | i'm less concerned about what this means to api than the code - i'm happy to say the api for the null-namespace is whatever was easiest to implement as long as it's half way defensible and we can document expected behavior | 21:06 |
| timburke | i'd be worried about prefix/delimiter listings of versions...\ | 21:06 |
| clayg | it would be very nice if the null namespace behavior or listing accounts and containers were pretty similar | 21:07 |
| timburke | agreed | 21:08 |
| *** NM has joined #openstack-swift | 21:09 | |
| clayg | and I think we *need* `<key-name>\x00<version>` so the "only starts with \x00" isn't really going to fly, right? | 21:09 |
| timburke | \x00<key>\x00<version>? *shrug* | 21:17 |
| timburke | i'm starting to get nervous about other parts of the null namespace things, though... i was trying to put p 682382 on top of it, hit an issue, started adding debug logging... | 21:18 |
| patchbot | https://review.opendev.org/#/c/682382/ - swift - WIP: New Object Versioning mode - 9 patch sets | 21:18 |
| timburke | and saw logs like "Got '412 Precondition Failed' for PUT /v1/AUTH_test/d" -- switching a %s to a %r, i got "Got '412 Precondition Failed' for PUT '/v1/AUTH_test/d\x00versions' (txn: tx0176b3794c764ca1aef39-005d93c1e9)" | 21:19 |
| clayg | oh shit!? logging is being sneaky with null terminated strings!? | 21:20 |
| clayg | that's scary/annoying 😞 | 21:20 |
| clayg | everything is terrible - let's give up 😭 | 21:21 |
| timburke | i was half hoping that LogStringFormatter would quote it for us... nope | 21:23 |
| timburke | i'm still not entirely sure where it gets dropped, either -- in our rsyslog emitter? in rsyslog itself? in rsyslog's formatter? | 21:25 |
| *** NM has quit IRC | 21:28 | |
| *** openstackstatus has quit IRC | 21:32 | |
| *** openstackstatus has joined #openstack-swift | 21:35 | |
| *** ChanServ sets mode: +v openstackstatus | 21:35 | |
| timburke | clayg, swift-get-nodes is also proving problematic: http://paste.openstack.org/show/780647/ | 22:19 |
| timburke | that "bash: warning: command substitution: ignored null byte in input" is annoying me... but i'm pretty sure i get why they do it | 22:20 |
| timburke | reminds me of https://bugs.launchpad.net/swift/+bug/1791302 | 22:23 |
| openstack | Launchpad bug 1791302 in OpenStack Object Storage (swift) "swift-get-nodes cli don't handle space in arguments" [Undecided,Invalid] - Assigned to Guillaume (vatlas) | 22:23 |
| timburke | something like https://review.opendev.org/#/c/600789/ may yet be useful... though, presumably, toggled by an option | 22:24 |
| patchbot | patch 600789 - swift - URL decode swift-get-node command line arguments. ... (ABANDONED) - 5 patch sets | 22:24 |
| openstackgerrit | Tim Burke proposed openstack/swift master: PDF Documentation Build tox target https://review.opendev.org/679898 | 22:47 |
| *** tkajinam has joined #openstack-swift | 22:50 | |
| *** BjoernT has quit IRC | 23:04 | |
| *** BjoernT has joined #openstack-swift | 23:33 | |
| *** rcernin has joined #openstack-swift | 23:33 | |
| mattoliverau | morning | 23:36 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!