*** lyrrad has quit IRC | 00:23 | |
*** acoles_ is now known as acoles | 08:16 | |
*** chsc has joined #swift3 | 14:20 | |
*** chsc has quit IRC | 14:20 | |
*** chsc has joined #swift3 | 14:20 | |
*** chsc_ has joined #swift3 | 15:09 | |
*** chsc has quit IRC | 15:10 | |
*** lyrrad has joined #swift3 | 15:52 | |
*** chsc_ has quit IRC | 16:53 | |
*** chsc has joined #swift3 | 17:44 | |
*** chsc has quit IRC | 17:44 | |
*** chsc has joined #swift3 | 17:44 | |
*** acoles is now known as acoles_ | 21:05 | |
*** acoles_ is now known as acoles | 21:07 | |
*** acoles is now known as acoles_ | 21:20 | |
*** acoles_ is now known as acoles | 21:21 | |
*** acoles is now known as acoles_ | 21:22 | |
*** acoles_ is now known as acoles | 21:24 | |
*** acoles is now known as acoles_ | 21:30 | |
*** acoles_ is now known as acoles | 21:31 | |
*** acoles is now known as acoles_ | 22:14 | |
kota_ | hello | 23:01 |
---|---|---|
kota_ | anyone else here? | 23:02 |
kota_ | hmm.. it seems better to skip today. | 23:04 |
kota_ | adressing to the next week. | 23:05 |
kota_ | looking at the logs, timburke already done to rebase existing patch but we are now forcusing on upstream swift crypto-review. | 23:06 |
kota_ | probably sig4 follow up is priority one though. | 23:06 |
timburke | bah, sorry | 23:06 |
timburke | yeah, i think crypto is starting to take priority for me | 23:06 |
kota_ | tiburke: :-) | 23:07 |
kota_ | tim* | 23:07 |
timburke | fwiw, i did a bit of follow up on https://review.openstack.org/#/c/331291/ and saw... | 23:08 |
patchbot | timburke: patch 331291 - swift3 - Don't try to read a body if the client didn't send... | 23:08 |
timburke | where'd i put it... | 23:08 |
kota_ | looking | 23:09 |
timburke | no, i'm just looking on my machine :-) | 23:09 |
kota_ | ah, yeah. that one. | 23:09 |
timburke | so trying to do a bulk delete with no body gave me a 400 / MalformedPOSTRequest | 23:10 |
kota_ | and s3 actually works for no body? | 23:11 |
timburke | curiously, if i did it without a content type set, it responded with 412 / PreconditionFailed (??) | 23:11 |
timburke | for bucket PUT, yes | 23:11 |
timburke | still haven't tried the other places (ACLs, multipart uploads) | 23:11 |
kota_ | oh what? bulk delete is a *POST* right? | 23:12 |
timburke | i expect them to behave more like bulk delete | 23:12 |
timburke | yep. but it also uses .xml()/.read() | 23:12 |
timburke | which was where things blew up | 23:12 |
kota_ | k, let me resolve the tangled ones. | 23:13 |
kota_ | so | 23:13 |
kota_ | bulk delete with no body -> 400 MalformedRequest | 23:13 |
kota_ | this seems correct, right? | 23:14 |
timburke | yep | 23:14 |
*** chsc has quit IRC | 23:14 | |
kota_ | bulk delete with no body without content type -> 412 PreconditionFaild | 23:15 |
kota_ | weird but possible if the content-type validated before the body validation. | 23:15 |
kota_ | maybe? | 23:15 |
timburke | yep. which reminds me of the 412s you can get from versioned_writes -- both seem like they don't really fit for a 412, but whatever | 23:16 |
timburke | definitely client error, so as long as it's a 4xx i suppose... | 23:16 |
kota_ | ah, ok. that is one we should fix. | 23:16 |
kota_ | and patch 331291 seems a different one to me, right? | 23:17 |
patchbot | kota_: https://review.openstack.org/#/c/331291/ - swift3 - Don't try to read a body if the client didn't send... | 23:17 |
kota_ | that hits an eventlet bug which is able to raise TypeError unexpectedly. | 23:17 |
kota_ | if the request have neither content-length nor transfer-encoding. | 23:18 |
kota_ | *has | 23:18 |
timburke | well, they're related. i need to make that patch more specific. i suppose another option would be to require eventlet>=0.18.0 (which includes https://github.com/eventlet/eventlet/commit/b443de3) | 23:19 |
timburke | i *think* that would also fix the issue i saw, but i'm not sure/haven't tested | 23:20 |
kota_ | yup, it *looks* to fix the problem. | 23:21 |
kota_ | thanks for the link. | 23:21 |
timburke | the only hesitation in requiring a newer eventlet is that our min swift version only requires eventlet>=0.9.15 | 23:22 |
timburke | ...which may be an argument for upping the min swift version... | 23:22 |
notmyname | FWIW, when https://github.com/eventlet/eventlet/pull/326 lands, I'll want to bump swift's version to use that | 23:23 |
timburke | even master only wants eventlet>=0.17.4 though | 23:23 |
kota_ | newer swift has larger min eventlet version, right now? | 23:23 |
kota_ | now looking at swift repo | 23:23 |
timburke | notmyname: thanks for the input. good to know | 23:23 |
kota_ | eventlet>=0.17.4 | 23:24 |
kota_ | now | 23:24 |
kota_ | notmyname: oh, thanks | 23:24 |
kota_ | hmm | 23:25 |
timburke | yep. so currently, nothing else requires 0.18.0 or later... which makes me more inclined to work around the bug, at least for a while | 23:25 |
timburke | (fwiw, 0.17.4 is what my company ships with its product, and the version on which i noticed the bug) | 23:26 |
kota_ | well | 23:27 |
kota_ | right now, I don't have strong opinion for this | 23:28 |
kota_ | but that's ok either to fix this with 0.17.4 or waiting 0.18.0 | 23:29 |
kota_ | but anyway, I'm realzing it may be a time to consider min swit version for swift3 in this cycle. | 23:29 |
timburke | fair enough. there's not a big rush; i still need to check the behavior for the other req.xml callers | 23:30 |
kota_ | bump mim version of swift, i mean. | 23:30 |
timburke | yeah, not a bad idea | 23:30 |
kota_ | i know we have some places to be compatible with older swift. | 23:30 |
timburke | fwiw, it looks like 2.3.0 was the last kilo release, 2.5.0 the last liberty | 23:32 |
kota_ | i'd like to set the new min version as... the min version which is not eol yet at that time. | 23:33 |
kota_ | 2.5.0 seems reasonable for now. | 23:34 |
kota_ | or | 23:34 |
kota_ | oh, timburke, when your change for object listing reverse mereged? | 23:34 |
timburke | lemme look... | 23:35 |
timburke | 2.6.0 | 23:35 |
kota_ | bah | 23:35 |
timburke | start of mitaka | 23:35 |
timburke | any particular place we'd want to use it? | 23:36 |
kota_ | maybe? | 23:36 |
kota_ | IIRC, though. | 23:36 |
timburke | the main thing i can think of is versioning, which really demands something like https://review.openstack.org/#/c/214922/ | 23:37 |
patchbot | timburke: patch 214922 - swift - Add "history" mode to versioned_writes middleware | 23:37 |
kota_ | yes | 23:38 |
kota_ | that one was top of swift community dash board until starting crypto-review :P | 23:39 |
timburke | yep :-) but still few reviews | 23:39 |
kota_ | hah, yeah | 23:40 |
timburke | it felt like a fairly large patch until i started going through the crypto stuff :P | 23:40 |
kota_ | and also versioned_writes are now complicated due to the implementation history | 23:41 |
kota_ | I also keep a patch for versioned_writes though | 23:42 |
kota_ | that was affected from copy middleware change... :/ | 23:43 |
timburke | yeah, that was somewhat unfortunate. at least it's smart enough to try the reverse listing first then gracefully revert | 23:43 |
timburke | ah, yeah... i'd nearly forgotten about that one | 23:43 |
kota_ | patch 209315 | 23:43 |
patchbot | kota_: https://review.openstack.org/#/c/209315/ - swift - Refactor versioned_writes middleware | 23:43 |
kota_ | right now, failed at only functests !?!? | 23:44 |
kota_ | missing branch at unit!?!? | 23:44 |
kota_ | that's the place I've stopped. | 23:44 |
kota_ | anyway, we could take a time to rework/review after crypto work. | 23:45 |
timburke | oh, weird... i guess we must be missing some coverage in the unit tests... yeah, i'll take a look after crypto | 23:46 |
kota_ | i'm expecting, not so big middleware pipeline change happen after that soon | 23:46 |
kota_ | or, i failed to rebase unit conflict :P | 23:46 |
timburke | oh, there was another bug that fell out of the sigv4 work. should be fixed in https://review.openstack.org/#/c/330888/ -- there's a related patch for swift that *should* also fix it...i think... https://review.openstack.org/#/c/331369/ | 23:49 |
patchbot | timburke: patch 330888 - swift3 - Stop using email.utils.mktime_tz | 23:49 |
patchbot | timburke: patch 331369 - swift - Always set swift processes to use UTC | 23:49 |
kota_ | timezone is always fighting to me :.8 | 23:50 |
kota_ | :.( | 23:51 |
timburke | yeah, i know the feeling | 23:51 |
timburke | every time i start looking at how dates and times are handled in python i lose several hours and get a headache :/ | 23:51 |
kota_ | but it seems nice work, thanks for taking it anyway. | 23:52 |
* kota_ will take a look | 23:53 | |
kota_ | is that all for today? we are closing to 1 hour spent. | 23:54 |
timburke | yeah, i think that's it from me | 23:54 |
kota_ | ok | 23:54 |
timburke | on the version bump, do you want to propose that or should i? | 23:54 |
kota_ | i can take it | 23:55 |
* kota_ will do later | 23:55 | |
timburke | sounds good | 23:55 |
kota_ | and... | 23:55 |
kota_ | i thought to ask you and bill about barcelona summit but | 23:56 |
kota_ | he is absent today, so it's ok to take it later. | 23:56 |
timburke | i've claimed my ATC pass, and booked some flights. i suppose i haven't officially gotten permission to go, but if not, looks like i'll have a nice vacation :-) | 23:57 |
kota_ | so quickly! | 23:57 |
kota_ | I've just done only my ATC pass registration yet. | 23:58 |
timburke | going a week early with my wife, so we wanted to arrange flights sooner rather than later | 23:58 |
kota_ | i see. | 23:58 |
timburke | especially since it's a little complicated - we're going to see her parents on our way out and leave Teagan with them | 23:59 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!