| *** cemason has quit IRC | 03:04 | |
| *** cemason has joined #swift3 | 03:04 | |
| kota_ | ugh, reviewing sig v4 makes me nurvas, it looks too different from Swift3/S3 implementaion almostly everything. I have to solve the tangled code step by step... | 06:48 |
|---|---|---|
| *** acoles_ is now known as acoles | 08:33 | |
| *** Shashikant86 has joined #swift3 | 08:41 | |
| *** Shashikant86 has quit IRC | 08:42 | |
| *** Shashikant86 has joined #swift3 | 08:43 | |
| *** Shashikant86 has quit IRC | 09:41 | |
| *** Shashikant86 has joined #swift3 | 09:45 | |
| *** Shashikant86 has quit IRC | 09:48 | |
| *** Shashikant86 has joined #swift3 | 09:56 | |
| openstackgerrit | Kota Tsuyuzaki proposed openstack/swift3: Re:implement AWS signature v4 https://review.openstack.org/301165 | 10:03 |
| openstackgerrit | Kota Tsuyuzaki proposed openstack/swift3: Re:implement AWS signature v4 https://review.openstack.org/301165 | 10:06 |
| kota_ | hopefully it is getting to be land-able... I was tired a bit to look at... | 10:09 |
| openstackgerrit | Kota Tsuyuzaki proposed openstack/swift3: Re:implement AWS signature v4 https://review.openstack.org/301165 | 10:13 |
| *** Shashikant86 has quit IRC | 10:50 | |
| *** Shashikant86 has joined #swift3 | 10:54 | |
| *** openstackgerrit has quit IRC | 11:18 | |
| *** openstackgerrit has joined #swift3 | 11:18 | |
| openstackgerrit | Kota Tsuyuzaki proposed openstack/swift3: Re:implement AWS signature v4 https://review.openstack.org/301165 | 11:19 |
| kota_ | weird, that unit failure looks to come from self.path_info -> self.path change | 11:20 |
| kota_ | probably I am missing something | 11:20 |
| openstackgerrit | Kota Tsuyuzaki proposed openstack/swift3: Re:implement AWS signature v4 https://review.openstack.org/301165 | 11:30 |
| *** Shashikant86 has quit IRC | 11:39 | |
| kota_ | ah....what? | 11:43 |
| kota_ | looking at the log http://logs.openstack.org/65/301165/20/check/gate-swift3-tox-keystone/9faafc6/console.html, canonical request is as host:localhost:8080:8080 ?? | 11:45 |
| kota_ | ah, that's what Andrey said in the original patch, right? | 11:45 |
| *** Shashikant86 has joined #swift3 | 11:52 | |
| *** Shashikant86 has quit IRC | 11:54 | |
| openstackgerrit | Kota Tsuyuzaki proposed openstack/swift3: Re:implement AWS signature v4 https://review.openstack.org/301165 | 11:58 |
| kota_ | anyway, the canonical string in string_to_sign matches the actuall public example from S3 so the calculation except host:port seems correct, right now... | 12:10 |
| kota_ | ah, only one error remaining | 12:23 |
| kota_ | yeah, that's probably what i touched in recent work. | 12:24 |
| kota_ | gotcha, it comes from buggy boto code it makes the host name as localhost:8080 and append port info 8080, it makes swob.Request.host_url to localhost:8080:8080 that violates URL syntax. | 13:06 |
| openstackgerrit | Kota Tsuyuzaki proposed openstack/swift3: Re:implement AWS signature v4 https://review.openstack.org/301165 | 13:20 |
| *** cemason1 has joined #swift3 | 13:21 | |
| *** cemason has quit IRC | 13:21 | |
| kota_ | ugh, still failed? | 13:35 |
| kota_ | aaaaaah, the issue seems correct but the newest patch resolve nothing. | 13:45 |
| openstackgerrit | Kota Tsuyuzaki proposed openstack/swift3: Re:implement AWS signature v4 https://review.openstack.org/301165 | 14:19 |
| kota_ | hopefully it can pass whole tests... | 14:26 |
| kota_ | Pylint doesn't like six, i understood. | 14:50 |
| timburke | kota_: on the doubled-up ports: yeah, we saw that with the original patch. will be fixed if boto merges https://github.com/boto/boto/pull/3513 (or https://github.com/boto/boto/pull/3181 although i don't like that approach as much) | 16:48 |
| *** lyrrad has joined #swift3 | 16:54 | |
| timburke | if the unit failure is the one i'm thinking of, i saw that before, too. get-utf8 subtest in test_canonical_string_v4? the path info should be unquoted in the env, so '/\xE1\x88\xB4' or u'/\u1234'.encode('utf-8') would more accurately reflect a real WSGI env (although i'm guessing you already figured that out) | 16:55 |
| *** cemason has joined #swift3 | 20:34 | |
| *** cemason1 has quit IRC | 20:34 | |
| openstackgerrit | Kota Tsuyuzaki proposed openstack/swift3: Re:implement AWS signature v4 https://review.openstack.org/301165 | 21:32 |
| *** acoles is now known as acoles_ | 22:01 | |
| *** bill_az_ has joined #swift3 | 22:02 | |
| kota_ | hello | 23:00 |
| bill_az_ | kota_: hi there | 23:00 |
| kota_ | hi bill_az_ | 23:01 |
| kota_ | i'm expecting timburke also will attend... | 23:02 |
| timburke | hi! thanks for the ping :) | 23:02 |
| kota_ | :-) | 23:03 |
| bill_az_ | Hi | 23:03 |
| kota_ | thanks for joining :) | 23:03 |
| kota_ | let's get started | 23:03 |
| kota_ | agenda is here, https://wiki.openstack.org/wiki/Meetings/swift3 | 23:03 |
| kota_ | looks like just one item for today. | 23:03 |
| kota_ | #topic sigv4 | 23:04 |
| kota_ | (does the command work?) | 23:04 |
| kota_ | yeah, that seems available only at #openstack-meeting channel | 23:05 |
| timburke | i think not; meetbot (or whatever) does that stuff | 23:05 |
| kota_ | ok | 23:05 |
| kota_ | recently I'm working on the signature v4 stuff again to address timburke's comment. | 23:06 |
| kota_ | comments, they reached 50 comments. | 23:06 |
| timburke | and i've been meaning to circle back and review all the latest changes. looking good so far! | 23:07 |
| timburke | lot's more to cover :( | 23:07 |
| kota_ | i hope so, some parts of them were hard because about some of them I didn't know the original Andrey's intention | 23:08 |
| kota_ | i think the latest one I pushed this morning can probably pass for all gates at gerrit. | 23:09 |
| kota_ | ( | 23:09 |
| kota_ | it seems still being running. | 23:09 |
| kota_ | or waiting at queue. | 23:09 |
| timburke | looking at http://status.openstack.org/zuul/ everything seemed to pass. we should probably see a notification soon | 23:10 |
| kota_ | with this work, I'm realizing to concern it is getting to big for one patch :/ | 23:11 |
| kota_ | I think, we still have minor issues (I added somewhere as NOTE) but if they're not serious, I'd like to merge it at first. | 23:11 |
| timburke | i was thinking about that too a bit. though now that i've gotten most of it loaded into my head, i don't really want to split it up | 23:12 |
| kota_ | timburke: if you find something serious, it's ok to make -1 to prevent to merge | 23:12 |
| kota_ | got a jenkins result! | 23:12 |
| kota_ | all green yey | 23:12 |
| bill_az_ | it is really only one file with big changes - not sure how you would split that up as changes are all related | 23:13 |
| timburke | agreed. if there's nothing that leaps out at me, the plan is to leave some comments and start addressing them in follow-up patch(es) | 23:14 |
| kota_ | bill_az_: yeah, request.py is too big and also the unit test is getting bigger rather than i expected | 23:15 |
| kota_ | timburke: ;-) | 23:15 |
| bill_az_ | kota_: when we're finished with this topic, i did have one question - I chatted with timburke last week but wanted to get your thoughts | 23:16 |
| kota_ | bill_az_: yup, you have floor | 23:17 |
| kota_ | scrolling back the log... | 23:17 |
| kota_ | about s3token? | 23:18 |
| bill_az_ | we are using swift3 w/ keystone auth, and we see big performance drop | 23:18 |
| bill_az_ | yes | 23:18 |
| kota_ | bill_az_: gotcha, that is probably problematic | 23:18 |
| bill_az_ | using tempauth performance is comparable, but with s3token much worse - because each request is going back to keystone | 23:18 |
| timburke | and worse, i think it's doing so twice :/ | 23:19 |
| bill_az_ | signature is only valid for short time - 15 min? - but I thought I had seen an earlier patch or blueprint that was about adding caching for signatures | 23:19 |
| kota_ | yes, it makes 2 requests which is un cachable to keystone. | 23:19 |
| bill_az_ | any ideas / suggestions on how we can improve this? | 23:20 |
| kota_ | ah, signature caching? | 23:21 |
| kota_ | can we go the way? I think the signature includes verb, path, any headers, and dates. | 23:21 |
| bill_az_ | yeah - if its for the same object to the same proxy it would help - but that's probably not much help | 23:22 |
| kota_ | even if we cache the signature, probably another request includes another info about them, i.e. signature will mis-match for the cache. | 23:22 |
| kota_ | make sense. | 23:23 |
| timburke | yeah, i'm fairly certain that any attempt to cache keystone responses in s3token will result in many many cache misses | 23:24 |
| bill_az_ | what are the two requests to keystone? looking at the code now - | 23:24 |
| timburke | one in s3token, one in auth_token | 23:25 |
| kota_ | the first is in s3token and the other is probably auth_token middleware | 23:25 |
| timburke | i want to make that less dumb, and now that s3token is in the swift3 repo, i think there's a better chance of it | 23:25 |
| kota_ | timburke: agreed | 23:26 |
| timburke | but i still haven't fully grokked how those interact or what needs to be populated where | 23:26 |
| timburke | (and probably ahead of that patch will be the use-swift's-http-client patch, which should help orient me somewhat) | 23:27 |
| kota_ | bill_az_: in current swift3 process, s3token attempts to get *token* to use *Swift* via s3 signature, and then, auth_token attempts to verify the *token* to keystone. | 23:27 |
| kota_ | even though the token is served just now in previous pipeline processs :/ | 23:28 |
| bill_az_ | ic - so s3token is basically converting signature to token and passing it on | 23:28 |
| kota_ | timburke: yup | 23:28 |
| kota_ | bill_az_: can i have the link for the blueprint for this? | 23:30 |
| kota_ | or ealier patch? | 23:31 |
| kota_ | anyways, we should file this issue somewhere online. | 23:31 |
| bill_az_ | if s3token is getting a token, then shouldnt authtoken be able to take advantage of caching and reduce keystone calls to one? | 23:31 |
| bill_az_ | kota_: I was not able to find that - maybe I dreamed it up... :-( | 23:32 |
| kota_ | bill_az_: if we could get all permissions like roles at the s3token request, we don't need auth_token process, i think. | 23:32 |
| kota_ | bill_az_: ok, np. we can make a new report by ourselvs :) | 23:33 |
| bill_az_ | If I can dig it up, I'll pass it on | 23:33 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!