*** mikecmpbll has quit IRC | 00:02 | |
*** evrardjp has quit IRC | 00:52 | |
openstackgerrit | Tim Burke proposed openstack/swift master: prefactor: Pass s3api Config to S3Requests https://review.opendev.org/729091 | 02:36 |
---|---|---|
openstackgerrit | Tim Burke proposed openstack/swift master: s3api: Log ratelimited requests as 498s https://review.opendev.org/729092 | 02:36 |
openstackgerrit | Tim Burke proposed openstack/swift master: s3api: Add config option to return 429s on ratelimit https://review.opendev.org/729093 | 02:36 |
*** gyee has quit IRC | 03:23 | |
*** psachin has joined #openstack-swift | 03:39 | |
*** threestrands has quit IRC | 04:15 | |
*** ccamacho has joined #openstack-swift | 06:32 | |
*** rpittau|afk is now known as rpittau | 06:51 | |
*** abelur has quit IRC | 07:26 | |
*** abelur has joined #openstack-swift | 07:28 | |
*** evrardjp has joined #openstack-swift | 07:31 | |
*** evrardjp has quit IRC | 07:32 | |
*** evrardjp has joined #openstack-swift | 07:34 | |
*** ccamacho has quit IRC | 07:40 | |
*** mikecmpbll has joined #openstack-swift | 07:58 | |
*** tkajinam has quit IRC | 08:02 | |
*** tkajinam has joined #openstack-swift | 10:30 | |
*** rpittau is now known as rpittau|bbl | 10:31 | |
*** rpittau|bbl is now known as rpittau | 12:04 | |
*** ccamacho has joined #openstack-swift | 12:18 | |
*** psachin has quit IRC | 13:35 | |
*** tkajinam has quit IRC | 14:11 | |
*** tkajinam has joined #openstack-swift | 14:11 | |
*** mikecmpbll has quit IRC | 15:00 | |
*** mikecmpbll has joined #openstack-swift | 15:03 | |
*** gyee has joined #openstack-swift | 15:46 | |
*** rpittau is now known as rpittau|afk | 16:22 | |
openstackgerrit | Tim Burke proposed openstack/swift master: s3api: Log ratelimited requests as 498s https://review.opendev.org/729092 | 16:30 |
clayg | Tim, my experience is that historically we've had enough trouble with our little environ hook passing that it's worth it do whatever is necessary to observe the expected behavior on p 729092 before I +A, if you don't have any specific recommendations for functionally testing I'm happy to pursue on my own. | 17:22 |
patchbot | https://review.opendev.org/#/c/729092/ - swift - s3api: Log ratelimited requests as 498s - 2 patch sets | 17:22 |
timburke | clayg, enable ratelimit and put it right of auth; something like `container_listing_ratelimit_0 = 10` and `max_sleep_time_seconds = 0.1` seemed to work well for me, though you could turn the max sleep all the way down to 0 if you want to *everything* 498s | 17:25 |
timburke | make s3 listings fast enough to trip ratelimiting, and look at logs to see that it's not logged as the same 503 that comes back to the client | 17:27 |
clayg | yuk, vsaio doesn't setup the config section for ratelimit by default it doesn't seem | 17:27 |
timburke | nope :-( | 17:27 |
clayg | for i in {1..20}; do aws s3api list-object --bucket test; done going to be good enough? | 17:28 |
timburke | it's simple enough, though: `use = egg:swift#ratelimit` | 17:28 |
clayg | yeah but how do you avoid the obvious yak of adding support to vsaio!!! 😁 | 17:28 |
timburke | start-up time may dominate; fwiw i did something like `for i in {1..20}; do aws s3 ls s3://test/ & done; time wait` so it's in parallel | 17:30 |
clayg | `tempurl s3api ratelimit tempauth ` | 17:31 |
clayg | ^? | 17:32 |
*** evrardjp has quit IRC | 17:32 | |
clayg | or, no i guess you're saying `tempurl s3api tempauth ratelimit` 🤔 | 17:33 |
timburke | yeha, i'd put it after auth -- that's the bit that'll be responsible for translating from /v1/test:tester/test to /v1/AUTH_test/test so you get a single limit for both swift and s3 | 17:33 |
*** evrardjp has joined #openstack-swift | 17:34 | |
clayg | timburke: what about p 729051 you started talking about cake and now I'm super excited about having ratelimits all over my pipeline!!! | 17:34 |
patchbot | https://review.opendev.org/#/c/729051/ - swift - ratelimit: Allow multiple placements - 2 patch sets | 17:34 |
timburke | all over the place! yeah, with that one, you can do something like `ratelimit s3api tempauth ratelimit` and not double-up book-keeping for swift requests | 17:35 |
clayg | yeah that sounds pretty freaking sexy | 17:37 |
timburke | do i remember this right? wasn't there some kind of issue with putting s3api in our default pipeline? trouble with devstack trying to have both s3api *and* swift3, something like that... | 18:00 |
*** mikecmpbll has quit IRC | 18:07 | |
*** mikecmpbll has joined #openstack-swift | 18:08 | |
openstackgerrit | Tim Burke proposed openstack/swift master: Add s3api to recommended pipeline https://review.opendev.org/729344 | 18:12 |
timburke | well, let's see what ^^^ does | 18:12 |
*** rcernin has quit IRC | 18:34 | |
clayg | oh wow, so swiftclient doesn't like to be ratelimited `Container GET failed: http://saio:8080/v1/AUTH_test/test?format=json 498 Rate Limited Slow down` 😞 | 18:36 |
timburke | whomp whomp | 18:37 |
timburke | for some reason i thought i'd had it respond reasonably well... *shrug* | 18:38 |
timburke | hey, wait -- https://github.com/openstack/python-swiftclient/commit/5187fd313 | 18:38 |
timburke | why don't we default that to true!? | 18:39 |
*** mikecmpbll has quit IRC | 19:23 | |
*** mikecmpbll has joined #openstack-swift | 19:23 | |
*** mikecmpbll has quit IRC | 19:46 | |
*** ccamacho has quit IRC | 21:09 | |
*** renich has joined #openstack-swift | 21:10 | |
clayg | timburke: FWIW I think I'm done reviewing p 729092 - but I hesitated on the +A since it hasn't been up very long and I'm not such a s3api expert. Ideally @rledisez or @kota_ could sweep through and knock it out if the trust my review and our comments and commit message make sense? | 21:22 |
patchbot | https://review.opendev.org/#/c/729092/ - swift - s3api: Log ratelimited requests as 498s - 2 patch sets | 21:22 |
*** rcernin has joined #openstack-swift | 22:56 | |
openstackgerrit | Tim Burke proposed openstack/swift master: ratelimit: Allow multiple placements https://review.opendev.org/729051 | 23:01 |
mattoliverau | morning | 23:20 |
*** renich has quit IRC | 23:41 |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!