| mattoliverau | lol | 00:00 |
|---|---|---|
| *** itlinux has joined #openstack-swift | 00:19 | |
| *** vint_bra1 has quit IRC | 00:30 | |
| *** vint_bra has joined #openstack-swift | 00:32 | |
| timburke | http://logs.openstack.org/27/531527/2/check/swift-tox-py27/fb13f40/job-output.txt.gz#_2018-01-08_22_59_52_213463 hmmm.... | 00:37 |
| timburke | Congratulations! .... Job timed out! | 00:37 |
| mattoliverau | i guess that means another recheck? | 00:43 |
| timburke | torgomatic's on it already | 00:44 |
| torgomatic | push the button, Frank | 00:44 |
| *** tovin07_ has joined #openstack-swift | 00:52 | |
| *** vint_bra has quit IRC | 01:37 | |
| kota_ | good morning | 01:38 |
| * kota_ is scrolling back the logs that are tons of lines... | 01:40 | |
| openstackgerrit | Matthew Oliver proposed openstack/swift master: Show devices marked as deleted on empty rings https://review.openstack.org/530839 | 01:42 |
| mattoliverau | Sigh.. damn pep8 | 01:42 |
| mattoliverau | timburke: clayg ^^ | 01:42 |
| mattoliverau | kota_: morning | 01:42 |
| kota_ | mattoliverau: hi | 01:42 |
| *** awkwardpenguin has quit IRC | 01:43 | |
| *** two_tired2 has joined #openstack-swift | 01:47 | |
| mattoliverau | clayg: as an aside, I know your a vim man like myself. I've been using https://github.com/frutiger/git-vimdiff quite a bit and find it really useful. gives you 'git vimdiff' which I use 'git vimdiff HEAD~' alot when review patches. It gives each file diff on a different tab, so I can jump back and forth and hack as required. I'm even lazy and.. https://github.com/matthewoliver/junk/blob/master/install_git_vimdiff.sh | 01:47 |
| mattoliverau | maybe you don't like tabs, or are happy with the normal git diff (using vimdiff) but I'm I like to jump around the patchset. | 01:48 |
| mattoliverau | s/i'm// | 01:48 |
| *** zhongjun has joined #openstack-swift | 01:52 | |
| openstackgerrit | Matthew Oliver proposed openstack/swift feature/deep: Use includes query param to request the SR that includes an object https://review.openstack.org/531661 | 01:57 |
| openstackgerrit | yanwei proposed openstack/swift master: Merge repeat code for rebalance There are three similar code segments in rebalance process as follows： https://review.openstack.org/531109 | 02:06 |
| openstackgerrit | Matthew Oliver proposed openstack/swift feature/deep: Use _root.db and _shard.db on sharded containers https://review.openstack.org/531350 | 02:14 |
| openstackgerrit | Samuel Merritt proposed openstack/swift master: proxy: make the right number of container updates https://review.openstack.org/531527 | 02:21 |
| *** armaan has quit IRC | 02:33 | |
| *** awkwardpenguin has joined #openstack-swift | 02:43 | |
| kota_ | ok, reached at the end of new logs. | 02:45 |
| kota_ | i loo | 02:45 |
| kota_ | oops | 02:45 |
| kota_ | it looks i should go to patch 531527 | 02:45 |
| patchbot | https://review.openstack.org/#/c/531527/ - swift - proxy: make the right number of container updates | 02:45 |
| kota_ | i don't have clear memory on the reason for the number of container updates but i like the smaller number if possible. | 02:47 |
| kota_ | IIRC, on those days, the number of commit ack was different too. maybe 2? so i couldn't choose less number of container_updates to prevent the container update lost. | 02:48 |
| *** awkwardpenguin has quit IRC | 02:49 | |
| *** bkopilov has quit IRC | 03:04 | |
| openstackgerrit | lei zhang proposed openstack/swift master: Fix the incorrect git.openstack.org source URL https://review.openstack.org/531989 | 03:05 |
| *** hoonetorg has quit IRC | 03:12 | |
| *** hoonetorg has joined #openstack-swift | 03:25 | |
| *** two_tired2 has quit IRC | 03:34 | |
| *** kei_yama has quit IRC | 03:37 | |
| *** gyee has quit IRC | 03:48 | |
| *** m_kazuhiro has joined #openstack-swift | 04:00 | |
| m_kazuhiro | good morning | 04:01 |
| *** kei_yama has joined #openstack-swift | 04:02 | |
| mattoliverau | m_kazuhiro: morning | 04:15 |
| m_kazuhiro | mattoliverau: morning! | 04:18 |
| kota_ | m_kazuhiro: o/ | 04:19 |
| m_kazuhiro | kota_: o/ | 04:19 |
| openstackgerrit | Merged openstack/swift master: Support existing builders with None _last_part_moves https://review.openstack.org/531905 | 04:22 |
| m_kazuhiro | mattoliverau: I added answer comments to your questions about legacy problem of general task queue on gerrit page. I want to get your opinion for them. https://review.openstack.org/#/c/517389/ | 04:22 |
| patchbot | patch 517389 - swift - WIP: Update object expirer to use general task que... | 04:22 |
| *** bkopilov has joined #openstack-swift | 04:23 | |
| mattoliverau | m_kazuhiro: oh yeah, I'll go take a look. | 04:23 |
| m_kazuhiro | mattoliverau: Thank you! And if you have time, please check internal client update patch for general task queue. clayg added +2 for this. https://review.openstack.org/#/c/530006/ | 04:25 |
| patchbot | patch 530006 - swift - Allow InternalClient to container/object listing w... | 04:25 |
| *** m_kazuhiro_ has joined #openstack-swift | 04:30 | |
| *** m_kazuhiro has quit IRC | 04:32 | |
| *** m_kazuhiro_ has quit IRC | 04:32 | |
| *** m_kazuhiro has joined #openstack-swift | 04:33 | |
| *** awkwardpenguin has joined #openstack-swift | 04:46 | |
| *** awkwardpenguin has quit IRC | 04:51 | |
| *** psachin has joined #openstack-swift | 05:04 | |
| *** links has joined #openstack-swift | 05:07 | |
| *** Renich_ has quit IRC | 05:22 | |
| *** hoonetorg has quit IRC | 05:55 | |
| *** hoonetorg has joined #openstack-swift | 05:59 | |
| *** links has quit IRC | 06:35 | |
| *** rcernin has quit IRC | 06:38 | |
| *** awkwardpenguin has joined #openstack-swift | 06:42 | |
| *** gkadam has quit IRC | 06:43 | |
| *** links has joined #openstack-swift | 06:51 | |
| *** hoonetorg has quit IRC | 06:56 | |
| *** pcaruana has joined #openstack-swift | 06:56 | |
| *** links has quit IRC | 07:03 | |
| *** armaan has joined #openstack-swift | 07:10 | |
| *** hoonetorg has joined #openstack-swift | 07:12 | |
| *** links has joined #openstack-swift | 07:14 | |
| *** awkwardpenguin has quit IRC | 07:18 | |
| *** links has quit IRC | 07:27 | |
| *** m_kazuhiro has quit IRC | 07:32 | |
| *** hoonetorg has quit IRC | 07:32 | |
| *** hseipp has joined #openstack-swift | 07:41 | |
| *** m_kazuhiro has joined #openstack-swift | 07:42 | |
| *** links has joined #openstack-swift | 07:45 | |
| *** notmyname has quit IRC | 07:49 | |
| *** PagliaccisCloud has quit IRC | 07:49 | |
| *** d0ugal has joined #openstack-swift | 07:52 | |
| *** notmyname has joined #openstack-swift | 07:52 | |
| *** ChanServ sets mode: +v notmyname | 07:52 | |
| *** PagliaccisCloud has joined #openstack-swift | 07:52 | |
| *** itlinux has quit IRC | 07:58 | |
| *** AJaeger has joined #openstack-swift | 07:58 | |
| AJaeger | swift3 cores, could you review https://review.openstack.org/#/c/530099/ to update your Zuul jobs, please? | 07:58 |
| patchbot | patch 530099 - swift3 - Import legacy tox jobs | 07:59 |
| *** neonpastor has quit IRC | 08:00 | |
| *** hoonetorg has joined #openstack-swift | 08:01 | |
| *** neonpastor has joined #openstack-swift | 08:02 | |
| *** gkadam has joined #openstack-swift | 08:13 | |
| *** tesseract has joined #openstack-swift | 08:20 | |
| *** psachin has quit IRC | 08:30 | |
| *** tovin07_ has quit IRC | 08:38 | |
| *** tovin07_ has joined #openstack-swift | 08:38 | |
| *** psachin has joined #openstack-swift | 08:40 | |
| *** geaaru has joined #openstack-swift | 08:48 | |
| *** cbartz has joined #openstack-swift | 08:55 | |
| acoles | good morning | 08:58 |
| acoles | mattoliverau: timburke clayg : nice work on ringbuilder fixes | 08:59 |
| *** awkwardpenguin has joined #openstack-swift | 09:20 | |
| *** flwang1 has quit IRC | 09:22 | |
| *** flwang has joined #openstack-swift | 09:24 | |
| *** nakagawamsa has joined #openstack-swift | 09:25 | |
| *** nakagawamsa has quit IRC | 09:25 | |
| *** awkwardpenguin has quit IRC | 09:25 | |
| *** hoonetorg has quit IRC | 09:32 | |
| openstackgerrit | Kota Tsuyuzaki proposed openstack/swift feature/s3api: Avoid global LOGGER instance https://review.openstack.org/529268 | 09:34 |
| *** jistr|off|mtg is now known as jistr | 09:36 | |
| kota_ | good morning acoles | 09:39 |
| acoles | kota_: o/ | 09:40 |
| kota_ | acoles: and a happy new year | 09:40 |
| acoles | thank you. and you too! | 09:40 |
| kota_ | acoles: sorry i was missing your message before the last vacation | 09:40 |
| kota_ | (and found after the vacation :( ) | 09:40 |
| acoles | oh well, better late than never :) | 09:40 |
| kota_ | AJaeger: done on patch 530099. thanks for the good work. | 09:42 |
| patchbot | https://review.openstack.org/#/c/530099/ - swift3 - Import legacy tox jobs | 09:42 |
| kota_ | timburke: i suppose you may prefer to patch 529268 rather than the old patch set. however since the work to write the patch takes over the year, i may put some mistakes in the patch so it's nice if you could take time to check the diff. | 09:45 |
| patchbot | https://review.openstack.org/#/c/529268/ - swift (feature/s3api) - Avoid global LOGGER instance | 09:45 |
| AJaeger | thanks, kota_ ! | 09:45 |
| *** armaan has quit IRC | 09:46 | |
| *** armaan has joined #openstack-swift | 09:47 | |
| *** hoonetorg has joined #openstack-swift | 09:49 | |
| *** armaan has quit IRC | 10:07 | |
| *** armaan has joined #openstack-swift | 10:07 | |
| *** d0ugal has quit IRC | 10:19 | |
| *** mvk has joined #openstack-swift | 10:20 | |
| *** d0ugal has joined #openstack-swift | 10:21 | |
| *** tovin07_ has quit IRC | 10:43 | |
| *** m_kazuhiro has quit IRC | 10:44 | |
| *** hoonetorg has quit IRC | 10:46 | |
| *** hoonetorg has joined #openstack-swift | 10:47 | |
| *** hseipp has quit IRC | 10:54 | |
| *** hseipp has joined #openstack-swift | 10:59 | |
| *** hoonetorg has quit IRC | 11:04 | |
| *** vint_bra has joined #openstack-swift | 11:17 | |
| *** kei_yama has quit IRC | 11:20 | |
| *** awkwardpenguin has joined #openstack-swift | 11:22 | |
| *** awkwardpenguin has quit IRC | 11:26 | |
| openstackgerrit | Alistair Coles proposed openstack/swift feature/deep: Turn down swiftclient and requests probe test logging level earlier https://review.openstack.org/532164 | 11:27 |
| *** vint_bra has quit IRC | 11:29 | |
| openstackgerrit | Alistair Coles proposed openstack/swift feature/deep: Turn down swiftclient and requests probe test logging level earlier https://review.openstack.org/532164 | 11:29 |
| *** vint_bra has joined #openstack-swift | 11:32 | |
| *** bkopilov has quit IRC | 11:36 | |
| *** mvk has quit IRC | 11:45 | |
| acoles | are the zuul checks smart enough to not run the dsvm jobs when the a patch only changes a unit test? e.g. p 531815 | 11:52 |
| patchbot | https://review.openstack.org/#/c/531815/ - swift - Fix intermittent check_delete_headers failure | 11:52 |
| acoles | timburke: ^^ you know *everything* ;) | 11:53 |
| openstackgerrit | Merged openstack/swift master: Allow InternalClient to container/object listing with prefix https://review.openstack.org/530006 | 11:59 |
| *** hoonetorg has joined #openstack-swift | 12:34 | |
| *** mvk has joined #openstack-swift | 12:47 | |
| *** zhurong_ has joined #openstack-swift | 12:56 | |
| *** hoonetorg has quit IRC | 13:03 | |
| *** zhurong_ has quit IRC | 13:07 | |
| *** bkopilov has joined #openstack-swift | 13:17 | |
| *** awkwardpenguin has joined #openstack-swift | 13:23 | |
| *** awkwardpenguin has quit IRC | 13:28 | |
| *** Dw_Sn has joined #openstack-swift | 13:34 | |
| Dw_Sn | Hello, I have an issue with one of my ring, it s -55.55 balance and now the rest of my luns are getting full and this one is free, can I run rebalance ? or check if there is a rebalance process running already ? | 13:35 |
| *** AJaeger has left #openstack-swift | 13:36 | |
| *** hseipp has quit IRC | 14:03 | |
| *** saint_ has joined #openstack-swift | 14:16 | |
| *** psachin has quit IRC | 14:28 | |
| *** awkwardpenguin has joined #openstack-swift | 14:30 | |
| *** awkwardpenguin has quit IRC | 14:35 | |
| openstackgerrit | Alistair Coles proposed openstack/swift feature/deep: Add state and state_timestamp to ShardRange https://review.openstack.org/531488 | 15:12 |
| openstackgerrit | Alistair Coles proposed openstack/swift feature/deep: Set shard range state during sharding https://review.openstack.org/531489 | 15:12 |
| *** saint_ has quit IRC | 15:14 | |
| *** hoonetorg has joined #openstack-swift | 15:16 | |
| openstackgerrit | Alistair Coles proposed openstack/swift master: Use _update_x_timestamp method in object controller DELETE method https://review.openstack.org/531447 | 15:23 |
| acoles | zaitcev: timburke: ^^ sorry to ask for your votes again but I liked timburke's suggestions | 15:26 |
| zaitcev | offset is something like 0.123_456, right? | 15:28 |
| acoles | zaitcev: correct | 15:28 |
| zaitcev | Is that even legal for clients to supply | 15:28 |
| zaitcev | why are they messing with our representation | 15:29 |
| zaitcev | is this for container-sync or something | 15:29 |
| *** ChanServ sets mode: +v jrichli | 15:30 | |
| zaitcev | I mean, it someone at proxy is trying to tell object server to delete at 0.123_456, it ought to be rejected with a 400. | 15:30 |
| acoles | in general clients can't supply x-timestamp, the gatekeeper prevents it. container sync is an exception. but internal client might get given an offset timestamp | 15:31 |
| zaitcev | Actually.... this is a test for proxy, not object server. | 15:31 |
| acoles | e.g. the reconciler uses offsets with internal client requests https://github.com/openstack/swift/blob/85d6cd30be965d3df2dca22d363aafd12146f869/swift/container/reconciler.py#L397-L431 | 15:31 |
| zaitcev | Makes sense, thanks | 15:32 |
| *** links has quit IRC | 15:36 | |
| *** hoonetorg has quit IRC | 15:41 | |
| *** hoonetorg has joined #openstack-swift | 15:54 | |
| *** ^andrea^ has quit IRC | 16:17 | |
| *** itlinux has joined #openstack-swift | 16:19 | |
| *** awkwardpenguin has joined #openstack-swift | 16:32 | |
| *** awkwardpenguin has quit IRC | 16:37 | |
| openstackgerrit | Merged openstack/swift master: Fix intermittent check_delete_headers failure https://review.openstack.org/531815 | 16:38 |
| *** geaaru has quit IRC | 16:48 | |
| *** hseipp has joined #openstack-swift | 16:50 | |
| *** itlinux has quit IRC | 16:56 | |
| *** gyee has joined #openstack-swift | 17:01 | |
| *** mvk has quit IRC | 17:06 | |
| *** awkwardpenguin has joined #openstack-swift | 17:18 | |
| *** itlinux has joined #openstack-swift | 17:18 | |
| notmyname | good morning | 17:19 |
| *** cbartz has quit IRC | 17:22 | |
| *** hseipp has quit IRC | 17:27 | |
| *** Dw_Sn has quit IRC | 17:29 | |
| *** JimCheung has joined #openstack-swift | 17:39 | |
| timburke | acoles: on the dsvm jobs -- yep :-) i did that in https://github.com/openstack-infra/project-config/commit/633b4c693 | 17:45 |
| acoles | timburke: nice | 17:47 |
| timburke | SO much faster | 17:47 |
| acoles | yup | 17:47 |
| *** awkwardpenguin has quit IRC | 17:48 | |
| *** gkadam has quit IRC | 17:49 | |
| *** tesseract has quit IRC | 17:57 | |
| *** hseipp has joined #openstack-swift | 18:19 | |
| *** hseipp has quit IRC | 18:29 | |
| *** itlinux has quit IRC | 18:38 | |
| *** itlinux has joined #openstack-swift | 18:42 | |
| *** itlinux has quit IRC | 18:46 | |
| *** itlinux has joined #openstack-swift | 18:49 | |
| *** itlinux has quit IRC | 18:52 | |
| *** itlinux has joined #openstack-swift | 18:54 | |
| *** itlinux has quit IRC | 18:55 | |
| *** gkadam has joined #openstack-swift | 18:57 | |
| *** JimCheung has quit IRC | 18:59 | |
| *** JimCheung has joined #openstack-swift | 19:01 | |
| notmyname | reminder, we should cut a release soon, so https://wiki.openstack.org/wiki/Swift/PriorityReviews is being updated with appropriate patches | 19:02 |
| notmyname | note that the queens cycle is ending soon (next month), so this may be a final release in queens, or we may have a smaller fast-follow release closer to the queens freeze date (feb 19) | 19:02 |
| clayg | interesting... | 19:18 |
| openstackgerrit | Clay Gerrard proposed openstack/swift master: Further extract a builder method to a util function https://review.openstack.org/532291 | 19:18 |
| *** itlinux has joined #openstack-swift | 19:18 | |
| notmyname | clayg: I find your "interesting..." comment interesting | 19:22 |
| clayg | i just hadn't realized we were so close to OS release - your point about the timelines is realistic I think - and it's a reminder that time moves fast | 19:22 |
| notmyname | ah. I thought your refactor was the interesting one :-) | 19:23 |
| notmyname | and I was trying to work out why | 19:23 |
| notmyname | yeah, the queens release kinda snuck up on me too | 19:23 |
| clayg | no, that's just trying to help up some newish contributor that's been poking around at stuff | 19:24 |
| notmyname | interesting... | 19:24 |
| notmyname | ;-) | 19:24 |
| *** itlinux has quit IRC | 19:24 | |
| clayg | following up on something they already have loaded in their brain seems like a good way to introduce them to different ideas about clarity and maintainability and collaboration... | 19:24 |
| clayg | it was orthogonal to my reaction to the release - and significantly LESS interesting in my mind ;) | 19:25 |
| *** hoonetorg has quit IRC | 19:25 | |
| *** gkadam has quit IRC | 19:32 | |
| *** itlinux has joined #openstack-swift | 19:36 | |
| *** itlinux has quit IRC | 19:41 | |
| clayg | timburke: looks like you pointed torgomatic at some issues with the pipeline insertion on patch 528106 - but didn't -1 | 19:57 |
| patchbot | https://review.openstack.org/#/c/528106/ - swift - Move CORS to middleware. | 19:57 |
| timburke | shit, i still have three pending drafts! i better finish that review! | 19:58 |
| timburke | i got side-tracked by wanting to write a CORS-functest harness, but i'm having a hard time imagining it being worth the (rather large) amount of work that requires | 19:59 |
| clayg | timburke: did you look at the existing functional tests for CORS? I feel like the existing functests are mostly focused on objects... but the unittests seem to do some assertions on container requests and strict mode about a few things... | 20:00 |
| clayg | ah... | 20:01 |
| clayg | is that mainly because or CORS support is so flakey/underused? | 20:01 |
| timburke | clayg: certainly seems to be -- but then, you saw my rant last week around http://eavesdrop.openstack.org/irclogs/%23openstack-swift/%23openstack-swift.2018-01-05.log.html#t2018-01-05T21:58:03 :P | 20:03 |
| clayg | yeah :D | 20:04 |
| timburke | you can't even auth against tempauth... now i'm wondering if SLO/DLO work... it's a mess | 20:04 |
| timburke | any time a middleware swallows a proxy-server response (or never even *calls* the proxy!) and instead makes up something from whole cloth, we've almost certainly broken CORS | 20:06 |
| timburke | given that we *haven't* had people screaming at us about it... it's probably underused to the point that we'll just continue to neglect it? torgomatic's patch is a good step toward *not*-neglect, but i'm increasingly feeling like the best way to make sure CORS works is a func test harness that simulates how a browser handles CORS | 20:11 |
| *** mvk has joined #openstack-swift | 20:11 | |
| timburke | and it turns out that's *also* really messy, and a lot of work, and you'd still be left wondering "did we *accurately* simulate a browser's behavior?" | 20:12 |
| timburke | maybe something like https://github.com/synacor/python-cors would be helpful? | 20:15 |
| *** hoonetorg has joined #openstack-swift | 20:21 | |
| *** armaan has quit IRC | 20:22 | |
| *** armaan has joined #openstack-swift | 20:23 | |
| *** armaan has quit IRC | 20:44 | |
| openstackgerrit | Alistair Coles proposed openstack/swift feature/deep: Set shard range state during sharding https://review.openstack.org/531489 | 20:44 |
| openstackgerrit | Alistair Coles proposed openstack/swift feature/deep: WIP NEEDS TEST Fix container listing while in sharding state https://review.openstack.org/532308 | 20:45 |
| notmyname | timburke: https://docs.openstack.org/swift/latest/cors.html#test-cors-page | 20:58 |
| notmyname | that's the one I was thinking of. not sure how helpful it is for what you're doing | 20:59 |
| timburke | notmyname: thanks -- reasonably interesting, but not so far off from what i'm already playing with. still manual :-( i might see what i can do to build out some sort of "automated" in-browser test suite | 21:06 |
| notmyname | it has the advantage of already being in our source repo :-) | 21:06 |
| *** pcaruana has quit IRC | 21:38 | |
| *** threestrands has joined #openstack-swift | 21:44 | |
| *** threestrands has quit IRC | 21:44 | |
| *** threestrands has joined #openstack-swift | 21:44 | |
| torgomatic | we should just give up on CORS entirely | 22:02 |
| torgomatic | the point of CORS is so the browser can provide or not provide data to the running JS application, but it's data that the browser has access to already | 22:02 |
| torgomatic | now, thanks to the miracle of speculative execution and cache-timing attacks, the JS application has access to the data anyway, so we don't need CORS at all! :p | 22:03 |
| *** rcernin has joined #openstack-swift | 22:06 | |
| openstackgerrit | Merged openstack/swift feature/deep: Turn down swiftclient and requests probe test logging level earlier https://review.openstack.org/532164 | 22:14 |
| notmyname | torgomatic: lol | 22:14 |
| openstackgerrit | Samuel Merritt proposed openstack/swift master: Limit object-expirer queue updates on object DELETE https://review.openstack.org/532342 | 22:23 |
| clayg | torgomatic: ^ this is pretty shitty | 22:26 |
| clayg | torgomatic: i missed that when you were talking about it earlier | 22:26 |
| torgomatic | clayg: the code, the situation, or both? :p | 22:26 |
| clayg | lol, well I don't immediately *love* X-Backend-Update-Expiring-Object-Queue: <true|false> | 22:27 |
| clayg | why not just don't send the expirer update headers? Or maybe better question - why isn't pop_queue *eventually* going to be everything we need? | 22:27 |
| torgomatic | clayg: it's already not sending expirer update headers | 22:28 |
| torgomatic | this is on object DELETE, so the object server is the only guy who knows if the object is expiring or not | 22:28 |
| clayg | I'm pretty sure the reconciler relies on it's own idempotency to eventually clean up a missed queue update | 22:28 |
| torgomatic | and he's got no rings, so he just drops an async_pending and calls it a day | 22:28 |
| clayg | oh sorry - i'm starting to catch on | 22:29 |
| mattoliverau | morning | 22:29 |
| torgomatic | yeah, the expirer will clean his own queue if you let him go; the early delete is an optimization so if you prematurely delete a bunch of expiring objects, the queue shrinks back down | 22:29 |
| torgomatic | morning! | 22:29 |
| clayg | meh... only if you haven't since changed your container divisor - or it wasn't in a container before we started doing the divisor_container - hash % 100 thing | 22:30 |
| timburke | and the upgrade path to making proxy-server send down expirer update headers and object-server rely solely on that involves losing expirer-queue updates | 22:30 |
| clayg | timburke: you mean have the proxy pre-calculate expirer update headers just in case | 22:31 |
| torgomatic | I'm not opposed to updating the expirer queue on object DELETE, but maybe let's not do it so much is all I'm saying | 22:31 |
| timburke | ya, like we already do for container updates | 22:32 |
| torgomatic | if you miss an update, things still work correctly, so let's not destroy the container servers in order to optimize them | 22:32 |
| torgomatic | you can't compute expirer update headers on DELETE; you don't know the object's X-Delete-At, so you can't figure out which shard to look up in the ring | 22:33 |
| clayg | this applies equally to PUT overwrite? | 22:33 |
| *** awkwardpenguin has joined #openstack-swift | 22:33 | |
| torgomatic | could be; I haven't looked | 22:33 |
| timburke | ah, good point. i was thinking about the PUT/POST path | 22:34 |
| torgomatic | clayg: ...yep, applies to PUT overwrite too, so I should probably get that code path | 22:34 |
| timburke | yup. only it get *even more fun* since we need to delete the old entry then add the new one | 22:34 |
| torgomatic | at least we get update headers for the new one, so it's not always going to async_pending | 22:35 |
| clayg | does the *expirer* itself somehow supress the generation of the async pendings tho? this is *any* DELETE or just DELETEs from the client? | 22:35 |
| torgomatic | clayg: currently, no, the expirer doesn't suppress anything, so the expirer issues a bunch of DELETE requests (creating async_pendings), then directly deletes the container rows, then the object updater eats the async_pendings and directly deletes the container rows a lot | 22:36 |
| torgomatic | I'm working on fixing that thing next now that I've got a way of suppressing the generation of those async_pendings | 22:36 |
| torgomatic | although I should probably fix up PUT-overwrite first | 22:42 |
| mattoliverau | while thinking about the expirer, which is good. We should also think about how far away is the general task queue which will also change how the expirer will operate. And can we glean something from what your working on now that we can put into the general queue design and expirer reimplementation? | 22:44 |
| torgomatic | mattoliverau: maybe? The general queue thing is about what the queue looks like; the stuff I'm doing is about how many enqueue/dequeue operations occur | 22:44 |
| *** itlinux has joined #openstack-swift | 22:47 | |
| mattoliverau | sure. I understand :) just want to make sure whatever brillant enqueue/dequeue optimisations you come up with making it known so anything using the new queue can take advantage (or copy) ;) Just saying keep on going but don't forget Kazuhiro is playing with reworking the expirer as well, and not sure how much of the changes he's making would conflict with what your playing with (patch 517389) | 22:54 |
| mattoliverau | maybe none, but want just wanted to put it in the back of everyones mind :) | 22:56 |
| torgomatic | :) | 22:57 |
| mattoliverau | Also, I've woken up to an email saying work'll send me to the PTG! now need to go deal with find the best 30+ hour flights. | 23:02 |
| *** awkwardpenguin has quit IRC | 23:04 | |
| *** awkwardpenguin has joined #openstack-swift | 23:04 | |
| torgomatic | hooray! I mean, not for the 30-hour flight part, but for the rest | 23:07 |
| *** awkwardpenguin has quit IRC | 23:10 | |
| timburke | mattoliverau: yay! | 23:26 |
| timburke | speaking of... i should go book tickets/travel/hotel... | 23:26 |
| torgomatic | timburke: you've got to call the hotel on the phone. They have a website, but it won't take the room-block code, and if you email them then they email you back asking for a credit-card number (in cleartext email!), so... | 23:28 |
| notmyname | mattoliverau: yay | 23:31 |
| *** kei_yama has joined #openstack-swift | 23:37 | |
| *** vint_bra has joined #openstack-swift | 23:49 | |
| openstackgerrit | Tim Burke proposed openstack/swift master: Add pipeline modification test for previously-recommended pipelines https://review.openstack.org/532360 | 23:56 |
| timburke | torgomatic: maybe handy? ^^^ | 23:56 |
| timburke | i kinda wanna just get it out of the git history dynamically... but the expectations are going to have to be specified out-of-band... | 23:58 |
| *** vint_bra has quit IRC | 23:58 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!