| *** mingyu has quit IRC | 00:11 | |
| *** NM has joined #openstack-swift | 00:13 | |
| *** Sukhdev_ has quit IRC | 00:16 | |
| *** alenavolk has quit IRC | 00:48 | |
| *** klrmn has quit IRC | 01:12 | |
| *** _JZ__ has joined #openstack-swift | 01:14 | |
| *** jeffli has joined #openstack-swift | 01:16 | |
| *** _JZ_ has quit IRC | 01:18 | |
| *** links has joined #openstack-swift | 01:30 | |
| clayg | i'm going to look at patch 448240 | 01:36 |
|---|---|---|
| patchbot | https://review.openstack.org/#/c/448240/ - swift - Enable per policy proxy config options | 01:36 |
| *** zhengyin has joined #openstack-swift | 01:41 | |
| zaitcev | http://paste.openstack.org/raw/610610/ | 01:42 |
| zaitcev | you've got to be kidding | 01:42 |
| *** chsc has joined #openstack-swift | 01:43 | |
| *** tovin07_ has joined #openstack-swift | 01:51 | |
| *** JimCheung has quit IRC | 01:52 | |
| clayg | kota_: do we have a patch to fixup http://saio:8080/v1/AUTH_test/doc/overview_erasure_code.html#global-ec | 01:52 |
| *** JimCheung has joined #openstack-swift | 01:52 | |
| clayg | oh sorry :\ | 01:52 |
| clayg | https://docs.openstack.org/developer/swift/overview_erasure_code.html#global-ec how bout (cc kota_ ) | 01:52 |
| clayg | redbo: can you help out zaitcev ? | 01:54 |
| *** JimCheung has quit IRC | 01:56 | |
| redbo | it probably needs go 1.8 | 01:58 |
| zaitcev | golang-1.7.5-2.fc25.x86_64 | 01:58 |
| zaitcev | But... why? This used to work on 1.7 and the code didn't change since you wrote it in 2014. | 01:59 |
| zaitcev | It's this construct | 01:59 |
| zaitcev | type HummingbirdServer struct { | 01:59 |
| zaitcev | http.Server | 01:59 |
| *** vint_bra has joined #openstack-swift | 01:59 | |
| redbo | http.Server grew a Shutdown method for gracefully shutting down in 1.8 | 02:00 |
| zaitcev | oh | 02:00 |
| *** NM has quit IRC | 02:02 | |
| kota_ | back to online | 02:05 |
| *** zhurong has joined #openstack-swift | 02:05 | |
| kota_ | clayg: not yet but good call | 02:06 |
| *** notmyname has quit IRC | 02:08 | |
| *** klrmn has joined #openstack-swift | 02:09 | |
| clayg | nice work redbo! | 02:12 |
| jeffli | quit | 02:12 |
| *** jeffli has quit IRC | 02:12 | |
| *** notmyname has joined #openstack-swift | 02:13 | |
| *** ChanServ sets mode: +v notmyname | 02:13 | |
| *** alenavolk has joined #openstack-swift | 02:32 | |
| *** lucasxu has joined #openstack-swift | 02:33 | |
| *** itlinux has joined #openstack-swift | 02:37 | |
| *** itlinux has quit IRC | 03:03 | |
| *** lucasxu has quit IRC | 03:06 | |
| *** lucasxu has joined #openstack-swift | 03:08 | |
| *** _JZ__ has quit IRC | 03:09 | |
| zaitcev | Error: Expected nil, but got: &errors.errorString{s:"Error setting up logger: Unable to create logger: open /var/log/swift/accountreplicator.log: no such file or directory"} | 03:12 |
| zaitcev | (no extra help needed, I'm just amazed that apparently everyone runs "go test -cover ./..." as root nowadays) | 03:12 |
| *** alenavolk has quit IRC | 03:15 | |
| *** lucasxu has quit IRC | 03:18 | |
| clayg | zaitcev: in a container - you're always root | 03:18 |
| *** vint_bra has quit IRC | 03:21 | |
| *** Sukhdev has joined #openstack-swift | 03:32 | |
| clayg | omg I should run my saio with write_affinity *all the time* i *love* handoffs! | 03:35 |
| clayg | and replication 😋 | 03:35 |
| clayg | OMG I have *three* different timestamps for my EC object right now - and *all* of my servers are running - why did you guys not tell me how much fun this wAS! | 03:39 |
| *** kei_yama has quit IRC | 04:00 | |
| *** kei_yama has joined #openstack-swift | 04:01 | |
| *** psachin has joined #openstack-swift | 04:05 | |
| *** klrmn has quit IRC | 04:05 | |
| *** zhurong has quit IRC | 04:08 | |
| *** Sukhdev has quit IRC | 04:11 | |
| clayg | OMG i was in less and then something happened and I was in nano!? | 04:11 |
| clayg | s'ok - I used stackoverflow to figure out how to exit - https://stackoverflow.blog/2017/05/23/stack-overflow-helping-one-million-developers-exit-vim/ | 04:11 |
| *** zhurong has joined #openstack-swift | 04:16 | |
| *** winggundamth has joined #openstack-swift | 04:27 | |
| *** alenavolk has joined #openstack-swift | 04:34 | |
| *** klrmn has joined #openstack-swift | 04:54 | |
| *** gyee has quit IRC | 05:03 | |
| *** chsc has quit IRC | 05:07 | |
| *** klrmn has quit IRC | 05:07 | |
| *** zhengyin has quit IRC | 05:11 | |
| *** zhengyin has joined #openstack-swift | 05:11 | |
| *** skudlik has joined #openstack-swift | 05:20 | |
| clayg | acoles: kota_: I spent the afternoon with patch 448240 - I +A'd it | 05:21 |
| patchbot | https://review.openstack.org/#/c/448240/ - swift - Enable per policy proxy config options | 05:21 |
| clayg | nice work! | 05:22 |
| kota_ | clayg: ! | 05:22 |
| clayg | :\ | 05:22 |
| clayg | did i miss something!? | 05:22 |
| kota_ | clayg: no no, much appreciated | 05:22 |
| clayg | oh, my pleasure - the thanks goes to you two (and timburke) - you all worked really hard on that feature | 05:23 |
| clayg | and that one guy that did that one patch way back in the day? where'd he go? Thanks to him too | 05:23 |
| *** _JZ_ has joined #openstack-swift | 05:23 | |
| kota_ | clayg: but I'm also on that patch and maybe less than 400 Lines rest to get the end of the patch :P | 05:23 |
| clayg | OH NO! | 05:23 |
| kota_ | but currently i didn't find issues to prevent the merge | 05:24 |
| clayg | kota_: http://eavesdrop.openstack.org/irclogs/%23openstack-swift/latest.log.html#t2017-05-25T01:36:40 | 05:24 |
| clayg | :\ | 05:24 |
| kota_ | i just am willing to understand whole even it was merged | 05:24 |
| clayg | excellent! | 05:25 |
| clayg | i'm going to need to take off | 05:25 |
| kota_ | yeah, it's your late time | 05:25 |
| kota_ | probably some of my comments will be included into timburke's follow up | 05:26 |
| *** cshastri has joined #openstack-swift | 05:36 | |
| *** NareshA has joined #openstack-swift | 05:46 | |
| NareshA | Hi folks | 05:46 |
| NareshA | swift stat giving 503 service unavailable error | 05:47 |
| *** alenavolk has quit IRC | 05:48 | |
| NareshA | which is the right place I can get instant help to debug my issue? | 06:04 |
| *** Dinesh_Bhor has quit IRC | 06:06 | |
| *** Dinesh_Bhor has joined #openstack-swift | 06:11 | |
| *** rcernin has joined #openstack-swift | 06:18 | |
| *** pcaruana has joined #openstack-swift | 06:20 | |
| openstackgerrit | Merged openstack/swift master: Enable per policy proxy config options https://review.openstack.org/448240 | 06:34 |
| *** wer has quit IRC | 06:44 | |
| *** _JZ_ has quit IRC | 07:16 | |
| *** oshritf has joined #openstack-swift | 07:29 | |
| *** zhurong has quit IRC | 07:30 | |
| *** oshritf has quit IRC | 07:32 | |
| *** oshritf has joined #openstack-swift | 07:33 | |
| *** oshritf has quit IRC | 07:36 | |
| *** oshritf has joined #openstack-swift | 07:39 | |
| *** zhurong has joined #openstack-swift | 07:45 | |
| openstackgerrit | Kota Tsuyuzaki proposed openstack/swift master: Follow up for affinity config per policy https://review.openstack.org/467920 | 07:49 |
| kota_ | timburke: perhaps, https://review.openstack.org/467920 can be picked to your follow up. That is for minor things I found in the reviews. | 07:54 |
| patchbot | patch 467920 - swift - Follow up for affinity config per policy | 07:54 |
| *** oshritf has quit IRC | 07:56 | |
| acoles | good morning | 07:57 |
| *** oshritf has joined #openstack-swift | 08:03 | |
| *** oshritf has quit IRC | 08:05 | |
| acoles | NareshA: start by looking in the proxy server logs | 08:10 |
| *** sheel has joined #openstack-swift | 08:13 | |
| NareshA | acoles: thank you. I figured out the error. Its working now. | 08:15 |
| acoles | NareshA: great! | 08:15 |
| *** oshritf has joined #openstack-swift | 08:16 | |
| *** NareshA has quit IRC | 08:16 | |
| *** oshritf has quit IRC | 08:18 | |
| *** joeljwright has joined #openstack-swift | 08:33 | |
| *** ChanServ sets mode: +v joeljwright | 08:33 | |
| *** oshritf has joined #openstack-swift | 08:38 | |
| *** oshritf has quit IRC | 08:41 | |
| *** zhurong has quit IRC | 08:54 | |
| *** zhurong has joined #openstack-swift | 09:04 | |
| *** oshritf has joined #openstack-swift | 09:13 | |
| openstackgerrit | Merged openstack/swift master: Clean up how PatchPolicies works https://review.openstack.org/460362 | 09:20 |
| *** zhurong has quit IRC | 09:35 | |
| *** mvk_ has quit IRC | 09:41 | |
| *** adriant has quit IRC | 09:58 | |
| *** tovin07_ has quit IRC | 10:00 | |
| *** mvk has joined #openstack-swift | 10:09 | |
| *** sheel has quit IRC | 10:22 | |
| *** mingyu has joined #openstack-swift | 10:48 | |
| openstackgerrit | Alistair Coles proposed openstack/swift master: Add link from policies overview to per-policy proxy-server conf https://review.openstack.org/467994 | 10:49 |
| *** mingyu has quit IRC | 10:53 | |
| *** kei_yama has quit IRC | 11:23 | |
| *** zhengyin has quit IRC | 11:36 | |
| openstackgerrit | Alistair Coles proposed openstack/swift master: Update Global EC docs with reference to composite rings https://review.openstack.org/468011 | 11:48 |
| acoles | timburke: I looked over https://review.openstack.org/#/c/450491 and it all looks sane but I did have one question. If the answer is trivial then I would not object to the patch being squashed with parent. | 12:08 |
| patchbot | patch 450491 - swift - fixup! Add container sharding to Swift containers | 12:08 |
| *** NM has joined #openstack-swift | 12:22 | |
| *** esnyder has quit IRC | 12:52 | |
| *** links has quit IRC | 12:54 | |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/swift master: Updated from global requirements https://review.openstack.org/88736 | 13:01 |
| *** oshritf has quit IRC | 13:06 | |
| *** chlong has quit IRC | 13:06 | |
| *** oshritf has joined #openstack-swift | 13:09 | |
| *** klamath has joined #openstack-swift | 13:09 | |
| *** klamath has quit IRC | 13:09 | |
| *** klamath has joined #openstack-swift | 13:10 | |
| *** psachin has quit IRC | 13:13 | |
| *** oshritf has quit IRC | 13:14 | |
| *** lucasxu has joined #openstack-swift | 13:16 | |
| *** lucasxu has quit IRC | 13:27 | |
| *** lucasxu has joined #openstack-swift | 13:27 | |
| *** lucasxu has quit IRC | 13:33 | |
| *** lucasxu has joined #openstack-swift | 13:34 | |
| *** oshritf has joined #openstack-swift | 13:40 | |
| *** lucasxu has quit IRC | 13:43 | |
| *** lucasxu has joined #openstack-swift | 13:43 | |
| *** oshritf has quit IRC | 13:49 | |
| *** oshritf has joined #openstack-swift | 13:53 | |
| *** oshritf has quit IRC | 14:14 | |
| *** _JZ_ has joined #openstack-swift | 14:14 | |
| *** oshritf has joined #openstack-swift | 14:16 | |
| *** oshritf has quit IRC | 14:23 | |
| *** oshritf has joined #openstack-swift | 14:25 | |
| *** chlong has joined #openstack-swift | 14:27 | |
| *** vint_bra has joined #openstack-swift | 14:27 | |
| *** lucasxu has quit IRC | 14:39 | |
| *** kestrels has joined #openstack-swift | 14:40 | |
| *** gyee has joined #openstack-swift | 14:50 | |
| *** lucasxu has joined #openstack-swift | 14:56 | |
| *** NM has left #openstack-swift | 14:59 | |
| *** cshastri has quit IRC | 15:06 | |
| openstackgerrit | Tim Burke proposed openstack/swift master: fixup! Replace slowdown option with *_per_second option https://review.openstack.org/467801 | 15:19 |
| openstackgerrit | Tim Burke proposed openstack/swift master: Log deprecation warnings for run_pause https://review.openstack.org/468065 | 15:27 |
| *** _JZ_ has quit IRC | 15:28 | |
| *** pcaruana has quit IRC | 15:38 | |
| *** pcaruana has joined #openstack-swift | 15:39 | |
| notmyname | good morning | 15:39 |
| notmyname | https://review.openstack.org/#/c/448240/ | 15:42 |
| patchbot | patch 448240 - swift - Enable per policy proxy config options (MERGED) | 15:42 |
| notmyname | it landed! | 15:42 |
| notmyname | thanks kota_ clayg timburke and acoles | 15:42 |
| *** jrichli has quit IRC | 15:47 | |
| *** jrichli has joined #openstack-swift | 15:49 | |
| *** joeljwright1 has joined #openstack-swift | 15:50 | |
| *** joeljwright has quit IRC | 15:53 | |
| *** oshritf has quit IRC | 15:54 | |
| openstackgerrit | Tim Burke proposed openstack/swift master: Log deprecation warnings for account_white/blacklist https://review.openstack.org/468077 | 15:54 |
| *** joeljwright1 has quit IRC | 15:56 | |
| *** lucasxu has quit IRC | 16:03 | |
| *** lucasxu has joined #openstack-swift | 16:03 | |
| mwheckmann | hello. Can someone elaborate on what the "post_quorum_timeout" does exactly? The doc isn't super clear. Specifically, I'm interested in the effect of quorum and response time back to clients. | 16:04 |
| mwheckmann | for writes especially | 16:04 |
| *** alenavolk has joined #openstack-swift | 16:08 | |
| *** chsc has joined #openstack-swift | 16:09 | |
| *** chsc has joined #openstack-swift | 16:09 | |
| openstackgerrit | Tim Burke proposed openstack/swift master: Log deprecation warning for allow_versions https://review.openstack.org/468084 | 16:11 |
| mwheckmann | while I'm at asking questions, here's another one: when using write_affinity, a DELETE verb on an object will return a 404 if said object has been replicated to all it's primaries yet. I'm getting end users who are complaining about this behaviour. I get that it's likely returning the 404 since it hasn't found the object in all it's primary locations, but is there anything that can be done | 16:12 |
| mwheckmann | about that? Of course, the delete actually still works in the eventually consistent sense: The obect dissapears from the container listing but is still GET'able until the handoffs have been cleared | 16:13 |
| mwheckmann | of course I meant if "..said object *hasn't* been replicated.." | 16:13 |
| acoles | mwheckmann: re post_quorum_timeout, assume the PUT is for a 3 replicas policy, IIRC once 2 replicas have been successfully PUT, we have a quorum and could return success to client right then. But then the 3rd replica write would get terminated, so instead the proxy waits a little longer to see if the 3rd replica write completes. But if the 3rd write is really slow then the proxy returns to client after post_quorum_timeout | 16:15 |
| acoles | . | 16:15 |
| clayg | I'm not really sure how the last laggard request is closed/cleaned up. | 16:18 |
| acoles | maybe 'terminated' was wrong word - the slow write's response wouldn't be considered in evaluating the client response | 16:20 |
| mwheckmann | acoles, clayg:hmm. interesting. I currently have post_quorum_timeout set to 0.5 sec and my experience with it is that it always waits for all replicas to finish. This is when I tested on a 6 replica globally replicated ring (3 regions) and when combined with write_affinity + write_affinity_node_count=3 it seems to always wait for all remote PUTs to finish. i.e since quorum for 6 replicas = 3, | 16:20 |
| mwheckmann | I would expect it to simple write to 3 local locations and that's it. Replicatos would handle the rest. | 16:20 |
| *** _JZ_ has joined #openstack-swift | 16:20 | |
| mwheckmann | basically I want to try to limit to 1 single local handoff location instead of 4 handoffs. (re: the issue I'm having with DELETEs). | 16:22 |
| *** alenavolk has quit IRC | 16:26 | |
| *** pcaruana has quit IRC | 16:30 | |
| *** JimCheung has joined #openstack-swift | 16:30 | |
| *** pcaruana has joined #openstack-swift | 16:30 | |
| *** alenavolk has joined #openstack-swift | 16:31 | |
| openstackgerrit | Tim Burke proposed openstack/swift master: Remove deprecated vm_test_mode option https://review.openstack.org/468099 | 16:36 |
| *** chsc has quit IRC | 16:37 | |
| *** alenavolk has quit IRC | 16:38 | |
| *** rcernin has quit IRC | 16:40 | |
| openstackgerrit | Tim Burke proposed openstack/swift master: Require that known-bad EC schemes be deprecated https://review.openstack.org/468105 | 16:56 |
| timburke | so much house-cleaning this morning! | 16:56 |
| *** mvk has quit IRC | 16:57 | |
| tdasilva | timburke: great! thank you!! | 17:01 |
| clayg | timburke: can you patch the vm_test_mode param in http://programmerthoughts.com/openstack/swift-on-pi/ as well - i know it came up recently on the ML as a document that was being used to guide new user/contributors through that configuration. ;) | 17:11 |
| timburke | clayg: i think i know a guy that can help with that | 17:13 |
| clayg | ok, i'll try to do some googles and track down the rest of them for you - but think of how much better swift will be once we don't have that deprecation warning in there anymore! | 17:14 |
| clayg | sorry forgot the ;) | 17:14 |
| timburke | clayg: think of how much better swift will be once we don't have this deprecation warning: https://github.com/openstack/swift/blob/2.14.0/swift/common/middleware/copy.py#L278-L283 | 17:15 |
| timburke | if we establish a history of not following through on deprecations, we'll *never* be able to remove them | 17:16 |
| tdasilva | timburke: when are we removing that one???? :D | 17:17 |
| timburke | tdasilva: i've got my calendar reminder set for 2018-02-26 | 17:18 |
| tdasilva | we should have asked operators last summit | 17:18 |
| timburke | started warning in ocata (2.13.0), when that's eol, i won't have any qualms | 17:19 |
| tdasilva | I hope ops show up at PTG, otherwise I have a feeling we won't see them until Vancouver | 17:20 |
| *** lucasxu has quit IRC | 17:20 | |
| *** Renich has joined #openstack-swift | 17:20 | |
| *** klrmn has joined #openstack-swift | 17:28 | |
| *** chsc has joined #openstack-swift | 17:36 | |
| *** chsc has joined #openstack-swift | 17:36 | |
| openstackgerrit | Tim Burke proposed openstack/swift master: Require that known-bad EC schemes be deprecated https://review.openstack.org/468105 | 17:42 |
| *** oshritf has joined #openstack-swift | 17:58 | |
| timburke | clayg: happy now? https://github.com/notmyname/swift_on_pi/pull/3 | 17:59 |
| clayg | trying to make clayg is non-goal - at best you can aim to make him "less unhappy" - but it's a looser's game | 18:00 |
| *** oshritf has quit IRC | 18:07 | |
| tdasilva | clayg, kota_, timburke, notmyname we haven't done a libec+pyeclib release in a while and there's some good stuff there since last release, wdyt??? | 18:22 |
| *** lucasxu has joined #openstack-swift | 18:25 | |
| openstackgerrit | Merged openstack/liberasurecode master: Allow to disable optimizations for portability https://review.openstack.org/467761 | 18:28 |
| *** JimCheung has quit IRC | 18:30 | |
| clayg | @tdasilva love it! new libec+pyeclib let's do it! | 18:31 |
| openstackgerrit | Thiago da Silva proposed openstack/pyeclib master: add six to test-requirements https://review.openstack.org/468125 | 18:32 |
| tdasilva | clayg: looking through outstanding patches... | 18:32 |
| clayg | @timburke jokes aside - you're right about "neglected and forgotten" being a bad deprecation "strategy" for things like post-as-copy - for things like the isa-l configs I'm *incredibly* happy to see better maintainers than I keeping their eye on the ball - we *really should* try to push operators that may be unaware to do the right thing. I think a ML post | 18:34 |
| clayg | to openstack-operators about that change is *also* a good idea - KUDOS | 18:34 |
| clayg | for stuff like vm_test_mode or some other rando config option with a few lines of compatibility shim - meh - #1 bigger fish #2 you can't fix *all* the internets rando tutorials - it's not unreasonable to find something someone wrote about say... mysql in the early 2000's and I can *still* copy and paste it into my linux machine word for word and *most* stuff | 18:36 |
| clayg | still works. | 18:36 |
| clayg | I know "2 years EOL'd" *seems* like a long time now, but it's not... | 18:36 |
| clayg | for stuff that is mainly just a burden to maintainers - and esp if that burden isn't that high - I value the smooth experience of the user/operator much higher than my own personal comfort | 18:37 |
| clayg | $0.02 | 18:38 |
| *** aselius has joined #openstack-swift | 18:40 | |
| *** Sukhdev has joined #openstack-swift | 18:43 | |
| mwheckmann | 18:55 | |
| *** pcaruana has quit IRC | 19:09 | |
| *** alenavolk has joined #openstack-swift | 19:10 | |
| *** mvk has joined #openstack-swift | 19:11 | |
| openstackgerrit | Tim Burke proposed openstack/swift master: Remove tempauth docs from deployment guide https://review.openstack.org/466953 | 19:35 |
| *** jamielennox is now known as jamielennox|away | 19:56 | |
| *** Sukhdev has quit IRC | 19:57 | |
| openstackgerrit | Tim Burke proposed openstack/swift master: Remove deprecated vm_test_mode option https://review.openstack.org/468099 | 20:03 |
| *** alenavolk has quit IRC | 20:04 | |
| *** chlong has quit IRC | 20:15 | |
| *** lucasxu has quit IRC | 20:21 | |
| *** lucasxu has joined #openstack-swift | 20:22 | |
| *** chlong has joined #openstack-swift | 20:33 | |
| *** Renich has quit IRC | 20:36 | |
| *** chlong has quit IRC | 20:37 | |
| *** chlong has joined #openstack-swift | 20:38 | |
| *** alenavolk has joined #openstack-swift | 20:40 | |
| *** skudlik has quit IRC | 20:40 | |
| *** vint_bra has quit IRC | 21:01 | |
| *** vint_bra has joined #openstack-swift | 21:02 | |
| *** itlinux has joined #openstack-swift | 21:05 | |
| *** chlong has quit IRC | 21:11 | |
| openstackgerrit | Tim Burke proposed openstack/swift master: Improve time_check output https://review.openstack.org/468168 | 21:20 |
| *** kestrels has quit IRC | 21:27 | |
| clayg | timburke: I +A'd patch 466953 - is that cool?! | 21:36 |
| patchbot | https://review.openstack.org/#/c/466953/ - swift - Remove tempauth docs from deployment guide | 21:36 |
| timburke | love it! | 21:36 |
| timburke | thanks clayg :-) | 21:36 |
| clayg | zaitcev: maybe I should have asked you first :\ | 21:36 |
| notmyname | timburke: thanks for updating the priority reviews page | 21:37 |
| notmyname | https://wiki.openstack.org/wiki/Swift/PriorityReviews <<-- has links to the current global ec follow-up patches | 21:37 |
| *** alenavolk has quit IRC | 21:37 | |
| *** alenavolk has joined #openstack-swift | 21:38 | |
| *** alenavolk has quit IRC | 21:38 | |
| *** lucasxu has quit IRC | 21:39 | |
| *** itlinux has quit IRC | 21:42 | |
| mattoliverau | Morning, I'm back from a short vacation, sorry I missed the meeting yesterday | 21:54 |
| timburke | good morning mattoliverau! | 21:54 |
| zaitcev | clayg: it's fine, we still have etc/proxy-server.conf-sample, so we do not totally lose the documentation. | 22:07 |
| clayg | zaitcev: yeah totally - it was a copy paste not a cut | 22:07 |
| *** itlinux has joined #openstack-swift | 22:22 | |
| notmyname | how are we supposed to model durability of global ec policies? if you've got 5+3 with 2x replication, how does the 2x replication factor into the durability calculation? it feels wrong that it would be equivalent to 10+6, but I don't know | 22:25 |
| timburke | notmyname: i've previously played around with a probabilistic approach in https://gist.github.com/tipabu/b614587d2e978df8438c0250cf353ebc -- if you don't care about timings (as i did when looking at duplication vs increased parities) and are happy to assume that `k` unique frags => can reconstruct, you can probably simplify it a bit, not even make an ECDriver, and speed it up a whole lot in the process | 22:32 |
| notmyname | interesting | 22:33 |
| timburke | then i used https://gist.github.com/tipabu/454d8859d997110ba0acbca36ab6b7ec to make graphs of # failures across the cluster vs. P[can_reconstruct] | 22:34 |
| notmyname | kota_ and cschwede_ had worked on http://notmyname.github.io/swift-durability-calculator/ previously (that's my fork of charz's for of kota_'s fork of cschwede_'s fork) | 22:34 |
| timburke | you could tweak it to have the y axis be log-scale, to simplify the "how many 9s" question | 22:35 |
| *** jamielennox|away is now known as jamielennox | 22:36 | |
| *** vint_bra has quit IRC | 22:36 | |
| timburke | yeah, i debated about a more-pure-math approach, but figured trying to properly account for duplicates would leave me much less satisfied with my correctness | 22:38 |
| *** itlinux has quit IRC | 23:07 | |
| notmyname | what's quorum size on a replicated ec policy? | 23:27 |
| timburke | notmyname: n_data * duplication_factor https://github.com/openstack/swift/blob/master/swift/common/storage_policy.py#L607 | 23:30 |
| notmyname | thanks | 23:30 |
| timburke | though i'm not really sure that's a *good* choice... | 23:31 |
| *** chsc has quit IRC | 23:32 | |
| mattoliverau | I thought you'd make it the same as it means things will still work in a region split brain scenario.. but I havent really thought about it hard yet cause no coffee ;) | 23:32 |
| timburke | er, (n_data + some per-ec-scheme value) * dup_factor | 23:33 |
| notmyname | yeah, and are they unique fragments? or per region? | 23:33 |
| timburke | yeah, that's what i'm thinking right now too... i think it was done to be defensive? | 23:33 |
| notmyname | or? | 23:33 |
| mattoliverau | Or do we only care about access is split brain not continuing to put | 23:33 |
| timburke | as i recall, the putter will preferentially send out unique frags, at least to handoffs | 23:34 |
| timburke | ie if both primaries for frag 1 are down and only one of the ones for frag 2, frag 1 is *definitely* getting prioritized for handoffs | 23:35 |
| *** _JZ_ has quit IRC | 23:43 | |
| *** alenavolk has joined #openstack-swift | 23:47 | |
| *** klamath has quit IRC | 23:48 | |
| *** alenavolk has quit IRC | 23:51 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!