| kota_ | jrichli: the bug you are looking at (https://bugs.launchpad.net/swift/+bug/1662159) looks like to have something related to https://bugs.launchpad.net/swift/+bug/1493303 | 00:00 |
|---|---|---|
| openstack | Launchpad bug 1662159 in OpenStack Object Storage (swift) "TCP leak between proxy and object-server on client disconnection" [Undecided,New] | 00:00 |
| openstack | Launchpad bug 1493303 in Ubuntu Cloud Archive liberty "[OSSA 2016-004] Swift proxy memory leak on unfinished read (CVE-2016-0738)" [Undecided,Triaged] | 00:01 |
| kota_ | jrichli: i don't look at it deelpy yet but the previous bug of proxy memory leak is for slo get, and the bug you are in looks slo put. | 00:01 |
| jrichli | kota_: oh, thanks. i had looked at a previous bug that we had that was similar, and fixed at the time of encryption footers. but I didn't see this one. | 00:01 |
| jrichli | kota_: except that when I chatted with jcaron, he said that he could reproduce the issue without slo. | 00:02 |
| kota_ | that may be good to dig some of resp or req body iter is closed correctly if the leak is available on your env. | 00:02 |
| kota_ | jrichli: ah, ok. | 00:03 |
| timburke | from my read, it seemed like slo just made it more obvious, since there would be a bunch of connections | 00:03 |
| jrichli | I have not yet reproduced the issue. jcaron also has tried using a fresh SAIO and could not. our going theory is that there is some config value they might run with that is not default. | 00:03 |
| jrichli | timburke: that is correct | 00:03 |
| kota_ | jrichli: kk, it needs more deeper eyes and the way to reproduce | 00:04 |
| kota_ | timburke: thx! | 00:04 |
| jrichli | I have looked at the different config options client_timeout and such, and even tested with different values. but nothing there so far. | 00:04 |
| kota_ | jrichli: just notify you to that ;-) | 00:04 |
| kota_ | ok | 00:04 |
| jrichli | kota_: Thanks! I will take a look | 00:04 |
| jrichli | I noticed that we have a "test_PUT_timeout_during_transfer_data" in the test_obj.py which sounds like it should be revealing this issue - or could if we find out what the real trigger is | 00:05 |
| clayg | jrichli: when I'm looking for connection leaks I always start with a slow client https://gist.github.com/clayg/a4c96c9b1d5819bf83e7 | 00:07 |
| jrichli | clayg: thanks! | 00:07 |
| kota_ | nice | 00:08 |
| clayg | then I netstat until I have all the connections in my notes - client -> proxy, proxy <- client, proxy -> storage, storage <- proxy | 00:08 |
| clayg | then I KILL things | 00:08 |
| kota_ | clayg: hello, sorry for blocking your patch :/ | 00:08 |
| clayg | kota_: it's no problem! | 00:08 |
| kota_ | clayg: but it looks you can read my thought via telepathy | 00:09 |
| * kota_ is reading the gerrit comment | 00:09 | |
| clayg | kota_: you explained your concern very clearly - I think it is valid - I think jrichli has a plan to fix everything tho - i'll e fine | 00:09 |
| kota_ | I was thinking we could have an option to add "handoff_only" on my bed. | 00:09 |
| kota_ | while asleep | 00:09 |
| clayg | lol! that was jrichli! :D | 00:09 |
| clayg | maybe *she* can read minds!? | 00:10 |
| jrichli | what?! that was clayg. | 00:10 |
| kota_ | so +1 to add handoff_only for both of replicator and reconstructor | 00:10 |
| jrichli | I only mentioned the handoff_only because clayg had mentioned it in the commit message :-) and maybe in a comment too | 00:10 |
| notmyname | clayg: do you remember (or still use) the mempeek middleware? do you use that for finding leaks? | 00:10 |
| kota_ | either is fine if handoff_first should be deprecated, for now, to me. | 00:11 |
| clayg | notmyname: I only barely remember it - I definately have never used it to troubleshoot an actual memory leak | 00:11 |
| clayg | kota_: jrichli: ok, based on your support I may try to spin something up ahead of the meeting | 00:11 |
| openstackgerrit | Tim Burke proposed openstack/swift master: Prevent traceback in object-server on client disconnect https://review.openstack.org/417187 | 00:12 |
| clayg | but I'm also trying to get a SwiftStack release out this week - and spend more time with global EC! | 00:12 |
| jrichli | clayg: typey, typey! | 00:12 |
| kota_ | clayg: yey! I think almost of your concern addressed (or commented) and I have to resolve a (small...??) gate failure which seems freaky | 00:13 |
| kota_ | in my global ec | 00:13 |
| clayg | kota_: yes - you are a rockstar - I'll make time before bed | 00:13 |
| kota_ | clayg: i could have another *crazy* thought for the handoff_first/only mode | 00:19 |
| kota_ | clayg: now we're thinking of only *rebalance* cases for that | 00:19 |
| kota_ | clayg: but I'm still thinking of global cluster case which could have a lot of handoffs | 00:20 |
| kota_ | in a region | 00:20 |
| kota_ | so perhaps, the concurrency for each *normal* replicator and *hanoff_only* replicator setting could be nice to have? | 00:21 |
| kota_ | e.g. normal * 2 and handoff_only * 1, it prioritize the ratio as 2:1 for primary: handoff | 00:21 |
| kota_ | not sure if it's convinient | 00:22 |
| kota_ | and it should be futher work after handoff only | 00:23 |
| kota_ | future | 00:23 |
| kota_ | just my dump of my idea. | 00:23 |
| clayg | kota_: that's a great thought! | 00:41 |
| clayg | kota_: I've been thinking about a "heartbeat" sort of idea | 00:42 |
| kota_ | clayg: k, I'll add it to ptg topics | 00:42 |
| clayg | I think hummingbird had something where it tried to target checking one primary part per second - or something like this - a target rate - I could see the target rate for retrying a failed handoff being much higher than the target rate of syncing primaries during a rebalance | 00:44 |
| *** zaitcev has quit IRC | 00:45 | |
| kota_ | oh, sounds nice | 00:45 |
| * kota_ should learn about REPCON protocol | 00:46 | |
| notmyname | FYI Intel Avoton chips are bricking some machines (I've seen these for storage, so people in here may be affected http://www.theregister.co.uk/2017/02/06/cisco_intel_decline_to_link_product_warning_to_faulty_chip/ | 00:56 |
| *** JimCheung has quit IRC | 00:58 | |
| *** JimCheung has joined #openstack-swift | 00:59 | |
| kota_ | hello JimCheung! | 01:01 |
| kota_ | JimCheung: sorry my delay for the patch 424353 but it's still in my rader | 01:01 |
| patchbot | https://review.openstack.org/#/c/424353/ - liberasurecode - Add Phazr.IO libphazr backend to liberasurecode | 01:01 |
| *** JimCheung has quit IRC | 01:03 | |
| *** caiobrentano has joined #openstack-swift | 01:22 | |
| *** _JZ_ has joined #openstack-swift | 01:48 | |
| *** caiobrentano has quit IRC | 01:51 | |
| *** tqtran has quit IRC | 01:53 | |
| *** winggundamth_ has joined #openstack-swift | 01:58 | |
| *** dmorita has quit IRC | 02:01 | |
| *** dmorita has joined #openstack-swift | 02:03 | |
| *** tanee is now known as tanee_away | 02:07 | |
| *** tanee_away is now known as tanee | 02:07 | |
| *** JimCheung has joined #openstack-swift | 02:48 | |
| *** JimCheung has quit IRC | 02:55 | |
| *** JimCheung has joined #openstack-swift | 02:56 | |
| openstackgerrit | gengchc2 proposed openstack/swift master: Use https instead of http for git.openstack.org https://review.openstack.org/430004 | 02:58 |
| *** JimCheung has quit IRC | 03:08 | |
| mahatic | good morning | 03:15 |
| mahatic | clayg: glad they're of help! | 03:15 |
| mahatic | re: reviews | 03:15 |
| *** JimCheung has joined #openstack-swift | 03:21 | |
| *** kei_yama has joined #openstack-swift | 03:22 | |
| *** links has joined #openstack-swift | 03:41 | |
| *** JimCheung has quit IRC | 03:51 | |
| *** winggundamth has joined #openstack-swift | 03:58 | |
| *** takashi has joined #openstack-swift | 03:59 | |
| *** winggundamth has quit IRC | 04:04 | |
| *** dmorita has quit IRC | 04:04 | |
| *** winggundamth has joined #openstack-swift | 04:08 | |
| *** bkopilov_ has joined #openstack-swift | 04:12 | |
| *** bkopilov has joined #openstack-swift | 04:12 | |
| *** SkyRocknRoll has joined #openstack-swift | 04:16 | |
| jrichli | clayg: seems that sleeps in the upload did the trick ;-) | 04:17 |
| jrichli | i dont have a solution yet, but i could reproduce it | 04:19 |
| *** dmorita has joined #openstack-swift | 04:23 | |
| *** winggundamth has quit IRC | 04:25 | |
| *** dmorita has quit IRC | 04:27 | |
| *** klrmn has quit IRC | 04:27 | |
| *** tonanhngo_ has quit IRC | 04:47 | |
| *** psachin has joined #openstack-swift | 04:49 | |
| *** winggundamth has joined #openstack-swift | 04:51 | |
| *** tonanhngo has joined #openstack-swift | 04:52 | |
| *** adu has joined #openstack-swift | 04:55 | |
| *** winggundamth has quit IRC | 05:01 | |
| *** adu has quit IRC | 05:03 | |
| *** _JZ_ has quit IRC | 05:05 | |
| *** adu has joined #openstack-swift | 05:11 | |
| *** winggundamth has joined #openstack-swift | 05:20 | |
| *** winggundamth has quit IRC | 05:25 | |
| *** dmorita has joined #openstack-swift | 05:36 | |
| *** adu has left #openstack-swift | 05:36 | |
| openstackgerrit | Kota Tsuyuzaki proposed openstack/swift master: EC Fragment Duplication - Foundational Global EC Cluster Support https://review.openstack.org/219165 | 05:39 |
| *** winggundamth has joined #openstack-swift | 05:40 | |
| *** winggundamth has quit IRC | 05:45 | |
| *** dmorita has quit IRC | 05:47 | |
| *** takashi has quit IRC | 05:49 | |
| *** sams-gleb has joined #openstack-swift | 05:50 | |
| *** tqtran has joined #openstack-swift | 05:51 | |
| *** links has quit IRC | 05:51 | |
| *** sams-gleb has quit IRC | 05:52 | |
| *** tqtran has quit IRC | 05:55 | |
| *** winggundamth has joined #openstack-swift | 05:58 | |
| *** JimCheung has joined #openstack-swift | 06:00 | |
| *** rcernin has joined #openstack-swift | 06:00 | |
| *** dmorita has joined #openstack-swift | 06:01 | |
| *** rcernin has quit IRC | 06:03 | |
| *** rcernin has joined #openstack-swift | 06:04 | |
| *** JimCheung has quit IRC | 06:05 | |
| *** winggundamth has quit IRC | 06:06 | |
| *** dmorita has quit IRC | 06:07 | |
| *** links has joined #openstack-swift | 06:07 | |
| *** mweshi has joined #openstack-swift | 06:13 | |
| *** sams-gleb has joined #openstack-swift | 06:15 | |
| *** links has quit IRC | 06:22 | |
| *** winggundamth has joined #openstack-swift | 06:26 | |
| openstackgerrit | Anh Tran proposed openstack/swift master: Typo fix: curent => current https://review.openstack.org/430085 | 06:35 |
| *** links has joined #openstack-swift | 06:38 | |
| *** sams-gleb has quit IRC | 06:44 | |
| *** sams-gleb has joined #openstack-swift | 06:45 | |
| *** winggundamth has quit IRC | 06:49 | |
| jrichli | well, i take it back. i might not have reproduced the real issue. | 07:03 |
| *** links has quit IRC | 07:07 | |
| *** mweshi_ has joined #openstack-swift | 07:12 | |
| *** mweshi has quit IRC | 07:14 | |
| *** links has joined #openstack-swift | 07:19 | |
| *** mweshi_ has quit IRC | 07:20 | |
| *** links has quit IRC | 07:30 | |
| *** tesseract has joined #openstack-swift | 07:31 | |
| *** tonanhngo has quit IRC | 07:36 | |
| *** ChubYann has quit IRC | 07:37 | |
| *** links has joined #openstack-swift | 07:42 | |
| *** pcaruana has joined #openstack-swift | 07:46 | |
| *** pcaruana has quit IRC | 07:48 | |
| *** pcaruana has joined #openstack-swift | 07:55 | |
| openstackgerrit | Merged openstack/swift master: Typo fix: curent => current https://review.openstack.org/430085 | 07:56 |
| *** jamielennox is now known as jamielennox|away | 07:59 | |
| *** takashi has joined #openstack-swift | 07:59 | |
| *** tonanhngo has joined #openstack-swift | 08:03 | |
| *** tonanhngo has quit IRC | 08:04 | |
| *** jamielennox|away is now known as jamielennox | 08:06 | |
| *** dmorita has joined #openstack-swift | 08:07 | |
| *** dmorita has quit IRC | 08:12 | |
| *** rledisez has joined #openstack-swift | 08:13 | |
| *** kei_yama has quit IRC | 08:32 | |
| *** Shashikant86 has joined #openstack-swift | 08:42 | |
| *** oshritf has joined #openstack-swift | 08:45 | |
| *** Shashikant86 has quit IRC | 08:46 | |
| *** oshritf has quit IRC | 08:47 | |
| *** geaaru has joined #openstack-swift | 08:49 | |
| *** Shashikant86 has joined #openstack-swift | 08:51 | |
| *** tqtran has joined #openstack-swift | 08:52 | |
| *** oshritf has joined #openstack-swift | 08:54 | |
| *** tqtran has quit IRC | 08:56 | |
| *** cbartz has joined #openstack-swift | 09:03 | |
| *** winggundamth has joined #openstack-swift | 09:16 | |
| *** silor has joined #openstack-swift | 09:17 | |
| *** jordanP has joined #openstack-swift | 09:19 | |
| *** winggundamth has quit IRC | 09:25 | |
| *** silor has quit IRC | 09:32 | |
| *** silor1 has joined #openstack-swift | 09:32 | |
| *** silor1 is now known as silor | 09:34 | |
| *** winggundamth has joined #openstack-swift | 09:34 | |
| *** Shashikant86 has quit IRC | 09:35 | |
| *** mariusv has quit IRC | 09:38 | |
| *** mariusv has joined #openstack-swift | 09:39 | |
| *** xlucas has joined #openstack-swift | 09:42 | |
| *** Shashikant86 has joined #openstack-swift | 09:46 | |
| *** winggundamth has quit IRC | 09:47 | |
| *** winggundamth has joined #openstack-swift | 09:49 | |
| *** winggundamth has quit IRC | 09:54 | |
| *** winggundamth has joined #openstack-swift | 09:55 | |
| *** Shashikant86 has quit IRC | 09:56 | |
| *** xlucas has quit IRC | 09:56 | |
| *** Shashikant86 has joined #openstack-swift | 09:57 | |
| *** Shashikant86 has quit IRC | 10:01 | |
| *** Shashikant86 has joined #openstack-swift | 10:03 | |
| *** winggundamth has quit IRC | 10:03 | |
| *** Shashikant86 has quit IRC | 10:04 | |
| *** mariusv has quit IRC | 10:10 | |
| *** winggundamth has joined #openstack-swift | 10:10 | |
| *** winggundamth has quit IRC | 10:16 | |
| *** Shashikant86 has joined #openstack-swift | 10:17 | |
| *** Shashikant86 has quit IRC | 10:18 | |
| *** Shashikant86 has joined #openstack-swift | 10:23 | |
| *** Shashikant86 has quit IRC | 10:23 | |
| *** winggundamth has joined #openstack-swift | 10:24 | |
| *** sams-gleb has quit IRC | 10:27 | |
| *** sams-gleb has joined #openstack-swift | 10:28 | |
| *** jtomasek has joined #openstack-swift | 10:32 | |
| *** sams-gleb has quit IRC | 10:32 | |
| jtomasek | Hi all, what is the recommended way to duplicate a swift container using API? Only way I see is create a new container and copy all objects to it | 10:36 |
| *** Shashikant86 has joined #openstack-swift | 10:36 | |
| *** xlucas has joined #openstack-swift | 10:40 | |
| *** oshritf has quit IRC | 10:43 | |
| *** Shashikant86 has quit IRC | 10:46 | |
| *** sams-gleb has joined #openstack-swift | 10:47 | |
| *** geaaru has quit IRC | 10:53 | |
| kota_ | jtomasek: AFAIK, it's only way to force it. Perhaps container-sync is an option to enable sync your container to another one with out copy for new incomming objects. | 10:55 |
| kota_ | jtomasek: the docs for container-sync is http://docs.openstack.org/developer/swift/overview_container_sync.html | 10:56 |
| jtomasek | kota_: ok, thanks, I'll check it | 10:57 |
| *** SkyRocknRoll has quit IRC | 10:57 | |
| *** geaaru has joined #openstack-swift | 11:05 | |
| *** mvk has quit IRC | 11:27 | |
| *** Shashikant86 has joined #openstack-swift | 11:28 | |
| ahale | perhaps if you wanted to move rather than copy you could use http://docs.openstack.org/developer/swift/overview_object_versioning.html | 11:37 |
| *** takashi has quit IRC | 11:56 | |
| *** Shashikant86 has quit IRC | 12:05 | |
| *** tonanhngo has joined #openstack-swift | 12:05 | |
| *** tonanhngo has quit IRC | 12:06 | |
| *** mvk has joined #openstack-swift | 12:07 | |
| *** cbartz has quit IRC | 12:12 | |
| *** cbartz has joined #openstack-swift | 12:13 | |
| *** cbartz has quit IRC | 12:13 | |
| *** cbartz has joined #openstack-swift | 12:15 | |
| *** bkopilov_ has quit IRC | 12:28 | |
| *** acoles_ is now known as acoles | 12:28 | |
| *** bkopilov has quit IRC | 12:28 | |
| *** links has quit IRC | 12:48 | |
| *** catintheroof has joined #openstack-swift | 12:50 | |
| *** tqtran has joined #openstack-swift | 12:53 | |
| *** tqtran has quit IRC | 12:58 | |
| *** xlucas has left #openstack-swift | 13:04 | |
| *** xlucas has joined #openstack-swift | 13:04 | |
| *** chlong has joined #openstack-swift | 13:14 | |
| *** klamath has joined #openstack-swift | 13:24 | |
| *** klamath has quit IRC | 13:24 | |
| *** klamath has joined #openstack-swift | 13:24 | |
| openstackgerrit | Merged openstack/swift master: Fix test_replicator assertion https://review.openstack.org/429673 | 13:37 |
| *** cdelatte has joined #openstack-swift | 13:43 | |
| *** remix_tj has quit IRC | 13:45 | |
| *** bkopilov_ has joined #openstack-swift | 13:51 | |
| *** bkopilov has joined #openstack-swift | 13:52 | |
| *** remix_tj has joined #openstack-swift | 13:53 | |
| *** cdelatte has quit IRC | 14:43 | |
| *** zaitcev has joined #openstack-swift | 14:48 | |
| *** ChanServ sets mode: +v zaitcev | 14:48 | |
| *** JimCheung has joined #openstack-swift | 14:53 | |
| *** tonanhngo has joined #openstack-swift | 14:58 | |
| *** JimCheung has quit IRC | 14:58 | |
| *** tonanhngo has quit IRC | 14:58 | |
| *** sams-gleb has quit IRC | 15:03 | |
| *** sams-gleb has joined #openstack-swift | 15:04 | |
| *** tqtran has joined #openstack-swift | 15:04 | |
| *** sams-gleb has quit IRC | 15:08 | |
| *** dmorita has joined #openstack-swift | 15:08 | |
| *** dmorita has quit IRC | 15:13 | |
| *** vint_bra has joined #openstack-swift | 15:13 | |
| *** vint_bra has quit IRC | 15:16 | |
| *** sams-gleb has joined #openstack-swift | 15:21 | |
| *** abqkawi1000 has joined #openstack-swift | 15:30 | |
| *** _JZ_ has joined #openstack-swift | 15:30 | |
| *** MVenesio has joined #openstack-swift | 15:32 | |
| *** abqkawi1000 has quit IRC | 15:33 | |
| *** abqkawi1000 has joined #openstack-swift | 15:34 | |
| *** cdelatte has joined #openstack-swift | 15:40 | |
| abqkawi1000 | joined hoping for some guidance on where I should be looking to fix an auth problem. Logs in my swift proxy box show "Identity server rejected authorization" "Unable to validate token: Identity server rejected auth necessary to fetch token data. | 15:44 |
| patchbot | Error: No closing quotation | 15:44 |
| *** tonanhngo has joined #openstack-swift | 15:59 | |
| *** cdelatte has quit IRC | 16:12 | |
| *** acoles is now known as acoles_ | 16:14 | |
| *** rcernin has quit IRC | 16:16 | |
| *** pcaruana has quit IRC | 16:17 | |
| *** tqtran has quit IRC | 16:20 | |
| *** klrmn has joined #openstack-swift | 16:29 | |
| *** openstackgerrit has quit IRC | 16:35 | |
| notmyname | good morning | 16:37 |
| notmyname | fungi: is this appropriate? https://review.openstack.org/#/c/430004/ | 16:38 |
| patchbot | patch 430004 - swift - Use https instead of http for git.openstack.org | 16:38 |
| fungi | notmyname: sure, that's a valid browseable url and we maintain trusted certs for the site | 16:39 |
| notmyname | fungi: kk. thanks | 16:40 |
| fungi | notmyname: if you want a more performant git url instead of a browser-oriented one, drop the /cgit from the path | 16:40 |
| fungi | then it will go through the smart git backend rather than the cgit rendering cgi | 16:41 |
| *** oshritf has joined #openstack-swift | 16:42 | |
| *** oshritf has quit IRC | 16:44 | |
| *** chsc has joined #openstack-swift | 16:47 | |
| *** chsc has joined #openstack-swift | 16:47 | |
| *** cdelatte has joined #openstack-swift | 16:49 | |
| *** psachin has quit IRC | 16:50 | |
| *** tqtran has joined #openstack-swift | 17:11 | |
| *** oshritf has joined #openstack-swift | 17:26 | |
| *** JimCheung has joined #openstack-swift | 17:27 | |
| *** oshritf has quit IRC | 17:34 | |
| *** ganders has joined #openstack-swift | 17:36 | |
| *** mvk has quit IRC | 17:37 | |
| *** nikhil|afk has joined #openstack-swift | 17:38 | |
| *** dmorita has joined #openstack-swift | 17:41 | |
| *** rledisez has quit IRC | 17:44 | |
| *** tonanhngo has quit IRC | 17:47 | |
| *** dmorita has quit IRC | 18:01 | |
| *** tesseract has quit IRC | 18:01 | |
| *** dmorita has joined #openstack-swift | 18:02 | |
| *** silor has quit IRC | 18:05 | |
| clayg | jrichli: I had posted a lengthy analysis of another socket leak backend connection buffer issue | 18:10 |
| *** winggundamth_ has quit IRC | 18:28 | |
| jrichli | clayg: where? oh no, you mean the ML? I guess i should really subscribe to that ... | 18:29 |
| jrichli | I see https://bugs.launchpad.net/swift/+bug/1466549 also was a similar issue in the past | 18:29 |
| openstack | Launchpad bug 1466549 in OpenStack Object Storage (swift) "[OSSA 2016-004] Download DLO objects leak connections when client kill connection (CVE-2016-0737)" [Undecided,Fix released] | 18:29 |
| *** openstackgerrit has joined #openstack-swift | 18:30 | |
| openstackgerrit | Christopher Bartz proposed openstack/python-swiftclient master: ISO 8601 timestamps for tempurl https://review.openstack.org/423377 | 18:30 |
| timburke | abqkawi1000: i saw a bit of the start of that in -keystone -- looked like swift was sending back a 503? that usually means the proxy was having trouble talking to the backend, like maybe the account server isn't running? | 18:31 |
| timburke | i think it could also mean that swift couldn't validate tokens against keystone; you might want to double check the creds for authtoken middleware and the roles assigned for that user | 18:32 |
| clayg | jrichli: i was thinking of my comments on this super old bug https://bugs.launchpad.net/swift/+bug/1568650 | 18:33 |
| openstack | Launchpad bug 1568650 in OpenStack Object Storage (swift) "Connection between client and proxy service does not closes" [High,Confirmed] - Assigned to drax (devesh-gupta) | 18:33 |
| clayg | afaik it's been confirmed for ~1yr | 18:33 |
| *** geaaru has quit IRC | 18:35 | |
| *** caiobrentano has joined #openstack-swift | 18:37 | |
| *** cbartz has quit IRC | 18:49 | |
| openstackgerrit | Tim Burke proposed openstack/swift master: Remove per-device reconstruction stats https://review.openstack.org/427476 | 19:01 |
| jrichli | clayg: ok, thanks. at first glance at the explanations, I can tell that I will need to take some time for a deeper look :-) | 19:04 |
| *** tonanhngo has joined #openstack-swift | 19:14 | |
| *** tonanhngo has quit IRC | 19:19 | |
| *** ChubYann has joined #openstack-swift | 19:32 | |
| *** xlucas has quit IRC | 19:33 | |
| *** stradling has joined #openstack-swift | 19:34 | |
| *** tonanhngo has joined #openstack-swift | 19:35 | |
| timburke | tdasilva: does the ring builder have any knowledge of how the ring will be used? i'm not sure it does... all of that information lives in swift.conf, yeah? | 19:38 |
| openstackgerrit | Merged openstack/swift master: Use https instead of http for git.openstack.org https://review.openstack.org/430004 | 19:39 |
| *** ganders has quit IRC | 19:44 | |
| *** ukaynar has joined #openstack-swift | 19:58 | |
| *** jordanP has quit IRC | 20:13 | |
| *** ukaynar_ has joined #openstack-swift | 20:16 | |
| *** ukaynar has quit IRC | 20:16 | |
| tdasilva | timburke: not sure I follow, what info live in swift.conf? | 20:25 |
| *** ukaynar has joined #openstack-swift | 20:35 | |
| *** ukaynar_ has quit IRC | 20:35 | |
| *** caiobrentano_ has joined #openstack-swift | 20:46 | |
| timburke | tdasilva: mapping from ring file to whether it's being used for acct, cont, or obj | 20:47 |
| *** jordanP has joined #openstack-swift | 20:48 | |
| timburke | like, i don't think we *can* have the ring builder prevent you from trying to increase the part_power of an acct/cont ring -- i don't think it has enough information to know that | 20:48 |
| *** joeljwright has joined #openstack-swift | 20:49 | |
| *** ChanServ sets mode: +v joeljwright | 20:49 | |
| *** caiobrentano has quit IRC | 20:49 | |
| *** caiobrentano_ has quit IRC | 20:50 | |
| tdasilva | timburke: Yes, I agree, but my concern was more out of practicality thinking of operators. The documentation talks about increasing partition power and it mentions objects, but it says nothing about account or containers | 20:55 |
| timburke | sure. yeah, i agree that we probably want to be more explicit in the docs | 20:55 |
| clayg | timburke: it's more like if we decide later to drop the whole function and punt to a library - someone is going to say "what about aarch64 - did anyone test it" - and then we're stuck with the reality that noone who is able to maintain the feature/moddule *ever* tested it on aarch64 | 20:55 |
| tdasilva | can we assume that operators just know that they can't increase part power for account and container rings? | 20:56 |
| *** ukaynar has quit IRC | 20:56 | |
| clayg | if we're not going to run it test it - i don't see how we can "support/maintain" it - how can we validate the use-case still works if we want to change the implementation? | 20:56 |
| *** ukaynar has joined #openstack-swift | 20:56 | |
| clayg | timburke: again, i'm all for fixing the tests | 20:57 |
| *** adu has joined #openstack-swift | 20:57 | |
| clayg | I think the unittest failure is pointing at bad coupling - using good clean pactching to avoid ties to the system running the tests squares that and the feature continues to work as it was written | 20:58 |
| timburke | clayg: same way we're doing it right now -- the thing *isn't supported* -- there's no update to the README or docs or setup.cfg implying otherwise. so if we need to change some implementation details, we change it and see if anyone balks | 21:00 |
| timburke | which is how we got where we are *right now*. he's not asking for full and proper support; he's asking for unit tests to pass. and if we can get unit tests pass *and* move closer to full and proper support, that seems better than papering over the issue and not knowing how well our tests reflect reality | 21:00 |
| clayg | i included a diff to fix the unittests | 21:00 |
| clayg | i think it's objectively better than the purposed change | 21:00 |
| clayg | I think if mattoliverau or torgomatic want to +A it i'm over it | 21:01 |
| clayg | but I also think if someone wants to change how io_prio/nice settings work - i'm going to send it your way to validate the change still works on aarch64 ;) | 21:01 |
| clayg | that's acctually my only beef - if I had a good way to validate the change on aarch64 I think I could "support" maintaining the subsystem on that arch - so i'd be find merging support - because I can test it? | 21:02 |
| clayg | idk, it doesn't seem insane to me | 21:02 |
| *** jordanP has quit IRC | 21:03 | |
| clayg | but I think choosing to do that should be orthogonal to fixing the test coupling with the aarch running the tests - i feel like i've said this - my argument is clear - but I see no reason that anyone should feel compelled to agree with me | 21:03 |
| clayg | i recognize the desire to "be helpful and merge things that people *in the communitiy like*" is very strong | 21:03 |
| clayg | and my threshold for YAGNI is maybe higher | 21:03 |
| clayg | it's totally subjective judgement | 21:04 |
| *** caiobrentano has joined #openstack-swift | 21:14 | |
| clayg | timburke: related to some of the stats output thread stuff you've been working on -> lp bug #1662672 | 21:18 |
| openstack | Launchpad bug 1662672 in OpenStack Object Storage (swift) "db replicator missing stats output" [Undecided,New] https://launchpad.net/bugs/1662672 | 21:18 |
| clayg | tdasilva: timburke: the part_power_increase relating to account/container sounds like a hozer | 21:20 |
| clayg | tdasilva: timburke: you could add something in some of the background ring reloading for a/c services to warn if they encounter something that looks part-power-increase-y | 21:21 |
| clayg | tdasilva: timburke: the relinker guy should know enough to error if you tried to envoke it on a non object ring | 21:21 |
| clayg | you could *warn* in the CLI if you try to do something part-power-increase-y on a ring named account/container - or maybe even anything not `object[-X]` ? | 21:22 |
| timburke | clayg: so how do you feel about the proprietary backends for libec? or, for that matter, jerasure? wasn't that part of your beef with https://review.openstack.org/#/c/425471/ -- "it's not something i care about and i'm not sure how to test it" so why have that support at all? | 21:22 |
| patchbot | patch 425471 - liberasurecode - Make free_rs_cauchy_desc tolerate partially-initia... | 21:22 |
| tdasilva | clayg: yeah, i was leaning towards the cli warning, it seems prudent to me | 21:24 |
| clayg | timburke: honestly my stance on jerasure is somewhat weak - being that it's a open source library and I'm a core maintainer for liberasurecode - at *some level* I should feel obliged to maintain jerasure support | 21:24 |
| clayg | ... but the reality is it has to sit below the already lowered capacity I have to prioritize changes for libec in general compared to ... e.g. upstream swift | 21:24 |
| timburke | clayg: the relinker is already smart enough to only look at object policies -- and it looks at all of them, checking if any seem to be in that transition state | 21:24 |
| timburke | you literally *can't* run it against an account/container ring | 21:25 |
| *** delattec has joined #openstack-swift | 21:25 | |
| clayg | timburke: but for something like Phzer.IO or w/e it's called - yeah i'm *super* nervous we can't test it - I *don't* think we should merge support - I think we should talk to Neutron and Cinder about how plugins to proprietary code is bad for everyone and what alternatives are available | 21:25 |
| *** cdelatte has quit IRC | 21:26 | |
| clayg | timburke: RE relinker - NICE | 21:26 |
| clayg | tdasilva: yeah a warning can't hurt - and something in the docs that goes out of it's way to explicitly state "i'm not sure where you might have got the idea this is related in anyway to a/c - but it is not - you can not increase part power on a/c" | 21:27 |
| clayg | ... of course cschwede probably already a plan underway :P | 21:27 |
| tdasilva | timburke: nice catch on symlink patch, I had not noticed that subtle change: https://review.openstack.org/#/c/232162/51/swift/common/middleware/symlink.py@247 | 21:30 |
| patchbot | patch 232162 - swift - Symlink implementation. | 21:30 |
| *** jordanP has joined #openstack-swift | 21:32 | |
| mattoliverau | morning, I'm back and I didn't sink or die while driving home.. now just have like a bazillion emails to catch up on. | 21:34 |
| notmyname | welcome back mattoliverau :-) | 21:35 |
| notmyname | mattoliverau: finally back to the stressful beach life after so many weeks away? | 21:35 |
| *** Jeffrey4l has quit IRC | 21:35 | |
| notmyname | mattoliverau: I have an ansible question for you that I hope you can help me with | 21:36 |
| *** Jeffrey4l has joined #openstack-swift | 21:36 | |
| notmyname | mattoliverau: in a role, I've got the mail.yaml. but there's a set of functionality that takes several tasks to do that I feel like should be in a different file. how do I "call" that different file from main.yaml? | 21:36 |
| *** MVenesio has quit IRC | 21:37 | |
| mattoliverau | yup, first time here this year | 21:37 |
| mattoliverau | can't wait to get down to the beach! | 21:37 |
| *** jordanP has quit IRC | 21:37 | |
| mattoliverau | notmyname: you can just create the other file in say the tasks folder, then simply include it. Where you include it in the file is where it'll play from | 21:38 |
| notmyname | oh ok | 21:38 |
| mattoliverau | notmyname: example coming | 21:38 |
| mattoliverau | notmyname: https://github.com/matthewoliver/swift-ansible/blob/master/roles/swift_common/tasks/main.yml#L7-L15 | 21:39 |
| notmyname | mattoliverau: I see. thanks | 21:40 |
| mattoliverau | are a bunch of include examples, the files they include you'll find in the tasks dir | 21:40 |
| *** jamielennox is now known as jamielennox|away | 21:42 | |
| *** catintheroof has quit IRC | 21:48 | |
| *** catintheroof has joined #openstack-swift | 21:49 | |
| *** ukaynar has quit IRC | 21:53 | |
| *** catintheroof has quit IRC | 21:54 | |
| *** ukaynar has joined #openstack-swift | 21:55 | |
| *** xlucas has joined #openstack-swift | 21:56 | |
| *** tesseract has joined #openstack-swift | 21:59 | |
| *** tesseract- has joined #openstack-swift | 22:00 | |
| *** jcaron1 has joined #openstack-swift | 22:06 | |
| *** xlucas has quit IRC | 22:06 | |
| *** tesseract has quit IRC | 22:07 | |
| *** tesseract- has quit IRC | 22:07 | |
| *** tesseract has joined #openstack-swift | 22:08 | |
| *** jcaron has quit IRC | 22:10 | |
| *** sams-gleb has quit IRC | 22:11 | |
| *** sams-gleb has joined #openstack-swift | 22:22 | |
| *** tesseract is now known as tesseract-RH | 22:26 | |
| *** tesseract-RH is now known as tesseract | 22:31 | |
| *** caiobrentano has quit IRC | 22:35 | |
| *** stradling has quit IRC | 22:36 | |
| *** jamielennox|away is now known as jamielennox | 22:42 | |
| *** sams-gleb has quit IRC | 22:57 | |
| *** sams-gleb has joined #openstack-swift | 22:58 | |
| *** tesseract has quit IRC | 23:00 | |
| *** tesseract has joined #openstack-swift | 23:00 | |
| *** tesseract-RH has joined #openstack-swift | 23:01 | |
| *** tesseract has quit IRC | 23:01 | |
| *** tesseract-RH has quit IRC | 23:01 | |
| *** sams-gleb has quit IRC | 23:02 | |
| *** JimCheung has quit IRC | 23:13 | |
| *** JimCheung has joined #openstack-swift | 23:13 | |
| *** joeljwright has quit IRC | 23:15 | |
| *** kei_yama has joined #openstack-swift | 23:23 | |
| *** chsc has quit IRC | 23:31 | |
| *** stradling has joined #openstack-swift | 23:33 | |
| *** adu has quit IRC | 23:36 | |
| *** klamath has quit IRC | 23:44 | |
| openstackgerrit | Jim Cheung proposed openstack/liberasurecode master: Add Phazr.IO libphazr backend to liberasurecode https://review.openstack.org/424353 | 23:46 |
| kota_ | good morning | 23:54 |
| kota_ | mattoliverau: welcome back! | 23:54 |
| kota_ | clayg, timburke: how's going global ec reviews? | 23:54 |
| kota_ | clayg, timburke: can i help you on something to progress (e.g. more fix, more tests or just waiting :P) it? | 23:55 |
| mattoliverau | kota_: morning | 23:56 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!