| *** david-lyle is now known as david-lyle_afk | 00:01 | |
| *** barra204 has joined #openstack-swift | 00:05 | |
| ho | good morning guys | 00:05 |
|---|---|---|
| mattoliverau | ho: morning | 00:09 |
| ho | mattoliverau: morning! morning! | 00:26 |
| notmyname | hi | 00:30 |
| ho | notmyname: hi! | 00:32 |
| *** Masahiro has joined #openstack-swift | 00:35 | |
| *** barra204 has quit IRC | 00:37 | |
| *** Masahiro has quit IRC | 00:39 | |
| *** elambert has quit IRC | 00:40 | |
| openstackgerrit | Kota Tsuyuzaki proposed openstack/swift: Fix slo constraint according to the error message https://review.openstack.org/145177 | 00:51 |
| *** ctennis has quit IRC | 00:52 | |
| *** ctennis has joined #openstack-swift | 00:52 | |
| *** Masahiro has joined #openstack-swift | 00:58 | |
| notmyname | overloaded ring weights. while I really like the idea as it's been explained to me, I feel that reviewing the patch is going to make my brain hurt | 01:14 |
| notmyname | if only it were commit messages that ran in prod instead of code | 01:14 |
| notmyname | /thanks torgomatic | 01:14 |
| torgomatic | if it was hard to write, it should be hard to read /s | 01:15 |
| notmyname | if commit messages were what ran in prod, I'd `git commit -am "made object server faster"` | 01:15 |
| notmyname | also, `git commit -am "implemented erasure codes"` | 01:16 |
| mattoliverau | lol | 01:16 |
| *** reed has quit IRC | 01:19 | |
| clayg | redbo: dfg: how up to date is git@github.com:redbo/goswiftobj.git with the other code you guys are trying out? | 01:19 |
| *** reed has joined #openstack-swift | 01:22 | |
| *** nosnos has joined #openstack-swift | 01:27 | |
| redbo | clayg: not at all, we've been working on it quite a bit. The plan is still to open source it, but general consensus is it should be not broken first. | 01:27 |
| *** lcurtis has quit IRC | 01:28 | |
| *** addnull has joined #openstack-swift | 01:30 | |
| *** reed has quit IRC | 01:40 | |
| ho | mattoliverau: In Dec 12, I talked about a patch https://review.openstack.org/#/c/89568 with you. | 02:02 |
| ho | mattoliverau: Then I sent an email to the owner but couldn't get any response. | 02:02 |
| ho | mattoliverau: I will wait for a response until Jan 12. | 02:02 |
| ho | mattoliverau: If I can not get any response, I would like to take it over to realize RBAC with Keystone. Is it possible? | 02:03 |
| *** haomaiwang has joined #openstack-swift | 02:08 | |
| ho | clayg: hi! | 02:31 |
| ho | clayg: if possible, i would like to have your re-review for https://review.openstack.org/#/c/133155/. | 02:33 |
| *** SkyRocknRoll has joined #openstack-swift | 02:39 | |
| *** SkyRocknRoll has quit IRC | 03:09 | |
| *** nosnos has quit IRC | 03:20 | |
| *** fandi has quit IRC | 03:23 | |
| *** chlong has quit IRC | 03:41 | |
| *** chlong has joined #openstack-swift | 03:42 | |
| *** chlong has quit IRC | 03:42 | |
| *** chlong has joined #openstack-swift | 03:44 | |
| *** chlong has quit IRC | 03:48 | |
| *** chlong has joined #openstack-swift | 03:48 | |
| *** ppai has joined #openstack-swift | 03:56 | |
| *** ho__ has joined #openstack-swift | 04:05 | |
| *** ho has quit IRC | 04:07 | |
| *** nosnos has joined #openstack-swift | 04:15 | |
| *** gyee has quit IRC | 04:31 | |
| openstackgerrit | Tushar Gohad proposed openstack/swift: Bump eventlet version to 0.16.0 https://review.openstack.org/145403 | 05:07 |
| *** ho has joined #openstack-swift | 05:11 | |
| *** ho__ has quit IRC | 05:13 | |
| *** addnull has quit IRC | 05:31 | |
| *** remix_auei has quit IRC | 05:37 | |
| *** remix_tj has joined #openstack-swift | 05:39 | |
| *** zaitcev has quit IRC | 05:42 | |
| *** lcurtis has joined #openstack-swift | 06:08 | |
| lcurtis | im wondering if there is anyone on who might be able answer why objects and containers keep disappearing on me in swift? | 06:10 |
| lcurtis | or maybe most obvious cause | 06:10 |
| ho | lcurtis: "keep disappearing"means not removed explicitly but looks disappering? | 06:16 |
| lcurtis | well, from horizon, the container disappears | 06:16 |
| lcurtis | and then swift list 'container' not found | 06:18 |
| lcurtis | so in that regard...yes to your question | 06:18 |
| ho | lcurtis: i see. can you share the env? | 06:19 |
| ho | lcurtis: can you use same account for the access? | 06:20 |
| lcurtis | ho: not sure what you mean | 06:20 |
| ho | lcurtis: i guess there are two candidates or more. first one is someone remove the objects. second one is you use different project-id for authentication. | 06:22 |
| lcurtis | no same project id | 06:22 |
| lcurtis | and user | 06:23 |
| ho | lcurtis: try to use same one. | 06:23 |
| lcurtis | yes i mean i use same project and user | 06:23 |
| *** km has joined #openstack-swift | 06:23 | |
| ho | lcurtis: sorry for mis-understanding. i have no idea... | 06:24 |
| lcurtis | thanks for your help all the same | 06:24 |
| lcurtis | i will try recreating... | 06:24 |
| lcurtis | good night all...be back on tomorrow | 06:24 |
| ho | lcurtis: good night! | 06:25 |
| wer | nite | 06:25 |
| mattoliverau | lcurtis: are you using the same proxy each time? | 06:27 |
| *** ajay_ has joined #openstack-swift | 06:27 | |
| lcurtis | hmmm...perhaps | 06:28 |
| ajay_ | I get: ImportError: No module named middleware.recon, when I start the swift deamon. Can anyone help? | 06:29 |
| lcurtis | using etc hosts | 06:29 |
| lcurtis | so is probably hitting same proxy | 06:30 |
| mattoliverau | lcurtis: just wondering if your hitting different proxies (load balenced) and the if the rings don't match then the proxy may look for a container in a different place.... thus disappearing. | 06:36 |
| mattoliverau | lcurtis: does the container come and go, or does it simply disappear? | 06:36 |
| lcurtis | thats what it seems like | 06:36 |
| mattoliverau | did you check the storage server logs? is there a permissions error or something causing the container to not actaully be saved to disk | 06:37 |
| mattoliverau | lcurtis: ^ | 06:37 |
| mattoliverau | lcurtis: point directly to the IP of a proxy to make sure your going through it, and create a container (I use curl usually) and then see if it disappears, if it does probably a backend thing | 06:39 |
| ho | ajay_: maybe description of confs is wrong,. first check [pipeline:main] section in the [account|container|object]-server.conf whether there is recon. | 06:40 |
| ho | ajay_: then check [filter:recon] section whether the descption is correct. | 06:40 |
| mattoliverau | if its the backend start with the logs, check permissions, etc. | 06:40 |
| mattoliverau | lcurtis: ^ | 06:40 |
| *** lcurtis has quit IRC | 06:42 | |
| *** addnull has joined #openstack-swift | 06:42 | |
| *** lcurtis has joined #openstack-swift | 06:42 | |
| mattoliverau | lcurtis: I'm sitting in car waiting at a train station for the wife to arrive, so probably wont be online for much longer.. but will be online tomorrow and I can try and help you work it out if you need :) | 06:42 |
| lcurtis | mattoliverau--no worries | 06:43 |
| lcurtis | mattoliverau That would be fantastic....i sincerely appreciate t | 06:43 |
| lcurtis | so many helpful people here | 06:44 |
| mattoliverau | Cools, looks like the train is here, better get off the phone so I can drive. Night all | 06:46 |
| *** addnull has quit IRC | 06:46 | |
| ajay_ | ho: Thanks for replying. I have used the conf file as it is given in here: http://docs.openstack.org/developer/swift/development_saio.html | 06:59 |
| *** Masahiro has quit IRC | 07:03 | |
| *** Masahiro has joined #openstack-swift | 07:03 | |
| *** sandywalsh has quit IRC | 07:11 | |
| *** lcurtis has quit IRC | 07:11 | |
| *** sandywalsh has joined #openstack-swift | 07:13 | |
| *** silor has joined #openstack-swift | 07:22 | |
| *** Masahiro has quit IRC | 07:24 | |
| *** Masahiro has joined #openstack-swift | 07:30 | |
| *** Masahiro has quit IRC | 07:33 | |
| ho | ajay_: thanks for the info. offcource there is no problem in the doc. is it possible to remove recon entry from the pipelines? Or check recon entry in swift.egg-info/entry_points.txt, unn.... | 07:55 |
| *** fandi has joined #openstack-swift | 08:00 | |
| *** pcaruana|afk| is now known as pcaruana | 08:04 | |
| ajay_ | ho: I removed recon from pipeline. It solved the problem but gives a new error: ImportError: No module named catch_errors | 08:06 |
| *** chlong has quit IRC | 08:07 | |
| ho | ajay_: removing recon entries means not use recon. i think if you remove catch_errors entry from the conf, you may get same error for other middlewares. | 08:13 |
| *** addnull has joined #openstack-swift | 08:14 | |
| ho | ajay_: one question. did you execute the procedure "cd $HOME/swift; sudo python setup.py develop; cd -" in step4? | 08:14 |
| ajay_ | ho: Yes. I followed the exact steps. | 08:17 |
| ajay_ | I removed recon and got all the demons running except proxy. | 08:19 |
| *** geaaru has joined #openstack-swift | 08:21 | |
| ho | ajay_: in your env, swift can solve app entries like [app:container-server] but can not solve filter entries | 08:22 |
| *** addnull has quit IRC | 08:22 | |
| ho | ajay_: [filter:catch_errors] | 08:22 |
| ho | ajay_: after you executed git clone in step3, you have swift dir. is it still remained in your env? | 08:25 |
| ajay_ | ho: Yes. It is there. | 08:27 |
| ajay_ | I removed catch_errors and all the middleware modules seems to give the same error. | 08:33 |
| ajay_ | ho^ | 08:33 |
| ajay_ | ho: When I do "python -c 'import swift.common.middleware.catch_errors; print "SUCCESS"'" I get success. | 08:36 |
| ho | ajay_: thanks. i think step4 doesn't work correctly, the line "cd $HOME/swift; sudo python setup.py build; sudo python setup.py install; cd -" may work instead of step4 | 08:38 |
| ajay_ | ho: No it still didn't work. Same error. | 08:42 |
| ajay_ | I had already done python setup.py develop. And now I did a python setup.py build and install over it. | 08:44 |
| *** addnull has joined #openstack-swift | 08:49 | |
| ajay_ | ho: /var/log/messages shows this: Executable '/usr/bin/swift-proxy-server' doesn't belong to any package and ProcessUnpackaged is set to 'no' | 08:51 |
| *** Masahiro_ has joined #openstack-swift | 08:57 | |
| *** pcaruana has quit IRC | 08:59 | |
| *** nosnos has quit IRC | 09:00 | |
| ho | ajay_: i see... the message does not directly related to the case. it is one of the effect of this trouble. | 09:01 |
| *** nosnos has joined #openstack-swift | 09:01 | |
| *** jordanP has joined #openstack-swift | 09:05 | |
| *** jordanP has joined #openstack-swift | 09:05 | |
| *** nosnos has quit IRC | 09:08 | |
| *** nosnos has joined #openstack-swift | 09:10 | |
| *** pcaruana has joined #openstack-swift | 09:12 | |
| *** jistr has joined #openstack-swift | 09:14 | |
| *** rledisez has joined #openstack-swift | 09:14 | |
| *** nellysmitt has joined #openstack-swift | 09:23 | |
| *** addnull has quit IRC | 09:33 | |
| *** acoles_away is now known as acoles | 09:41 | |
| *** esmute has quit IRC | 10:03 | |
| *** esmute has joined #openstack-swift | 10:06 | |
| *** addnull has joined #openstack-swift | 10:18 | |
| *** rledisez has quit IRC | 10:20 | |
| *** rledisez has joined #openstack-swift | 10:35 | |
| *** chlong has joined #openstack-swift | 10:38 | |
| *** Masahiro_ has quit IRC | 10:41 | |
| *** haomaiwang has quit IRC | 10:42 | |
| *** Masahiro has joined #openstack-swift | 10:43 | |
| *** NM has joined #openstack-swift | 10:45 | |
| *** Masahiro has quit IRC | 11:00 | |
| *** Masahiro has joined #openstack-swift | 11:00 | |
| *** Masahiro has quit IRC | 11:06 | |
| *** chlong has quit IRC | 11:08 | |
| *** mahatic has joined #openstack-swift | 11:31 | |
| NM | Morning! | 11:32 |
| NM | ajay_: did you solve your problem? | 11:32 |
| ajay_ | Yes. Thanks. NM | 11:32 |
| NM | What happened? | 11:32 |
| ajay_ | Ho, thank you | 11:32 |
| ajay_ | [14:04] <ho> ajay_: thanks. i think step4 doesn't work correctly, the line "cd $HOME/swift; sudo python setup.py build; sudo python setup.py install; cd -" may work instead of step4 | 11:33 |
| ajay_ | This worked for me. | 11:33 |
| ho | ajay_: great! | 11:34 |
| *** Masahiro has joined #openstack-swift | 11:47 | |
| *** Masahiro has quit IRC | 11:47 | |
| *** Masahiro has joined #openstack-swift | 11:48 | |
| *** ppai has quit IRC | 11:51 | |
| ho | good night! | 11:51 |
| *** ho has quit IRC | 11:52 | |
| *** km has quit IRC | 11:53 | |
| *** david-lyle_afk is now known as david-lyle | 11:58 | |
| *** ajay_ has quit IRC | 12:03 | |
| *** ppai has joined #openstack-swift | 12:04 | |
| *** lpabon has joined #openstack-swift | 12:04 | |
| *** fandi has quit IRC | 12:08 | |
| openstackgerrit | Prashanth Pai proposed openstack/swift: fsync() on directories https://review.openstack.org/126923 | 12:19 |
| *** Masahiro has quit IRC | 12:26 | |
| openstackgerrit | Merged openstack/swift: Only move too-close-together replicas when they can spread out. https://review.openstack.org/140879 | 12:31 |
| *** tdasilva has quit IRC | 12:31 | |
| *** dmsimard_away is now known as dmsimard | 12:33 | |
| *** miqui has quit IRC | 12:36 | |
| *** nosnos has quit IRC | 12:43 | |
| *** nshaikh has joined #openstack-swift | 12:46 | |
| *** fandi has joined #openstack-swift | 12:50 | |
| openstackgerrit | Merged openstack/swift: Avoid unnecessary unlink() on every successful PUT https://review.openstack.org/143943 | 12:57 |
| *** lpabon has quit IRC | 13:08 | |
| *** addnull has quit IRC | 13:08 | |
| *** david-lyle is now known as david-lyle_afk | 13:08 | |
| *** Masahiro has joined #openstack-swift | 13:27 | |
| *** ppai has quit IRC | 13:30 | |
| *** Masahiro has quit IRC | 13:32 | |
| *** SkyRocknRoll has joined #openstack-swift | 13:51 | |
| *** ppai has joined #openstack-swift | 13:55 | |
| *** NM has quit IRC | 13:56 | |
| *** tdasilva has joined #openstack-swift | 13:57 | |
| *** bill_az has joined #openstack-swift | 14:03 | |
| *** nshaikh has quit IRC | 14:06 | |
| openstackgerrit | Mahati proposed openstack/swift: Implement OPTIONS verb for storage nodes. https://review.openstack.org/140103 | 14:11 |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/swift: Updated from global requirements https://review.openstack.org/88736 | 14:14 |
| *** ppai has quit IRC | 14:19 | |
| *** SkyRocknRoll has quit IRC | 14:41 | |
| *** lpabon has joined #openstack-swift | 14:42 | |
| *** SkyRocknRoll has joined #openstack-swift | 14:55 | |
| *** jistr has quit IRC | 14:57 | |
| *** jistr has joined #openstack-swift | 14:59 | |
| *** annegentle has joined #openstack-swift | 15:08 | |
| *** Masahiro has joined #openstack-swift | 15:16 | |
| *** Masahiro has quit IRC | 15:21 | |
| *** NM has joined #openstack-swift | 15:22 | |
| *** dkingshott has joined #openstack-swift | 15:35 | |
| *** elambert has joined #openstack-swift | 15:48 | |
| *** zaitcev has joined #openstack-swift | 16:04 | |
| *** ChanServ sets mode: +v zaitcev | 16:04 | |
| *** cdelatte has joined #openstack-swift | 16:15 | |
| *** cdelatte has quit IRC | 16:16 | |
| *** cdelatte has joined #openstack-swift | 16:16 | |
| *** jistr has quit IRC | 16:18 | |
| *** nellysmitt has quit IRC | 16:21 | |
| *** reed has joined #openstack-swift | 16:25 | |
| *** jistr has joined #openstack-swift | 16:34 | |
| *** gyee has joined #openstack-swift | 16:40 | |
| *** atan8 has joined #openstack-swift | 16:49 | |
| *** annegentle has quit IRC | 16:55 | |
| *** SkyRocknRoll has quit IRC | 16:59 | |
| notmyname | good morning | 17:00 |
| mahatic | good morning! | 17:00 |
| notmyname | mahatic: how are you? | 17:01 |
| mahatic | notmyname, I'm good. Went through ringmd5 | 17:02 |
| mahatic | also, this one looks like it's ready for review https://review.openstack.org/#/c/140103/ | 17:03 |
| notmyname | mahatic: great | 17:04 |
| *** Masahiro has joined #openstack-swift | 17:05 | |
| peluse | good morning! | 17:05 |
| mahatic | it should have been ready yesterday except for my blatant mistake ;) | 17:05 |
| notmyname | peluse: will you be around in a couple of hours for the team meeting? | 17:06 |
| peluse | yup | 17:08 |
| *** atan8 has quit IRC | 17:09 | |
| notmyname | great, thanks. I'll try to give a summary of EC from this week, but extra color and detail will probably best come from you :-) | 17:09 |
| *** Masahiro has quit IRC | 17:10 | |
| notmyname | mahatic: I'm going through email. I'll look at your patch right after | 17:12 |
| peluse | notmyname, no problema | 17:12 |
| mahatic | notmyname, sure, no problem | 17:14 |
| * notmyname select all + mark as read | 17:14 | |
| notmyname | /notreally | 17:14 |
| *** atan8 has joined #openstack-swift | 17:15 | |
| *** jordanP has quit IRC | 17:24 | |
| *** rledisez has quit IRC | 17:25 | |
| notmyname | mahatic: I'm looking right now. I suspect there will be a few other things to add to this patch. I'll leave comments in gerrit | 17:25 |
| mahatic | notmyname, sure | 17:26 |
| *** jistr has quit IRC | 17:26 | |
| *** atan8 has quit IRC | 17:29 | |
| notmyname | mahatic: ok. just 2 things I think | 17:38 |
| notmyname | mahatic: first, and importantly, what you have right now looks really good! | 17:39 |
| notmyname | mahatic: if there weren't additional requirements, I'd +2 it as-is | 17:39 |
| mahatic | notmyname, cool | 17:39 |
| notmyname | however, there are 2 things I think need to be handled | 17:39 |
| notmyname | first the simpler one | 17:39 |
| mahatic | okay | 17:40 |
| notmyname | ...thinking... | 17:41 |
| mahatic | :D alright | 17:41 |
| notmyname | http://tools.ietf.org/html/rfc7231#section-4.3.7 <-- defines the OPTIONS verb | 17:41 |
| notmyname | the initial thing we've talked about (the recon checker) will make an OPTIONS request to * since it's looking for server capabilities instead of the capabilities at a particular resource on that server | 17:42 |
| mahatic | right | 17:43 |
| notmyname | my initial thought is that the OPTIONS should maybe only respond to OPTIONS *, but then I was thinking if the response would actually be any different for a specific path | 17:43 |
| notmyname | hmm..I guess an OPTIONS request to a specific path should probably have the right response code (eg 200 vs 404) | 17:44 |
| mahatic | sorry, was looking at the code, (reading what you wrote now) | 17:45 |
| *** EmilienM is now known as EmilienM|afk | 17:45 | |
| notmyname | ya, it should. see the OPTIONS support in the proxy. (however, it also looks like the proxy's OPTIONS support doesn't handle * properly. that should get fixed later) | 17:46 |
| mahatic | you are right about the response code part | 17:46 |
| notmyname | hmm | 17:47 |
| *** david-lyle_afk is now known as david-lyle | 17:47 | |
| notmyname | well, that's not actually what the proxy OPTIONS request does either | 17:47 |
| notmyname | it checks to see if it's a valid path (format) and checks auth. but it doesn't actually check the particular resource (eg 404) | 17:48 |
| openstackgerrit | Merged openstack/swift: Add undocumented options to keystoneauth sample config https://review.openstack.org/145266 | 17:48 |
| notmyname | and that happens pretty much just because those other checks come before the dispatch to the OPTIONS method | 17:48 |
| mahatic | hmm, i'm looking | 17:49 |
| notmyname | mahatic: ok, for now that's probably all ok. ie as-is the storage node responds to any path in the OPTIONS request | 17:50 |
| notmyname | I don't think there's the requirement to return eg 404 if the object isn't there. that's a HEAD request | 17:50 |
| notmyname | not OPTIONS | 17:50 |
| *** lcurtis has joined #openstack-swift | 17:51 | |
| notmyname | so "if valid path or *, return 200 OK with the right headers" seems reasonable. and that's what you have. although you didn't specifically code for *, it works | 17:51 |
| notmyname | ok, issue 1 taken care of :-) | 17:51 |
| mahatic | notmyname, as i see now, proxy OPTIONS does not handle much right? It has more to do with CORS? | 17:52 |
| notmyname | rigth | 17:52 |
| mahatic | hmm | 17:52 |
| notmyname | and there's _no_way_ we're doing CORS on the storage nodes | 17:52 |
| mahatic | :D nice ;) | 17:52 |
| notmyname | torgomatic may have other feelings on this matter, and we should probably all study the RFC definition for OPTIONS to clarify what do to if the resource isn't found on the server | 17:52 |
| notmyname | ok, for the second issue (and in the interest of time, since I have something in 7 minutes) | 17:53 |
| notmyname | there is one case you aren't handling | 17:53 |
| mahatic | sure, go ahead | 17:53 |
| notmyname | the allowed_methods should be updated based on the "replication_server" config settion. ie if it's a replication server, a call to non-REPLICATE will respond with HTTPMethodNotAllowed. that needs to be reflected in the Allow header on the OPTIONS response | 17:55 |
| notmyname | coding that up and testing it, I suspect, will be "fun"* | 17:55 |
| notmyname | * not really | 17:55 |
| notmyname | I think that's what's missing from this patch before it can land | 17:55 |
| mahatic | heh, and I am still comprehending what I should be doing ;) | 17:56 |
| notmyname | I'll add that to gerrit comments | 17:56 |
| mahatic | sure | 17:56 |
| mahatic | notmyname, I don't know this - so, a replication server will not respond to any headers? | 17:58 |
| notmyname | mahatic: ok, I'll be back online later to offer any clarity needed (or you can ask others here while I'm not around) | 17:58 |
| mahatic | okay | 17:58 |
| notmyname | mahatic: no, a replication server will only handle replication verbs. not normal client-request verbs (PUT/GET/etc) | 17:58 |
| notmyname | mahatic: see http://docs.openstack.org/developer/swift/replication_network.html#dedicated-replication-network | 17:59 |
| notmyname | ok, I gotta run. I'll be back for the team meeting in an hour | 17:59 |
| mahatic | notmyname, great, thanks for the link | 17:59 |
| mahatic | sure | 17:59 |
| *** geaaru has quit IRC | 17:59 | |
| lcurtis | hello all...hopefully a few more quick questions as i set up cluster | 18:07 |
| lcurtis | is it possible to have different number of replicas in different region | 18:07 |
| tdasilva | lcurtis: I think you should be able to do that with different storage policies | 18:08 |
| lcurtis | and if we do 3 replicas, for example, can the 3rd replica be in different region | 18:08 |
| *** acoles is now known as acoles_away | 18:08 | |
| tdasilva | lcurtis: check this out: https://swiftstack.com/blog/2012/09/16/globally-distributed-openstack-swift-cluster/ | 18:09 |
| *** jdaggett_ is now known as jdaggett | 18:10 | |
| lcurtis | yes of course...thanks tdasilva...had read this before but it has been a while | 18:11 |
| lcurtis | will have to revisit | 18:11 |
| tdasilva | lcurtis: I think that should help with the replica in different region question, then you just need to figure out how to deploy your storage policies for different number of replicas | 18:13 |
| lcurtis | tdasilva...is this latest doc from swiftstack that you know of? | 18:13 |
| tdasilva | lcurtis: humm...not sure on that one, it was the first that came up on google :-) | 18:15 |
| lcurtis | lol | 18:15 |
| lcurtis | right thats why i asked okay... | 18:15 |
| tdasilva | lcurtis: i remember reading it before, not sure if they have a more recent one | 18:15 |
| lcurtis | lol...yes me too | 18:15 |
| lcurtis | trying to be the sme as I go... does not always work | 18:16 |
| lcurtis | thanks very much for the link tho...jogged my memory | 18:16 |
| tdasilva | welcome :) | 18:17 |
| lcurtis | perhaps you know....is it possible to change number of replicas without affecting data | 18:18 |
| tdasilva | lcurtis: not sure I understand your question | 18:19 |
| tdasilva | lcurtis: do you mean you have data that is triple replicated at the moment and you want to reduce to 2x? | 18:20 |
| lcurtis | so is it a very impactful operation to increase replicas on the fly with a larger cluster | 18:20 |
| lcurtis | or vice versa | 18:20 |
| lcurtis | so if we start with 2 and want to go to 3 | 18:20 |
| tdasilva | hum...never actually tried that to be honest, so you would get a much better answer from the experts | 18:22 |
| lcurtis | yep...scouring documentation today | 18:23 |
| lcurtis | good to know before we start ingesting tons of data i spose | 18:23 |
| tdasilva | my guess is that technically it is possible, but it might be impact your cluster pretty heavily for a while... | 18:23 |
| lcurtis | yes indeed | 18:27 |
| tdasilva | lcurtis: just found this: http://docs.openstack.org/admin-guide-cloud/content/section_fractional-replicas.html plus if you are able to set up a dedicated replication network, that could help with lessening the impact on your cluster | 18:29 |
| *** pberis has quit IRC | 18:29 | |
| lcurtis | yes just found that a few minutes ago | 18:30 |
| lcurtis | you rock | 18:30 |
| lcurtis | thanks fo much for that | 18:30 |
| *** pberis has joined #openstack-swift | 18:30 | |
| lcurtis | i feel a whole lot better for choosing swift for the community alone | 18:30 |
| *** silor has quit IRC | 18:31 | |
| tdasilva | lcurtis: make sure to test changing replica counts in a saio environment first...like I said before I never actually tried this myself...adding to my todo list now :-) | 18:34 |
| clayg | notmyname: my calendar still doesn't work - what time is the swift meeting | 18:34 |
| lcurtis | ;) yes indeed...guess we better set up a sandbox real quick | 18:35 |
| tdasilva | clayg: hi, meeting is in 25 min | 18:36 |
| *** bkopilov has quit IRC | 18:41 | |
| *** silor has joined #openstack-swift | 18:42 | |
| torgomatic | notmyname: i looked into OPTIONS *, but I couldn't get curl to make such a request, so limiting the path to * makes it annoying to test. | 18:43 |
| *** annegentle has joined #openstack-swift | 18:52 | |
| *** lihkin has joined #openstack-swift | 18:53 | |
| *** Masahiro has joined #openstack-swift | 18:53 | |
| *** kota_ has joined #openstack-swift | 18:54 | |
| *** acoles_away is now known as acoles | 18:56 | |
| *** bkopilov has joined #openstack-swift | 18:57 | |
| *** cutforth has joined #openstack-swift | 18:57 | |
| *** Masahiro has quit IRC | 18:58 | |
| notmyname | clayg: swift meeting is now | 18:59 |
| notmyname | torgomatic: really? curl http://foo/* doesn't work? | 18:59 |
| *** lihkin1 has joined #openstack-swift | 19:00 | |
| torgomatic | No, that is a path of /*, so the request line is OPTIONS /* HTTP/1.1 | 19:01 |
| torgomatic | Basically a path not starting with / doesn't fit an URL, so it's really obnoxious to generate such a request. | 19:01 |
| *** imkarrer has quit IRC | 19:02 | |
| notmyname | ah ok | 19:02 |
| *** lihkin has quit IRC | 19:02 | |
| zaitcev | ok how? I don't understand what Sam is saying. | 19:02 |
| zaitcev | Needs to drop the slash? | 19:03 |
| torgomatic | I couldn't even get requests (the library) to do it from a repl; that's when I gave up. | 19:03 |
| zaitcev | I see. telnet(1) is too old-school. | 19:03 |
| zaitcev | Okay I'm sorry already | 19:03 |
| zaitcev | jeez | 19:04 |
| zaitcev | My sinus hurts | 19:04 |
| zaitcev | And I called my daugher a "stupid little b*tch" because she did not want to listen to my explanations of proper braking procedures. | 19:05 |
| zaitcev | something is not right with my mind | 19:05 |
| notmyname | wow | 19:05 |
| *** silor has quit IRC | 19:14 | |
| *** annegentle has quit IRC | 19:20 | |
| *** annegentle has joined #openstack-swift | 19:21 | |
| *** itarchitectkev has joined #openstack-swift | 19:21 | |
| itarchitectkev | notmyname, so Swift all works great when keystone is running http, but moving to ssl https I get auth errors. I don't see any ssl errors but could do with finding out how I can debug. | 19:22 |
| notmyname | itarchitectkev: hmmm...interesting | 19:24 |
| notmyname | itarchitectkev: I'm in the swift team meeting right now, but let me know what you've done and what you're seeing | 19:25 |
| swifterdarrell | itarchitectkev: first Q that comes to mind: is everything using "real" or self-signed certs? are you having to do --insecure type stuff? | 19:25 |
| itarchitectkev | notmyname, no problem. I've only just started to tackle this today so not in any hurry. | 19:25 |
| itarchitectkev | swifterdarrell, self signed certs. is there an --insecure option to some configs in swift? | 19:26 |
| itarchitectkev | context: keystone https with self certs work fine with nova, glance, cinder, keystone. | 19:26 |
| swifterdarrell | itarchitectkev: not sure, but my first bet on any problem that shows up with https is cert trust ;) | 19:27 |
| swifterdarrell | itarchitectkev: anyway, it's something to dig into | 19:27 |
| notmyname | itarchitectkev: ya, in the swift CLI you have to tell it that self-signed certs are ok | 19:27 |
| itarchitectkev | yeah, when I was troubleshooting nova stuff and having to use --insecure, I saw SSL errors in the debug. I don't see them with swift --debug | 19:27 |
| itarchitectkev | notmyname, ah balls. missed that option. still not working, but that implies I've something else wrong. | 19:28 |
| *** atan8 has joined #openstack-swift | 19:29 | |
| zaitcev | that option is only for client. but authtoken middleware must have something similar | 19:29 |
| itarchitectkev | zaitcev, yeah - in the nova/glance/etc they have an option to do an insecure check. can't see that in swift. | 19:31 |
| zaitcev | https://github.com/openstack/keystonemiddleware/blob/master/keystonemiddleware/auth_token.py#L262 | 19:31 |
| cschwede | itarchitectkev: https://github.com/openstack/python-keystoneclient/blob/master/keystoneclient/middleware/auth_token.py#L272 | 19:32 |
| cschwede | oh, zaitcev was faster. never mind :) | 19:32 |
| notmyname | hmm...I think I just saw a patch this morning to add docs for keystone options that are currently not documented. maybe that's one? | 19:32 |
| itarchitectkev | but I'm not seeing ssl errors in the debug. | 19:33 |
| itarchitectkev | cschwede, zaitcev nice. I'll give that a go. | 19:33 |
| zaitcev | our etc/proxy-server.conf.example hasn't got right lines for keystonemiddleware, we need to refresh | 19:33 |
| zaitcev | can't send people to source all the time like now | 19:34 |
| acoles | notmyname: that was keystoneauth, this is authtoken | 19:34 |
| *** mahatic has quit IRC | 19:34 | |
| notmyname | oh. ok | 19:35 |
| acoles | would be neat if we could crossreference the keystonemiddleware config samples rather than maintaining our own | 19:37 |
| notmyname | acoles: "just" port everything to oslo.config. done! | 19:37 |
| *** dank has joined #openstack-swift | 19:37 | |
| notmyname | (don't really do that) | 19:38 |
| *** mahatic has joined #openstack-swift | 19:39 | |
| *** dkingshott has quit IRC | 19:41 | |
| glange | haha | 19:43 |
| acoles | notmyname: you gave me like 11 seconds to make a start before retracting your permission :D | 19:45 |
| *** kota_ has quit IRC | 19:45 | |
| notmyname | :-) | 19:46 |
| notmyname | itarchitectkev: meeting done. do you have your next steps? or do you still have outstanding questions? | 19:46 |
| mattoliverau | cschwede: i wonder if we could have a delayed delete option which just sets the X-Delete-After flag to a OP set amount and told the user in the Delete response | 19:46 |
| *** kota_ has joined #openstack-swift | 19:48 | |
| cschwede | mattoliverau: that was my first idea. unfortunately this requires either fast-post, at least a metadata update. i was looking for a „cheap“ option (in terms of replication, traffic, etc) | 19:48 |
| * torgomatic predicts that delayed-delete will end up with a bad case of scope creep | 19:48 | |
| mattoliverau | cschwede: acoles just nees to get fast-post working :P | 19:48 |
| zaitcev | maybe John can rule delayed-delete with iron fist | 19:49 |
| zaitcev | The job of PTL is to say "NO" | 19:49 |
| notmyname | NOOOOOOOOOOOOoooooo........ | 19:49 |
| acoles | mattoliverau: i'm trying man :S | 19:49 |
| acoles | mattoliverau: will you make hackathon? | 19:50 |
| mattoliverau | acoles: yup, I have approval :) | 19:50 |
| acoles | mattoliverau: cool | 19:50 |
| torgomatic | first it's ask an administrator to recover some deleted objects | 19:51 |
| torgomatic | then it's self-service recovery of deleted objects | 19:51 |
| torgomatic | and then, of course, users will want to list their deleted objects | 19:51 |
| acoles | torgomatic: then its automatic? | 19:51 |
| mattoliverau | torgomatic: lol, so photorec meddleware :P | 19:51 |
| notmyname | it's not automatic. it's torgomatic! (new marketing slogan) | 19:52 |
| torgomatic | and that ignores the question of who pays for storage of deleted objects | 19:52 |
| mattoliverau | lol | 19:52 |
| torgomatic | not to mention configurable retention periods based on a hundred million different characteristics | 19:52 |
| *** atan8 has quit IRC | 19:52 | |
| acoles | notmyname: time for a new channel topic? :) | 19:52 |
| mattoliverau | +1 :P | 19:53 |
| mattoliverau | if someone out there wants to make t-shirts for the hackathon: "it's not automatic. it's torgomatic!" | 19:54 |
| *** ChanServ changes topic to "It's not automatic, it's torgomatic!" | 19:54 | |
| * tdasilva thinks notmyname has new idea for swift shirt | 19:54 | |
| itarchitectkev | notmyname, sorry (kids bedtime here) - yes. I've got stuff to go on. thanks. (thanks zaitcev swifterdarrell). | 19:54 |
| mattoliverau | tdasilva: great minds :) | 19:55 |
| *** kota_ has quit IRC | 19:55 | |
| notmyname | itarchitectkev: great! | 19:55 |
| tdasilva | lol | 19:55 |
| torgomatic | and then some bright spark is going to say "hey, why not store all deleted objects using erasure-coding so it's cheap to keep them around?" | 19:55 |
| mattoliverau | lol, and the swift buzz word cycle is complete :P | 19:56 |
| * cschwede takes notes from torgomatic ideas | 19:57 | |
| itarchitectkev | notmyname zaitcev swifterdarrell -- adding insecure worked. I don't know why I didn't try that originally considering I used that to get over the nova/glance/etc stuff before pointing the in the direction of the certs. *shakes head* | 19:58 |
| notmyname | itarchitectkev: great! | 19:58 |
| mattoliverau | I might go find some breakfast, bbs | 19:58 |
| acoles | cschwede: i think we have seen cases where an undelete 'button' would have been very welcome | 19:59 |
| torgomatic | cschwede: just in case you thought it'd be relatively easy ;) | 19:59 |
| *** acoles is now known as acoles_away | 20:03 | |
| cschwede | acoles_away: yeah, me too. the question is if this feature is accessible by an user or only by an operator | 20:04 |
| cschwede | torgomatic: hehe, i never thought this will be easy :) so, lot of options, lot of stuff to think about. thanks for your input! | 20:07 |
| tdasilva | cschwede: what do you think of obj. versioning handling deletes? | 20:07 |
| torgomatic | and finally, it'll gain the ability to send an email when a super-important deleted file has only 72 hours of life before expunging... that's how you know the feature is done: it includes sending email ;) | 20:07 |
| cschwede | tdasilva: might be an option too. depends if this is consumable by an end-user? or if an user can enable/disable this (like the current versioning)? in this case there is no protection for credential abuse | 20:09 |
| cschwede | tdasilva: my first idea had operators without any end-users in mind; for example IT departments that use Swift as distributed storage for their applications | 20:10 |
| notmyname | cschwede: new plan. just disable the DELETE verb ;-) | 20:10 |
| cschwede | torgomatic: oh, easy. we simply integrate an email server into Swift. And to make it more comfortable, also an receiving mail server, so you can send a request by mail ;) | 20:11 |
| notmyname | cschwede: and of course swift will support the full smtp protocol so that it can be a central MTA to store all your mail | 20:11 |
| torgomatic | now we're talking enterprise quality! | 20:11 |
| cschwede | notmyname: that was also on my list (disable DELETE), but after some discussions i discarded that | 20:12 |
| tdasilva | cschwede: I guess I need to read up more on ACL to understand what a user could be allowed not allowed to do... | 20:13 |
| cschwede | tdasilva: so, my main idea was that it should be impossible to deactive using the API. Only enable/disable on the storage servers | 20:14 |
| tdasilva | cschwede: i see | 20:14 |
| *** fandi has quit IRC | 20:17 | |
| *** EmilienM|afk is now known as EmilienM | 20:25 | |
| *** cutforth has quit IRC | 20:34 | |
| *** pberis has quit IRC | 20:34 | |
| *** pberis has joined #openstack-swift | 20:34 | |
| *** fandi has joined #openstack-swift | 20:36 | |
| *** NM has quit IRC | 20:38 | |
| *** annegentle has quit IRC | 20:39 | |
| *** annegentle has joined #openstack-swift | 20:42 | |
| *** Masahiro has joined #openstack-swift | 20:42 | |
| *** Masahiro has quit IRC | 20:47 | |
| *** tdasilva has quit IRC | 21:01 | |
| *** itarchitectkev has quit IRC | 21:01 | |
| *** openstackgerrit has quit IRC | 21:05 | |
| *** openstackgerrit has joined #openstack-swift | 21:06 | |
| *** elambert has quit IRC | 21:11 | |
| *** annegentle has quit IRC | 21:14 | |
| *** NM has joined #openstack-swift | 21:27 | |
| *** elambert has joined #openstack-swift | 21:32 | |
| *** NM has quit IRC | 21:37 | |
| *** NM has joined #openstack-swift | 21:38 | |
| *** annegentle has joined #openstack-swift | 21:40 | |
| *** NM has quit IRC | 21:41 | |
| *** elambert has quit IRC | 21:46 | |
| *** annegentle has quit IRC | 21:48 | |
| *** annegentle has joined #openstack-swift | 21:49 | |
| notmyname | CFP for the vancouver summit just opened. deadline is Feb 9 | 21:52 |
| notmyname | https://www.openstack.org/summit/vancouver-2015/call-for-speakers | 21:52 |
| *** chlong has joined #openstack-swift | 21:56 | |
| *** chlong has quit IRC | 22:03 | |
| *** annegentle has quit IRC | 22:10 | |
| openstackgerrit | Samuel Merritt proposed openstack/swift: Add notion of overload to swift-ring-builder https://review.openstack.org/141452 | 22:20 |
| openstackgerrit | Samuel Merritt proposed openstack/swift: Add dispersion command to swift-ring-builder https://review.openstack.org/144432 | 22:20 |
| *** Masahiro has joined #openstack-swift | 22:31 | |
| *** Masahiro has quit IRC | 22:36 | |
| *** lpabon has quit IRC | 22:38 | |
| lcurtis | another nagging question...if we have multiple proxy servers, how are they kept consistent in a high load /volume environment? or is there a quorum leader per se? | 22:43 |
| notmyname | lcurtis: what needs to be consistent between them? | 22:43 |
| lcurtis | well...where to write data | 22:44 |
| lcurtis | or where an object is stored | 22:44 |
| notmyname | that's determined by the ring file + math. it's deterministic, so there's no need for the proxies to share state | 22:44 |
| notmyname | that being said, it's fair to ask what happens when you have 2 proxy servers with different versions of the ring | 22:45 |
| *** annegentle has joined #openstack-swift | 22:45 | |
| *** elambert has joined #openstack-swift | 22:46 | |
| *** annegent_ has joined #openstack-swift | 22:46 | |
| notmyname | in which case, assuming you aren't using radically different rings (ie you're using the included tools and pushing ring files with every rebalance), the ring doesn't change too much from one rebalance to another (it's proportional to the amount of capacity changed) | 22:46 |
| notmyname | and therefore the locations are largely stable | 22:46 |
| notmyname | lcurtis: https://swiftstack.com/blog/2013/02/25/data-placement-in-swift/ and https://swiftstack.com/blog/2012/11/21/how-the-ring-works-in-openstack-swift/ | 22:47 |
| lcurtis | yes had started reading | 22:47 |
| lcurtis | notmyname do you work for swiftstack? | 22:47 |
| notmyname | lcurtis: yes, I do | 22:48 |
| lcurtis | you are the swift-meistro | 22:48 |
| lcurtis | thanks so much for your guidance | 22:48 |
| notmyname | no more than many others here, but I have been around a while :-) | 22:49 |
| notmyname | clayg: peluse: you'll probably enjoy this talk https://www.usenix.org/conference/osdi14/technical-sessions/presentation/muralidhar <-- facebook's warm storage. cross-DC erasure codes (RS14,4+XOR) | 22:50 |
| *** annegentle has quit IRC | 22:50 | |
| *** dmsimard is now known as dmsimard_away | 22:50 | |
| clayg | notmyname: yeah thanks | 22:52 |
| notmyname | also I thought that it's cool that the recording of that talk is hosted in Swift (RAX cloud files) ;-) | 22:53 |
| *** annegent_ has quit IRC | 22:54 | |
| tacticus | notmyname: nice talk. | 22:56 |
| *** chlong has joined #openstack-swift | 23:22 | |
| *** elambert has quit IRC | 23:32 | |
| *** elambert has joined #openstack-swift | 23:32 | |
| *** lihkin1 has quit IRC | 23:33 | |
| *** elambert has quit IRC | 23:37 | |
| *** kei_yama has joined #openstack-swift | 23:39 | |
| *** lcurtis has quit IRC | 23:48 | |
| *** km has joined #openstack-swift | 23:51 | |
| *** ho has joined #openstack-swift | 23:54 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!