| *** byeager has quit IRC | 00:03 | |
| *** byeager has joined #openstack-swift | 00:03 | |
| *** byeager has quit IRC | 00:08 | |
| *** judd7 has quit IRC | 00:09 | |
| *** judd7 has joined #openstack-swift | 00:12 | |
| *** Midnightmyth has quit IRC | 00:22 | |
| *** matsuhashi has joined #openstack-swift | 00:30 | |
| *** NM has joined #openstack-swift | 01:17 | |
| *** hurricanerix has quit IRC | 01:25 | |
| *** zaitcev has quit IRC | 01:30 | |
| openstackgerrit | Jon Snitow proposed a change to openstack/swift: Privileged acct ACL header, new ACL syntax, TempAuth impl. https://review.openstack.org/63227 | 01:33 |
|---|---|---|
| *** nosnos has joined #openstack-swift | 01:37 | |
| *** csd has quit IRC | 01:43 | |
| *** shri has quit IRC | 01:43 | |
| *** NM has left #openstack-swift | 01:59 | |
| *** zackf has quit IRC | 02:28 | |
| *** chandankumar_ has joined #openstack-swift | 02:36 | |
| *** lpabon has quit IRC | 02:42 | |
| *** zackf has joined #openstack-swift | 02:43 | |
| *** chandankumar_ has quit IRC | 02:48 | |
| *** hugokuo has quit IRC | 02:51 | |
| *** minnear has quit IRC | 02:52 | |
| *** hugokuo has joined #openstack-swift | 02:52 | |
| *** minnear has joined #openstack-swift | 02:52 | |
| *** byeager has joined #openstack-swift | 03:03 | |
| *** chandankumar_ has joined #openstack-swift | 03:09 | |
| *** chandankumar_ has quit IRC | 03:16 | |
| *** matsuhashi has quit IRC | 03:30 | |
| *** matsuhashi has joined #openstack-swift | 03:31 | |
| *** zackf has quit IRC | 03:34 | |
| *** matsuhashi has quit IRC | 03:35 | |
| *** dencaval has quit IRC | 03:38 | |
| *** dencaval has joined #openstack-swift | 03:38 | |
| *** peluse has quit IRC | 03:40 | |
| *** peluse has joined #openstack-swift | 03:40 | |
| *** gyee has quit IRC | 03:44 | |
| *** byeager has quit IRC | 03:57 | |
| *** raies has joined #openstack-swift | 04:46 | |
| raies | notmyname: can you please suggest best material on cross origin resource sharing ?? | 04:47 |
| *** matsuhashi has joined #openstack-swift | 04:55 | |
| *** zackf has joined #openstack-swift | 05:00 | |
| *** zackf has quit IRC | 05:08 | |
| *** jasondotstar has quit IRC | 05:10 | |
| *** ppai has joined #openstack-swift | 05:10 | |
| *** nshaikh has joined #openstack-swift | 05:11 | |
| *** matsuhashi has quit IRC | 05:34 | |
| *** matsuhashi has joined #openstack-swift | 05:35 | |
| *** krast has quit IRC | 05:35 | |
| *** nosnos has quit IRC | 05:38 | |
| *** nosnos has joined #openstack-swift | 05:41 | |
| *** zackf has joined #openstack-swift | 05:45 | |
| *** SkyRocknRoll has joined #openstack-swift | 05:45 | |
| *** Anju has joined #openstack-swift | 06:14 | |
| Anju | hii all, which version of python swift needs | 06:15 |
| Anju | ? | 06:15 |
| hugokuo | Anju: 2.6~2.7 | 06:17 |
| Anju | hugokuo: can I use 2.7.1? | 06:18 |
| hugokuo | Anju: As I know.. Yes. We are running 2.7.3 on our box | 06:18 |
| Anju | hugokuo: can you tell me | 06:18 |
| Anju | where is is mentioned ? | 06:18 |
| Anju | hugokuo: is 2.7 backward compatible to 2.4 | 06:20 |
| hugokuo | Anju: I don't really sure about if there's any conflict with 2.4 sry | 06:21 |
| *** zackf has quit IRC | 06:21 | |
| Anju | okk hugokuo : but please tell me where the python requirement is mentioned | 06:21 |
| hugokuo | Anju: As a note in developer guide. It requires 2.6+ http://docs.openstack.org/developer/swift/getting_started.html?highlight=python#system-requirements | 06:22 |
| *** matsuhashi has quit IRC | 06:23 | |
| Anju | hugokuo: thanks a lot | 06:23 |
| *** matsuhas_ has joined #openstack-swift | 06:25 | |
| *** sourabh__ has quit IRC | 06:28 | |
| *** early has quit IRC | 06:39 | |
| *** early has joined #openstack-swift | 06:47 | |
| raies | Hi.. can someone tell me about CORS practical implementation ?? | 06:52 |
| *** psharma has joined #openstack-swift | 07:09 | |
| raies | notmyname: it is possible to give a fake Origin header in the request during Cross origin resource sharing, then how it is effecient to handle these requests. | 07:19 |
| *** madhuri has joined #openstack-swift | 07:22 | |
| madhuri | Hi, can anyone tell me what does fast POST mean? | 07:25 |
| *** mjseger has quit IRC | 07:28 | |
| *** SkyRocknRoll_ has joined #openstack-swift | 07:43 | |
| SkyRocknRoll | raies, here you go https://developer.mozilla.org/en/docs/HTTP/Access_control_CORS | 07:44 |
| SkyRocknRoll | raies, http://www.html5rocks.com/en/tutorials/cors/ | 07:44 |
| *** SkyRocknRoll has quit IRC | 07:44 | |
| *** SkyRocknRoll_ is now known as SkyRocknRoll | 07:44 | |
| *** SkyRocknRoll has quit IRC | 07:45 | |
| *** SkyRocknRoll has joined #openstack-swift | 07:48 | |
| SkyRocknRoll | Rei | 07:49 |
| *** Midnightmyth has joined #openstack-swift | 07:52 | |
| *** markd has quit IRC | 07:52 | |
| *** chandankumar_ has joined #openstack-swift | 07:59 | |
| *** chandankumar_ has quit IRC | 08:03 | |
| *** chandankumar_ has joined #openstack-swift | 08:03 | |
| *** xga has joined #openstack-swift | 08:10 | |
| *** thomaschaaf has joined #openstack-swift | 08:10 | |
| *** Midnightmyth has quit IRC | 08:12 | |
| *** chandankumar_ has quit IRC | 08:15 | |
| *** thomaschaaf has quit IRC | 08:16 | |
| *** dencaval has quit IRC | 08:19 | |
| *** dencaval has joined #openstack-swift | 08:20 | |
| *** thomaschaaf has joined #openstack-swift | 08:20 | |
| *** chandankumar_ has joined #openstack-swift | 08:40 | |
| *** chandankumar_ has quit IRC | 08:46 | |
| *** nacim has joined #openstack-swift | 08:47 | |
| *** thomaschaaf has quit IRC | 08:53 | |
| openstackgerrit | Tristan Cacqueray proposed a change to openstack/python-swiftclient: Add SSL certificate verification by default https://review.openstack.org/69187 | 09:00 |
| tristanC | Hello folks, this last changeset should be good for review | 09:01 |
| tristanC | And now I wonder what kind of ssl tests should be added, as most of the work is done by the requests module | 09:03 |
| *** matsuhas_ has quit IRC | 09:04 | |
| *** matsuhashi has joined #openstack-swift | 09:04 | |
| *** xga_ has joined #openstack-swift | 09:07 | |
| *** xga has quit IRC | 09:07 | |
| *** mlipchuk has joined #openstack-swift | 09:10 | |
| *** matsuhashi has quit IRC | 09:17 | |
| *** matsuhashi has joined #openstack-swift | 09:19 | |
| *** matsuhashi has quit IRC | 09:51 | |
| *** nosnos has quit IRC | 09:51 | |
| *** mehmetali has joined #openstack-swift | 09:56 | |
| *** xga_ has quit IRC | 09:56 | |
| *** xga has joined #openstack-swift | 09:57 | |
| *** _bluev has joined #openstack-swift | 10:00 | |
| *** xga_ has joined #openstack-swift | 10:04 | |
| *** xga has quit IRC | 10:07 | |
| acoles | dfg: sorry, just seen your message. | 10:33 |
| *** gvernik has joined #openstack-swift | 10:58 | |
| openstackgerrit | Balazs KOSSOVICS proposed a change to openstack/swift: 503 instead of 404 if account server fails https://review.openstack.org/69576 | 11:00 |
| *** xga_ has quit IRC | 11:08 | |
| *** zul has quit IRC | 11:20 | |
| *** zul has joined #openstack-swift | 11:24 | |
| *** gvernik has quit IRC | 11:34 | |
| *** tristanC has quit IRC | 11:36 | |
| *** tristanC has joined #openstack-swift | 11:36 | |
| *** marcusvrn has joined #openstack-swift | 11:41 | |
| erlon | marcusvrn: ping | 11:53 |
| erlon | marcusvrn: ;) | 11:53 |
| marcusvrn | erlon: nice! :) | 11:54 |
| marcusvrn | portante: hey all, I'm trying to adapt the existing in-memory diskfile interface example to Swift, but it still saves containers and objects on sdb1 and not in memory, could someone help me to understand what is happening? | 11:57 |
| openstackgerrit | A change was merged to openstack/swift: Merge branch 'milestone-proposed' (1.12.0) into merge_branch https://review.openstack.org/69666 | 11:58 |
| koolhead17 | Can someone take this question https://ask.openstack.org/en/question/11145/swift-data-migration-from-essex-to-havana/ | 12:04 |
| *** ppai has quit IRC | 12:04 | |
| *** Trixboxer has joined #openstack-swift | 12:16 | |
| *** haomaiwang has quit IRC | 12:18 | |
| *** haomaiwang has joined #openstack-swift | 12:18 | |
| *** gvernik has joined #openstack-swift | 12:18 | |
| *** ppai has joined #openstack-swift | 12:22 | |
| portante | marcusvrn: I'll be able to help later when I get into the office, around 9:30 AM EST | 12:22 |
| marcusvrn | portante: ok! thks!! | 12:25 |
| *** mehmetali has quit IRC | 12:47 | |
| *** AbyssOne__ is now known as a1|away | 12:54 | |
| *** chandankumar_ has joined #openstack-swift | 13:01 | |
| *** chandankumar_ has quit IRC | 13:05 | |
| *** chandankumar_ has joined #openstack-swift | 13:06 | |
| *** gvernik has quit IRC | 13:10 | |
| *** chandankumar_ has quit IRC | 13:30 | |
| *** dencaval has quit IRC | 13:33 | |
| *** ppai has quit IRC | 13:34 | |
| *** dencaval has joined #openstack-swift | 13:34 | |
| *** ppai has joined #openstack-swift | 13:51 | |
| *** chandankumar_ has joined #openstack-swift | 13:51 | |
| *** chandankumar_ has quit IRC | 13:51 | |
| *** haomaiwang has quit IRC | 14:10 | |
| *** ppai has quit IRC | 14:10 | |
| *** haomaiwang has joined #openstack-swift | 14:16 | |
| *** gvernik has joined #openstack-swift | 14:18 | |
| *** IRCInfoCustomer has joined #openstack-swift | 14:22 | |
| *** IRCInfoCustomer has left #openstack-swift | 14:22 | |
| *** gvernik has quit IRC | 14:28 | |
| *** nacim has quit IRC | 14:36 | |
| *** tongli has joined #openstack-swift | 14:37 | |
| *** byeager has joined #openstack-swift | 14:37 | |
| *** haomaiwang has quit IRC | 14:42 | |
| *** haomaiwang has joined #openstack-swift | 14:43 | |
| *** chandankumar_ has joined #openstack-swift | 14:48 | |
| *** nacim has joined #openstack-swift | 14:49 | |
| *** Midnightmyth has joined #openstack-swift | 14:51 | |
| *** chandankumar_ has quit IRC | 14:51 | |
| *** chandankumar_ has joined #openstack-swift | 14:53 | |
| *** chandankumar_ has quit IRC | 14:57 | |
| *** chandankumar_ has joined #openstack-swift | 14:58 | |
| portante | marcusvrn: back | 14:58 |
| *** psharma has quit IRC | 14:58 | |
| *** Midnightmyth has quit IRC | 15:01 | |
| *** chandankumar_ has quit IRC | 15:02 | |
| *** chandankumar_ has joined #openstack-swift | 15:03 | |
| marcusvrn | portante: hey! | 15:07 |
| portante | good morning (or afternoon) | 15:07 |
| *** chandankumar_ has quit IRC | 15:07 | |
| marcusvrn | portante: good morning/afternoon (afternoon here) | 15:09 |
| portante | so how are things going? | 15:09 |
| portante | you mentioned having some issues earlier with the in-memory object server | 15:09 |
| marcusvrn | portante: yep....I intended to run swift with in-memory diskfile, but I don't know if I did sth wrong...because when I upload an object, it is saved on sdb1 | 15:11 |
| portante | what does your object server .conf file look like and what version of swift are you using? | 15:12 |
| marcusvrn | portante: I'm using master swift and my .conf is like object-server.conf-sample | 15:16 |
| portante | so can you post your "use = " line for the object-server section? | 15:16 |
| marcusvrn | portante: for sure! "use = egg:swift#object" | 15:18 |
| portante | change that to "use = egg:swift#mem_object" | 15:19 |
| portante | right now the mechanism for using a different object server is to change the use line to load the correct server code | 15:20 |
| marcusvrn | hmmm...nice! but changing it, I don't need to change obj/server.py, right? | 15:21 |
| portante | right | 15:21 |
| portante | it is using the obj/mem_server.py instead | 15:21 |
| marcusvrn | and if I create a new diskfile, which parameter i need to write in "use=" line? | 15:23 |
| marcusvrn | where do i need to define the parameter, in this case "mem_object"? | 15:28 |
| *** zul has quit IRC | 15:30 | |
| portante | so you should write your new DiskFile class, and then follow the mem_server.py file, cloning that to use your new diskfile | 15:30 |
| *** zul has joined #openstack-swift | 15:30 | |
| dfg | acoles: its alright- i want to add some account lvl ratelimiting. Basically to prevent a problem user from taking too much of the system's resources. I thought a good place to add the flag was using your sysmeta stuff. I think it would work but wanted your opinion on in. | 15:32 |
| dfg | acoles: in your patch you didn't set up any specific means of setting this data did you? I was thinking of just adding in the functionality and if anybody would need to use it they'd just set it with a direct client or something but I'm pretty sure that won't go over very well. idk | 15:33 |
| acoles | dfg: sounds like it could be a candidate for sysmeta | 15:33 |
| *** chandankumar_ has joined #openstack-swift | 15:34 | |
| *** hurricanerix has joined #openstack-swift | 15:34 | |
| acoles | dfg: when you say "means of settingthis data" do you mean a function to add a sysmeta header? | 15:35 |
| dfg | ya | 15:35 |
| *** hurricanerix has quit IRC | 15:35 | |
| acoles | dfg: i did put a convenience method in like add_sysmeta(key, value) but it got reviewed-away because nothing was calling it :) | 15:35 |
| dfg | :) | 15:36 |
| *** hurricanerix has joined #openstack-swift | 15:36 | |
| *** byeager has quit IRC | 15:36 | |
| *** byeager has joined #openstack-swift | 15:36 | |
| dfg | thats fine- i just didn't know what your thoughts were on how it'd be used. We could set it with a direct client, or there could be some weird backdoor middleware or something. i kinda like the direct client better | 15:37 |
| acoles | dfg: so currently you need to add to req.headers - there is a get_sysmeta_prefix function to avoid typing x-account-sysmeta- or whatever, in request_helpers.py | 15:37 |
| dfg | acoles: but you'd have to add those headers in some middleware beneath the gatekeeper right? | 15:37 |
| marcusvrn | portante: alright! i'll try to run in-memory object server, changing "use=" line and after that, i'll try to write my diskfile!! tks so much!!!!! | 15:37 |
| acoles | dfg: clients can't set sysmeta though - gatekeeper prevents that, deliberately | 15:37 |
| *** chandankumar_ has quit IRC | 15:38 | |
| dfg | acoles: ya- i'd use like swiftly -D- skip all the middleware. this would just be for sys admin use for example | 15:38 |
| portante | marcusvrn: welcome | 15:38 |
| dfg | or direct client | 15:38 |
| acoles | dfg: so yes, pattern would be client sends x-foo, middleware validates and copies to x-account-sysmeta-foo, which is then automaticaaly persisted (assuming its an account request) | 15:38 |
| dfg | ya- thats what i was thinking. hmmm | 15:39 |
| *** chandankumar_ has joined #openstack-swift | 15:39 | |
| acoles | dfg: oh, so if you have a backdoor for sysadmin client and send straight to proxy_Server then you could just send the sysmeta header. the gatekeeper is there to stop users ever tinkering | 15:40 |
| *** byeager has quit IRC | 15:41 | |
| *** IRTermite has left #openstack-swift | 15:41 | |
| *** chandankumar_ has quit IRC | 15:45 | |
| *** chandankumar_ has joined #openstack-swift | 15:46 | |
| *** mjseger has joined #openstack-swift | 15:47 | |
| dfg | acoles: ya- thats the idea. it would work. how would that sound as a general way of setting those flags. assuming people are going to start using them for different types of sysadmin type things. | 15:47 |
| *** chandankumar_ has quit IRC | 15:49 | |
| *** chandankumar_ has joined #openstack-swift | 15:50 | |
| acoles | dfg: i guess if its only ever sysadmin then its ok to subvert the gatekeeper. | 15:50 |
| acoles | dfg: maybe others will scream objections... | 15:51 |
| acoles | dfg: btw, here's the function that didn't make it, in case its of any use (was in request_helpers.py) : http://paste.openstack.org/show/62106/ | 15:52 |
| portante | we are using git not subversion | 15:52 |
| *** zackf has joined #openstack-swift | 15:52 | |
| *** chandankumar_ has quit IRC | 15:52 | |
| acoles | portante: :) | 15:53 |
| marcusvrn | portante: sry...just one more question....i'm trying to run servers ("startmain"), but it returns me an error: "KeyError: 'getpwnam(): name not found: swift'" | 15:55 |
| mjseger | awhile back I brought up the fact that 1K PUTs were slower then 2K PUTs and after others confirmed it was because of the nagel algorithm, I thought we confirmed that too and someone said they were making the change to set TCP_NODELAY client-side | 15:56 |
| mjseger | but when I clone the latest version of swiftclient and grep for TCP, I don't see that set anywhere. is this still on someone's todo list? | 15:56 |
| dfg | acoles: alright cool. i'll code something up and see what people say about it. i'm sure it'll go just great... :) | 15:59 |
| *** byeager has joined #openstack-swift | 16:01 | |
| marcusvrn | portante: i fixed it! Sry for that question! :) | 16:04 |
| *** lpabon has joined #openstack-swift | 16:06 | |
| *** nshaikh has left #openstack-swift | 16:07 | |
| portante | marcusvrn: no problem, just got back to my desk and saw your question | 16:08 |
| portante | mjseger: we set NDELAY server side, for sure, not sure about the client | 16:10 |
| mjseger | oooooh, I don't thnk that will solve the problem, will it? | 16:10 |
| mjseger | it's the client-side where nagel is causing the extra 0.04 sec delay, at least I think so | 16:11 |
| portante | I don't think so either | 16:11 |
| mjseger | portante: so doesn't that mean users will continue to see slower 1k puts? | 16:12 |
| mjseger | the thing that concerns me is people love to run benchmarks (me too ;)) using 1k, 10k etc object sizes and swift does NOT do well with 1k | 16:13 |
| mjseger | and I don't want to see if get a bum rap | 16:13 |
| Dieterbe | any tips on rsync max_connections settings? | 16:15 |
| Dieterbe | i see we still have it set to 2, and also been seeing "max connections reached" errors. | 16:15 |
| marcusvrn | portante: now it's working, but when i upload an object, in log file i got this:[29/Jan/2014:16:06:37 +0000] "PUT /sdb1/802/AUTH_test/teste1"[29/Jan/2014:16:06:37 +0000] "HEAD /sdb1/135/AUTH_test/teste1/test.txt" ...so it still saving object on sdb, right? | 16:15 |
| Dieterbe | since we have beefy machines, i wonder if i can crank this up to 20 or something? | 16:15 |
| portante | ah | 16:15 |
| portante | so that is because the proxy server does not know that it is only in-memory, and the rings you are using are setup to use disks from before | 16:16 |
| *** otherjon_ has joined #openstack-swift | 16:16 | |
| portante | mjseger: there seems to be two things to consider here | 16:17 |
| portante | 1. behavior of the python-swiftclient code in the face of small objects as it relates to the nagel algorithm | 16:18 |
| portante | 2. behavior of the swift server in the face of small objects | 16:18 |
| mjseger | understood | 16:18 |
| mjseger | so it sounds like someone has address the server-side but not the client side | 16:19 |
| portante | someone can address the swift client as well, but I would be careful to not lump them together | 16:19 |
| mjseger | I was pretty sure when we had this conversation notmyname; suggested a patch which i tried and it worked. then he (or someone else) said they would commit it. | 16:20 |
| portante | I would consider doing benchmarks against the swift server using a client on which you have total control (or near total control) of the messages being sent | 16:20 |
| portante | that notmyname! | 16:20 |
| portante | :) | 16:20 |
| *** haomaiwa_ has joined #openstack-swift | 16:22 | |
| *** zaitcev has joined #openstack-swift | 16:24 | |
| *** haomaiwang has quit IRC | 16:26 | |
| marcusvrn | portante: i'm looking for documentation that explain how ring works and how configure it, but actually i didn't find sth about that. | 16:32 |
| peluse | marcusvrn: check out this video: https://swiftstack.com/blog/2012/11/21/how-the-ring-works-in-openstack-swift/ | 16:36 |
| portante | marcusvrn: and http://docs.openstack.org/developer/swift/admin_guide.html | 16:36 |
| *** gyee has joined #openstack-swift | 16:46 | |
| *** byeager_ has joined #openstack-swift | 16:56 | |
| *** byeager has quit IRC | 16:56 | |
| *** acoles_ has joined #openstack-swift | 16:59 | |
| *** acoles_ has quit IRC | 16:59 | |
| *** byeager has joined #openstack-swift | 17:20 | |
| *** byeager_ has quit IRC | 17:23 | |
| *** mlipchuk has quit IRC | 17:29 | |
| marcusvrn | peluse: portante thks for the links | 17:32 |
| *** nacim has quit IRC | 17:34 | |
| portante | welcome | 17:36 |
| peluse | yup, you bet | 17:36 |
| *** nacim has joined #openstack-swift | 17:42 | |
| *** csd has joined #openstack-swift | 17:43 | |
| *** pheadron has joined #openstack-swift | 17:48 | |
| *** pheadron1 has quit IRC | 17:50 | |
| *** nacim has quit IRC | 17:52 | |
| *** Midnightmyth has joined #openstack-swift | 17:54 | |
| *** fbo is now known as fbo_away | 18:04 | |
| *** byeager has quit IRC | 18:12 | |
| *** byeager has joined #openstack-swift | 18:13 | |
| *** _bluev has quit IRC | 18:13 | |
| *** byeager has quit IRC | 18:17 | |
| *** shri has joined #openstack-swift | 18:49 | |
| *** nacim has joined #openstack-swift | 18:57 | |
| *** zackf1 has joined #openstack-swift | 19:09 | |
| *** krtaylor has quit IRC | 19:10 | |
| *** zackf has quit IRC | 19:12 | |
| *** zackf1 has quit IRC | 19:23 | |
| *** dencaval has quit IRC | 19:28 | |
| *** dencaval has joined #openstack-swift | 19:28 | |
| *** dencaval has quit IRC | 19:32 | |
| *** byeager has joined #openstack-swift | 19:33 | |
| *** gvernik has joined #openstack-swift | 19:33 | |
| *** dencaval has joined #openstack-swift | 19:33 | |
| *** zackf has joined #openstack-swift | 19:48 | |
| *** krtaylor has joined #openstack-swift | 19:56 | |
| *** nacim has quit IRC | 20:01 | |
| *** byeager has quit IRC | 20:02 | |
| *** byeager has joined #openstack-swift | 20:03 | |
| *** byeager has quit IRC | 20:05 | |
| *** byeager has joined #openstack-swift | 20:05 | |
| *** wayneeseguin has joined #openstack-swift | 20:10 | |
| *** dencaval has quit IRC | 20:18 | |
| *** byeager has quit IRC | 20:18 | |
| *** dencaval has joined #openstack-swift | 20:18 | |
| *** byeager has joined #openstack-swift | 20:19 | |
| *** byeager has quit IRC | 20:24 | |
| *** csd has quit IRC | 20:24 | |
| *** byeager has joined #openstack-swift | 20:33 | |
| *** byeager has quit IRC | 20:42 | |
| *** byeager has joined #openstack-swift | 20:42 | |
| *** byeager_ has joined #openstack-swift | 20:45 | |
| *** byeager has quit IRC | 20:45 | |
| *** pberis_ has quit IRC | 20:53 | |
| *** pberis has quit IRC | 20:53 | |
| *** byeager_ has quit IRC | 20:57 | |
| *** byeager has joined #openstack-swift | 20:57 | |
| *** byeager has quit IRC | 21:01 | |
| *** byeager has joined #openstack-swift | 21:02 | |
| *** shri has quit IRC | 21:08 | |
| *** shri has joined #openstack-swift | 21:11 | |
| wer | hrm. I was just trying to be nice and push a node out before I removed it.... and adjusted the weight from 3000 to 10 on each of it's drives. And now it is pulling tons of data. meanwhile I have a node with the same weight as all the nodes that is getting no data. | 21:26 |
| wer | the whole reason I was pushing this node out in the the first place is that I was migrating to one zone in that rack. So the node that is filling up in zone 1. And the other nodes are in zone 0. Before the change zone one's usage was higher. which made sense. But now it is going to fill up :/ | 21:32 |
| wer | so the answer is.... I must have done something stuupid. | 21:33 |
| torgomatic | wer: yeah, the partition placement algorithm prioritizes durability over even distribution, so if you've got two zones and three replicas, a minimum of 1/3 of your data will reside in the smaller of the two zones | 21:33 |
| torgomatic | if you want to take drives out, set their weights to 0, I think | 21:33 |
| torgomatic | that should make it so they have no partitions allocated, so the data should then move off them | 21:34 |
| wer | I'll try that in an hour :) I knew I should have set them to 0, but I wanted to see it work how I expected. Previous to pushing it out it didn't have 1/3 despite being the only node. But perhaps I didn't rebalance enough between all these steps to know. | 21:35 |
| torgomatic | ah yeah, you might have wound up in a situation where 1/3 of your partitions wanted to be in that zone, but due to min_part_hours, they couldn't move yet | 21:36 |
| wer | yup :) | 21:36 |
| torgomatic | so after $min_part_hours hours, you rebalanced with lower weights, and the allocated data went *up* | 21:36 |
| wer | And the choice to move to one zone was to avoid this scenario where that zone couldn't provide 1/3 of the space. I would rather leverage the host/disk smartness given the size of my cluster :) | 21:37 |
| wer | does a handoff occur when 1/3 is not available in a given zone? Or does it start doing node/disk distribution at that point? | 21:38 |
| torgomatic | wer: if you want to stop it now, swift-ring-builder has a (deliberately) undocumented command called "pretend_min_part_hours_passed" | 21:38 |
| torgomatic | wer: there's no handoffs involved; it's all about the placement of primary nodes | 21:39 |
| wer | so if that zone filled up what would I expect? | 21:39 |
| Dieterbe | does tox insist on python 2.6 ? i'm running it on python 2.7.6 and i get: ERROR: InterpreterNotFound: python2.6 | 21:40 |
| torgomatic | badness? things get pretty unhappy with full disks... you'd get failed object PUTs at a minimum, possibly lost container updates (can't write an async_pending if there's no space), and probably other things too | 21:40 |
| wer | torgomatic: nice! jut reinforces my decision :) | 21:41 |
| wer | tyty | 21:41 |
| torgomatic | np | 21:41 |
| *** csd has joined #openstack-swift | 21:49 | |
| wer | sweet torgomatic now the ring looks like I expected. The data stopped migrating into zone1 with all the weights set to 0.... letting it soak waiting for magic. | 21:50 |
| wer | my first lie to swift. | 21:51 |
| *** lpabon has quit IRC | 21:55 | |
| *** tongli has quit IRC | 22:00 | |
| wer | So I'll have a similar thing when I add a new region. I suppose I should just add everything at once. If I add a new zone, add it all at once. If I remove a zone, set it to 0 first before removing nodes. | 22:02 |
| wer | or scary. | 22:02 |
| wer | What's crazy is watching a machine with 24 3TB disk fill at %2 and hour..... btw the node set with a weight of 0 is not egressing at all. But I'll let it have some time before I poke. | 22:04 |
| wer | *an hour :) | 22:04 |
| clayg | notmyname: can we like assign this but to pbr -> https://bugs.launchpad.net/swift/+bug/1273985 | 22:06 |
| *** byeager has quit IRC | 22:10 | |
| *** byeager has joined #openstack-swift | 22:10 | |
| openstackgerrit | David Goetz proposed a change to openstack/swift: Add a way to ratelimit all writes to an account https://review.openstack.org/69995 | 22:11 |
| *** byeager has quit IRC | 22:15 | |
| clayg | like the paste error when trying to load some middleware that is totally installed has come up before right? | 22:17 |
| torgomatic | yeah, it's what's in your egg-info being stale | 22:20 |
| clayg | torgomatic: yeah i've seen it like that in dev, but that normally means I cloned new stuff and didn't setup.py develop again | 22:44 |
| clayg | you think the issue in that bug was installed from source with setup.py develop? | 22:45 |
| portante | torgomatic, clayg: i have also see that, where I did not properly setup.py develop the tree in my SAIO | 22:56 |
| * portante above is missing an "n" | 22:56 | |
| openstackgerrit | paul luse proposed a change to openstack/swift: Add Storage Policy Support for /Info https://review.openstack.org/69700 | 22:57 |
| *** Midnightmyth has quit IRC | 22:58 | |
| *** shri1 has joined #openstack-swift | 23:03 | |
| *** _bluev has joined #openstack-swift | 23:05 | |
| *** shri has quit IRC | 23:05 | |
| *** krtaylor has quit IRC | 23:22 | |
| torgomatic | clayg: I have no idea what that guy's problem is; it's probably either that or he's got a mix of packages and source installs | 23:23 |
| openstackgerrit | paul luse proposed a change to openstack/swift: Add Storage Policy Support for /Info https://review.openstack.org/69700 | 23:32 |
| torgomatic | I love how those get filed as bugs against Swift, though. It's not Swift's fault that someone doesn't know how Python works. | 23:37 |
| *** shri1 has left #openstack-swift | 23:42 | |
| *** csd has quit IRC | 23:59 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!