| shri | notmyname: I have some follow up questions about the regions. Let me send out an email with those | 00:01 |
|---|---|---|
| *** haomaiw__ has quit IRC | 00:02 | |
| *** kenhui has joined #openstack-swift | 00:02 | |
| *** annegent_ has quit IRC | 00:10 | |
| *** pconstantine_ has quit IRC | 00:15 | |
| *** kenhui has quit IRC | 00:17 | |
| *** matsuhashi has joined #openstack-swift | 00:21 | |
| *** kevinc___ has joined #openstack-swift | 00:23 | |
| *** mwstorer has quit IRC | 00:24 | |
| *** kevinc_ has quit IRC | 00:26 | |
| *** kevinc___ has quit IRC | 00:28 | |
| *** dmorita has joined #openstack-swift | 00:31 | |
| *** shri has quit IRC | 00:39 | |
| *** haomaiwa_ has joined #openstack-swift | 00:59 | |
| *** kota_ has joined #openstack-swift | 01:04 | |
| *** annegent_ has joined #openstack-swift | 01:05 | |
| *** annegent_ has quit IRC | 01:10 | |
| *** haomaiwa_ has quit IRC | 01:31 | |
| *** nsquare has quit IRC | 01:40 | |
| *** kenhui has joined #openstack-swift | 01:43 | |
| *** nosnos has joined #openstack-swift | 01:44 | |
| *** mattoliverau has quit IRC | 01:48 | |
| *** charz has joined #openstack-swift | 01:49 | |
| *** kota_ has quit IRC | 01:51 | |
| *** kota_ has joined #openstack-swift | 01:51 | |
| *** mattoliverau has joined #openstack-swift | 01:51 | |
| *** baojg has joined #openstack-swift | 01:52 | |
| *** kenhui has quit IRC | 01:52 | |
| *** kota_ has quit IRC | 02:01 | |
| *** zz_wasmum is now known as wasmum | 02:03 | |
| *** wasmum is now known as zz_wasmum | 02:10 | |
| *** baojg has quit IRC | 02:12 | |
| *** annegent_ has joined #openstack-swift | 02:15 | |
| *** kenhui has joined #openstack-swift | 02:18 | |
| *** zaitcev has quit IRC | 02:25 | |
| *** dmsimard is now known as dmsimard_away | 02:28 | |
| *** baojg has joined #openstack-swift | 02:32 | |
| *** charz has quit IRC | 02:33 | |
| *** charz has joined #openstack-swift | 02:34 | |
| *** dmsimard_away is now known as dmsimard | 02:35 | |
| *** charz has quit IRC | 02:35 | |
| *** dmsimard is now known as dmsimard_away | 02:35 | |
| *** charz has joined #openstack-swift | 02:38 | |
| *** kenhui has quit IRC | 02:40 | |
| *** charz has quit IRC | 02:45 | |
| *** baojg has quit IRC | 02:48 | |
| *** jyoti_ranjan has joined #openstack-swift | 02:50 | |
| *** nosnos has quit IRC | 02:54 | |
| *** nosnos has joined #openstack-swift | 02:54 | |
| *** CybergeekDK has quit IRC | 02:54 | |
| *** byeager_away has quit IRC | 02:57 | |
| *** nosnos has quit IRC | 02:59 | |
| *** jyoti_ranjan has quit IRC | 02:59 | |
| *** CybergeekDK has joined #openstack-swift | 03:00 | |
| *** byeager_away has joined #openstack-swift | 03:00 | |
| *** nosnos has joined #openstack-swift | 03:05 | |
| *** nsquare has joined #openstack-swift | 03:08 | |
| *** matsuhashi has quit IRC | 03:18 | |
| *** matsuhashi has joined #openstack-swift | 03:19 | |
| *** zhiyan_ is now known as zhiyan | 03:19 | |
| *** charz has joined #openstack-swift | 03:24 | |
| *** matsuhashi has quit IRC | 03:24 | |
| *** baojg has joined #openstack-swift | 03:27 | |
| *** kenhui has joined #openstack-swift | 03:29 | |
| *** kenhui has quit IRC | 03:34 | |
| *** nosnos has quit IRC | 03:34 | |
| *** nosnos has joined #openstack-swift | 03:35 | |
| *** nosnos has quit IRC | 03:39 | |
| *** annegent_ has quit IRC | 03:47 | |
| *** annegent_ has joined #openstack-swift | 03:50 | |
| *** annegent_ has quit IRC | 03:51 | |
| *** baojg has quit IRC | 03:59 | |
| *** jyoti_ranjan has joined #openstack-swift | 04:08 | |
| *** matsuhashi has joined #openstack-swift | 04:10 | |
| *** bkopilov has joined #openstack-swift | 04:12 | |
| *** nosnos has joined #openstack-swift | 04:19 | |
| *** nshaikh has joined #openstack-swift | 04:21 | |
| *** baojg has joined #openstack-swift | 04:24 | |
| *** ppai has joined #openstack-swift | 04:37 | |
| *** nshaikh has quit IRC | 04:38 | |
| *** psharma has joined #openstack-swift | 04:39 | |
| notmyname | reading the ML thread(s) about removing hacking rules, I'm very glad we've got an opt-in system to them | 04:40 |
| *** ChanServ changes topic to "Currently in Swift 2.0 RC phase -- read all about it at http://lists.openstack.org/pipermail/openstack-dev/2014-June/038357.html | master is open for dev work" | 04:40 | |
| *** gyee has quit IRC | 04:42 | |
| *** kopparam has joined #openstack-swift | 04:42 | |
| *** nshaikh has joined #openstack-swift | 04:46 | |
| portante | notmyname: yes, and it seems a good path to take to look at what projects opt-in and which opt-out to find out what the "real" concensus is, not just what folks say on MLs or in blue prints | 04:46 |
| portante | I am not sure I agree that strong code style enforcement materially helps reach the goal of allowing folks to easily move between projects | 04:47 |
| notmyname | portante: yup. every time stuff like this comes up, I find out over beers at a conference that there is much less consensus than a ML may imply | 04:47 |
| notmyname | portante: first, show me that there is a strong desire or current practice of a high degree of cross project contribution (and I suspect those that do are generally good devs anyway) | 04:48 |
| * portante should just shut up now ... not much good comes from keeping my mouth open on these topics | 04:49 | |
| notmyname | portante: especially at such an hour? ;-) | 04:49 |
| portante | :) | 04:50 |
| portante | 9:50 PM, not so late, is it? :) | 04:50 |
| portante | somewhere it is almost 9 AM, so can't be so bad, right? | 04:51 |
| notmyname | heh. I'm on kid duty for the first part of this week, so I'm just catching up on stuff now. I've got a presentation to give tomorrow at lunch (and the marketing VP said "sure would be nice to have a blog post tonight on it") | 04:52 |
| notmyname | hmm...I probably shouldn't use a $ icon for my "vendor extensions" slide should I? | 04:54 |
| portante | sure, what's in a word anyways? | 04:54 |
| *** kopparam_ has joined #openstack-swift | 04:57 | |
| *** kopparam has quit IRC | 04:58 | |
| *** kenhui has joined #openstack-swift | 05:00 | |
| *** nsquare has quit IRC | 05:01 | |
| *** Kbee has joined #openstack-swift | 05:02 | |
| openstackgerrit | A change was merged to openstack/swift: Change assertCalledWith to assert_called_with https://review.openstack.org/99124 | 05:04 |
| *** kenhui has quit IRC | 05:04 | |
| notmyname | now what's the icon I use for "you"? as in "you can create cool new functionality for your swift cluster with storage policies" | 05:06 |
| *** ajc_ has joined #openstack-swift | 05:07 | |
| notmyname | clearly http://4.bp.blogspot.com/_2kjisMm3M9Y/R72hXbFjHjI/AAAAAAAAC7s/iI8mGFZIahY/s1600-h/wallpaper_batman_animated_series2.png | 05:09 |
| mattoliverau | its only 3:15pm here :P | 05:15 |
| notmyname | mattoliverau: yes, but you live in australia, so I imagine your normal conversation starts with "how many times did an animal try to kill you today" | 05:21 |
| notmyname | mattoliverau: I wanted to say thanks for all the reviews you've been doing lately. great stuff! | 05:22 |
| mattoliverau | notmyname: lol, that is true, so when I say I'm off to lunch and I don't come back, start to panic :P | 05:22 |
| notmyname | hehe | 05:22 |
| *** pconstantine_ has joined #openstack-swift | 05:23 | |
| Kbee | notmyname: For a very large swift deployment( 4PB).. Is it good to have acocunt and container as combined with proxy nodes.. and object nodes seperate | 05:23 |
| Kbee | OR.. proxy seperate and account and container seperate | 05:24 |
| mattoliverau | notmyname: no problem at all, still learning the code base, but thanks to all the code reviews (and the swift text I got in Atlanta) I think I'm getting a good grasp on it. I've now got 2 bug patches in.. low priority ones that its a start :) | 05:24 |
| notmyname | Kbee: well it all depends, doesn't it? ;-) | 05:24 |
| Kbee | notmyname, :) true.. But is is it OK to combine proxy and account and container nodes.. any known issues or problems | 05:25 |
| goodes | good morning all | 05:25 |
| notmyname | Kbee: it depends on the hardware you have and the usage patterns for the data. let me ask you this: where are you currently seeing resource contention in the cluster? | 05:26 |
| mattoliverau | morning goodes | 05:26 |
| *** haomaiwa_ has joined #openstack-swift | 05:26 | |
| goodes | mattoliverau: g'day | 05:27 |
| mattoliverau | goodes: you sound like and aussie :) | 05:27 |
| mattoliverau | *an | 05:27 |
| Kbee | notmyname, I'm not aware beforehand of resource contention... but had these two options before me.. hence wanted to figure out one. | 05:28 |
| goodes | mattoliverau: Perth born and breed | 05:28 |
| Kbee | notmyname, to avoid any obvious pitfalls if any | 05:29 |
| notmyname | Kbee: sure, you can combine account and containers and proxies. generally I'd recommend keeping them separate in larger clusters, but that's mostly because you can better optimize hardware and it's less complexity for problems (ie one more SKU vs more possible resource contention from various processes on the box) | 05:29 |
| mattoliverau | goodes: aaah nice! Perth is a beautiful place :) | 05:29 |
| notmyname | goodes: I was in perth in january. beautiful. some of the best beaches I've ever seen | 05:29 |
| Kbee | notmyname, Thanks.. | 05:30 |
| notmyname | Kbee: is this a cluster you are designing or have running today? | 05:30 |
| goodes | not many in the world compare | 05:30 |
| Kbee | notmyname, we are planning to deploy | 05:30 |
| notmyname | Kbee: can I ask where and for what use case? I find it very useful to know who's using swift and for what (and it's terribly hard to track) | 05:31 |
| Kbee | notmyname, This is just for internal purpose for dev-testing. | 05:32 |
| mattoliverau | goodes: I was in bali a few years ago and was wondering why people loved it so much, then I realised that beaches in Australia are awesome and unlike most other places in the world.. makes you realise how lucky one is to be in Australia :) | 05:32 |
| notmyname | Kbee: as a general recommendation, I think you should use 3 SKUs in large clusters: 1 proxy, 1 account+container, and 1 object | 05:32 |
| Kbee | notmyname, thanks.. | 05:33 |
| notmyname | Kbee: think of it this way. if you have one SKU for proxy+account+container, that box has to have enough cpu and ram to do the packet shuffling for the cluster and also enough ram to keep dbs in page cache and enough IOPS to keep up with the listing updates. that's an expensive box | 05:33 |
| notmyname | Kbee: but if you have one proxy (with NICs and CPU and some RAM) and one for account+container (with SSDs and a weaker CPU and less RAM) then it's probably less complexity to manage as you keep it running | 05:34 |
| Kbee | notmyname, Totally Agree.. Wanted to confirm my understanding .. | 05:35 |
| notmyname | Kbee: 4PB for internal dev-testing? can you share what company? | 05:35 |
| Kbee | notmyname, Not eactly 4PB i extrapolated that.. we are having few terabytes and just wanted to extrapolate to that value | 05:36 |
| *** nsquare has joined #openstack-swift | 05:36 | |
| *** baojg has quit IRC | 05:40 | |
| goodes | mattoliverau: have been living overseas for a long while now and work keeps taking me all over the world and whereever I go keep getting reminded of paraphrasing that Croc Dundee line "That's not a beach/surf ...." | 05:41 |
| mattoliverau | goodes: lol, where are you now? | 05:42 |
| *** baojg has joined #openstack-swift | 05:42 | |
| *** nsquare has quit IRC | 05:48 | |
| goodes | now the we have policies have we considered adding to swiftclient a move object I.e. sever side copy + (optional confirm, HEAD) + delete - seems to me many people will be implementing this logic | 05:49 |
| notmyname | goodes: ya, and it will be tricky to get right. but doing it once in the official client is better than everybody doing it and half getting it wrong | 05:50 |
| goodes | especially if you need to confirm write perms for source and dest | 05:52 |
| goodes | but do it only once if you're going to batch | 05:52 |
| *** kopparam_ has quit IRC | 05:53 | |
| notmyname | I think the hard part is getting the retries and cleanup handled properly when it half-fails | 05:53 |
| goodes | and for slo/dlo you would want to confirm all the parts are copied before deleting the source objects | 05:54 |
| goodes | if you think that it's worthwhile, I may take a stab at it | 05:55 |
| *** kopparam has joined #openstack-swift | 05:56 | |
| notmyname | goodes: yes, seems good. it would need to be some SDK function and then wrap that in the CLI. ie do both parts, not just something in the CLI | 05:56 |
| *** bkopilov has quit IRC | 05:58 | |
| *** bkopilov has joined #openstack-swift | 05:59 | |
| *** haomaiwa_ has quit IRC | 06:00 | |
| goodes | notmyname: started looking at the new swiftservice - step in the right direction | 06:00 |
| goodes | notmyname: still trying to get my head around the whole 'process' - may need some help doing it right | 06:01 |
| notmyname | goodes: no problem. I'm here to help | 06:01 |
| goodes | notmyname: thx | 06:02 |
| ppai | mattoliverau, any luck on reseller_prefix ? | 06:06 |
| openstackgerrit | Clay Gerrard proposed a change to openstack/swift: Fix issues with test_wsgi.py and Storage Policies https://review.openstack.org/102052 | 06:13 |
| mattoliverau | Hey ppai, welcome back! Yeah I asked the guys in the know this moring and they said it's not something that is easy to do. As the reseller_prefix is prepended to the account name, means that the location hash for every account, container and object is effected. Which in turn would mean every thing in the cluster would need to be rehashed. So unfortuantly there is no easy way to do this. It's one of those parameters you | 06:14 |
| mattoliverau | want to get right before rolling out the cluster. | 06:14 |
| notmyname | mattoliverau: +1 | 06:15 |
| ppai | mattoliverau, Thanks! So what is the point of making it configurable option..Confused. | 06:15 |
| *** ppai_ has joined #openstack-swift | 06:16 | |
| mattoliverau | ppai: so can potentailly use multiple authentication middleware's, you need this to be configured so the middle whare know which belongs to which.. or if you simply just don't like the default 'auth' :) | 06:17 |
| mattoliverau | wow, my english is good today.. sorry about that | 06:17 |
| notmyname | the original point of the reseller_prefix is so that you can have multiple auth systems integrated into the same cluster at one time, and you can clearly distinguish between the sets of accounts each are responsible for | 06:18 |
| mattoliverau | ppai: what notmyname said ^^ :) | 06:19 |
| notmyname | eg you can have a public auth system and also an internal one for corporate or test use. and that kind of thing can help make billing and utilization tracking easier and also allow for perhaps even custom app integration | 06:19 |
| notmyname | eg you might have an app that have lots of users and that app could be a reseller for a whole set of accounts in swift | 06:20 |
| *** ppai has quit IRC | 06:20 | |
| ppai_ | Ah, thanks! These auth middlewares would be in same pipeline! ? | 06:20 |
| notmyname | ppai_: yes, exactly | 06:21 |
| ppai_ | notmyname, mattoliverau, thanks guys! | 06:23 |
| mattoliverau | ppai_: anytime | 06:23 |
| ppai_ | todo for self: try with tempauth and swauth in same proxy pipeline.. | 06:23 |
| mattoliverau | notmyname: thanks :) I was trying to help ppai_ yesterday but got stuck at the reseller_prefix, thus why I asked this moring (my time) | 06:24 |
| notmyname | yup. no worries | 06:24 |
| ahale | yep we have three (i think) reseller prefixes in our cluster, for all the reasons notmyname mentioned.. all next to each other in the pipeline | 06:25 |
| ppai_ | cool :) | 06:26 |
| notmyname | ahale: I still think it would have been really cool to expose that at RAX and to allow partners/customers to implement the auth. ie delegated auth to groups of cloud files accounts. | 06:27 |
| notmyname | ahale: there be dragons there, but I still think it would be cool | 06:27 |
| ahale | yeah that would be really cool | 06:27 |
| *** nsquare has joined #openstack-swift | 06:27 | |
| ahale | i would worry about proxies getting hung up talking to auth endpoints beyond our control though | 06:28 |
| notmyname | yes. dragons ;-) | 06:28 |
| ahale | hehe | 06:28 |
| ahale | I wonder if the barbican folks have considered those dragons in their remote hsm stuff they wanted to plug into swift | 06:29 |
| notmyname | but if you could do some clever revenue share, and then the "cloud application" that the third party has written actually removes the "buy our app and bing your cloud creds" requirement. instead "buy our app (or free) and get an account at RAX" | 06:29 |
| *** bvandenh has joined #openstack-swift | 06:29 | |
| notmyname | ahale: hmm..I'm sure they'll be told if it's proposed and people are paying attention :-) | 06:29 |
| ahale | big if's :) | 06:30 |
| notmyname | the "proxy talking to external services" doesn't worry me too much in a business sense. it's exactly the same thing that happens now with normal corporate partnerships. define and SLA and monitor it. it's actually _exactly_ the same thing cloud files does with the CDN :-) | 06:31 |
| notmyname | that is, the edge->origin connection isn't completely controlled by either party and contracts and monitoring tools enforce that it stays up and fingers don't get pointed when something happens | 06:32 |
| ahale | yeah that part can be handled with contracts .. its more memories of the headaches when identity calls start piling up, memcaches get overwhelmed and start failing | 06:33 |
| notmyname | oh yeah. doesn't make the dragons go away. just hopefully keeps them penned up :-) | 06:34 |
| ahale | adding more potential poits of failure like that in a busy system would be pretty nasty from ops side | 06:34 |
| ahale | in a perfect world it would be really nice though to let people stand up their own auths though, you could do some neat things with corp identity systems and access/cache appliances | 06:36 |
| ahale | ah I should get to the office and start fixing replicators :) | 06:38 |
| notmyname | hmm..but maybe that is just a local maxima. better auth that does [insert magic thing here] may be better than setting up lots of auth silos | 06:38 |
| notmyname | ahale: good times | 06:38 |
| notmyname | I got my presentation (preso for the aussies) done. looks like it's just about time for me to turn in | 06:39 |
| ahale | indeed - stupid replicators >.< | 06:39 |
| ahale | (or stupid drives, or servers) | 06:40 |
| *** ujjain has joined #openstack-swift | 06:40 | |
| *** ujjain has quit IRC | 06:42 | |
| *** ujjain has joined #openstack-swift | 06:42 | |
| mattoliverau | woo aussies! | 06:45 |
| *** ppai__ has joined #openstack-swift | 06:52 | |
| *** ppai_ has quit IRC | 06:56 | |
| goodes | aussie aussie aussie !! | 06:58 |
| *** nshaikh has quit IRC | 07:01 | |
| *** psharma has quit IRC | 07:08 | |
| *** ujjain has quit IRC | 07:09 | |
| *** openstackgerrit has quit IRC | 07:10 | |
| *** psharma has joined #openstack-swift | 07:11 | |
| *** jamiehannaford has joined #openstack-swift | 07:12 | |
| *** jamiehannaford has quit IRC | 07:13 | |
| *** jyoti_ranjan has quit IRC | 07:18 | |
| *** mmcardle has joined #openstack-swift | 07:23 | |
| *** kopparam has quit IRC | 07:25 | |
| *** mmcardle has quit IRC | 07:34 | |
| *** mmcardle has joined #openstack-swift | 07:38 | |
| *** nsquare has quit IRC | 07:39 | |
| *** nacim has joined #openstack-swift | 07:41 | |
| *** foexle has joined #openstack-swift | 07:42 | |
| *** jyoti_ranjan has joined #openstack-swift | 07:46 | |
| *** dANOKELOFF has joined #openstack-swift | 07:49 | |
| *** kopparam has joined #openstack-swift | 07:56 | |
| *** joeljwright has joined #openstack-swift | 07:59 | |
| *** nacim has quit IRC | 08:09 | |
| *** kopparam has quit IRC | 08:15 | |
| *** kopparam has joined #openstack-swift | 08:15 | |
| *** kopparam has quit IRC | 08:20 | |
| *** kopparam has joined #openstack-swift | 08:22 | |
| *** ujjain has joined #openstack-swift | 08:23 | |
| *** nacim has joined #openstack-swift | 08:24 | |
| *** nacim has quit IRC | 08:28 | |
| *** nshaikh has joined #openstack-swift | 08:33 | |
| *** Kbee has quit IRC | 08:35 | |
| *** Kbee has joined #openstack-swift | 08:36 | |
| *** nacim has joined #openstack-swift | 08:41 | |
| *** ujjain has quit IRC | 08:42 | |
| *** ujjain has joined #openstack-swift | 08:42 | |
| *** jamiehannaford has joined #openstack-swift | 08:58 | |
| *** kopparam has quit IRC | 09:03 | |
| *** dmorita has quit IRC | 09:07 | |
| *** kopparam has joined #openstack-swift | 09:07 | |
| *** Kbee has quit IRC | 09:09 | |
| *** Kbee has joined #openstack-swift | 09:09 | |
| *** jamiehannaford has quit IRC | 09:17 | |
| *** jyoti_ranjan has quit IRC | 09:18 | |
| *** jyoti_ranjan has joined #openstack-swift | 09:26 | |
| *** dANOKELOFF has quit IRC | 09:27 | |
| *** zhiyan is now known as zhiyan_ | 09:28 | |
| *** baojg_ has joined #openstack-swift | 09:29 | |
| *** baojg has quit IRC | 09:30 | |
| *** pconstantine has quit IRC | 09:31 | |
| *** pconstantine_ is now known as pconstantine | 09:31 | |
| *** pconstantine_ has joined #openstack-swift | 09:32 | |
| *** kopparam has quit IRC | 09:33 | |
| *** openstackgerrit has joined #openstack-swift | 09:46 | |
| *** psharma has quit IRC | 09:51 | |
| *** jyoti_ranjan has quit IRC | 10:05 | |
| *** jamiehannaford has joined #openstack-swift | 10:09 | |
| *** jamiehannaford has quit IRC | 10:10 | |
| *** dANOKELOFF has joined #openstack-swift | 10:10 | |
| *** jyoti_ranjan has joined #openstack-swift | 10:14 | |
| *** kopparam has joined #openstack-swift | 10:14 | |
| *** nosnos has quit IRC | 10:19 | |
| *** nosnos has joined #openstack-swift | 10:19 | |
| *** ppai__ has quit IRC | 10:20 | |
| *** matsuhashi has quit IRC | 10:23 | |
| *** matsuhashi has joined #openstack-swift | 10:24 | |
| *** nosnos has quit IRC | 10:24 | |
| *** joeljwright has left #openstack-swift | 10:25 | |
| *** matsuhashi has quit IRC | 10:28 | |
| *** mkollaro has joined #openstack-swift | 10:33 | |
| *** dANOKELOFF has quit IRC | 10:37 | |
| *** haomaiwa_ has joined #openstack-swift | 10:54 | |
| *** Kbee has quit IRC | 11:07 | |
| *** dmsimard_away is now known as dmsimard | 11:33 | |
| *** bsdkurt has quit IRC | 11:34 | |
| *** bsdkurt has joined #openstack-swift | 11:35 | |
| *** ppai has joined #openstack-swift | 11:37 | |
| *** haomaiwa_ has quit IRC | 11:38 | |
| *** Longgeek_ has joined #openstack-swift | 11:39 | |
| *** kopparam has quit IRC | 11:46 | |
| *** kopparam has joined #openstack-swift | 11:47 | |
| *** kopparam has quit IRC | 11:51 | |
| *** joeljwright has joined #openstack-swift | 11:53 | |
| *** mkollaro has quit IRC | 12:03 | |
| *** lpabon has joined #openstack-swift | 12:04 | |
| *** nacim has quit IRC | 12:05 | |
| *** mmcardle has quit IRC | 12:08 | |
| *** dANOKELOFF has joined #openstack-swift | 12:11 | |
| *** mmcardle has joined #openstack-swift | 12:15 | |
| *** baojg_ has quit IRC | 12:16 | |
| *** baojg has joined #openstack-swift | 12:16 | |
| *** baojg has quit IRC | 12:17 | |
| *** kopparam has joined #openstack-swift | 12:17 | |
| *** baojg has joined #openstack-swift | 12:17 | |
| *** nacim has joined #openstack-swift | 12:21 | |
| *** kopparam has quit IRC | 12:22 | |
| *** baojg has quit IRC | 12:22 | |
| *** kopparam has joined #openstack-swift | 12:26 | |
| *** mmcardle has quit IRC | 12:26 | |
| *** mmcardle has joined #openstack-swift | 12:30 | |
| dANOKELOFF | hi everyone, when i went to start swift with : swift-init all start i have error : could not bind to @ip after trying 30s. How i can solve this error ? | 12:39 |
| *** mmcardle has quit IRC | 12:42 | |
| ahale | do you have something already listening on the port you're trying to start stuff on ? | 12:45 |
| *** ppai has quit IRC | 12:45 | |
| dANOKELOFF | yes | 12:47 |
| dANOKELOFF | i try somethings | 12:47 |
| *** NM1 has joined #openstack-swift | 12:48 | |
| *** CaioBrentano has joined #openstack-swift | 12:49 | |
| *** ajc_ has quit IRC | 12:50 | |
| *** flowouffff is now known as fuz1on | 12:50 | |
| *** mkollaro has joined #openstack-swift | 13:02 | |
| *** baojg has joined #openstack-swift | 13:06 | |
| *** gholt has quit IRC | 13:07 | |
| *** bkopilov has quit IRC | 13:08 | |
| *** gholt has joined #openstack-swift | 13:09 | |
| *** ChanServ sets mode: +v gholt | 13:09 | |
| *** nacim has quit IRC | 13:11 | |
| *** kopparam has quit IRC | 13:21 | |
| *** Longgeek_ has quit IRC | 13:22 | |
| *** nacim has joined #openstack-swift | 13:28 | |
| *** annegent_ has joined #openstack-swift | 13:30 | |
| *** bboris has joined #openstack-swift | 13:33 | |
| bboris | hi | 13:34 |
| peluse_ | notmyname: wondering if we need to supply a policy patch for ceilometer? http://docs.openstack.org/developer/ceilometer/measurements.html#object-storage-swift | 13:35 |
| bboris | what is the fastest way to delete containers? | 13:35 |
| bboris | for testing purposes | 13:35 |
| bboris | i'm testing write performance and i must delete containers quickly, preferably not all the objects one by one through http | 13:36 |
| *** jasondotstar has joined #openstack-swift | 13:47 | |
| *** zz_wasmum is now known as wasmum | 13:48 | |
| *** mkollaro1 has joined #openstack-swift | 13:59 | |
| *** mkollaro has quit IRC | 13:59 | |
| *** jamiehannaford has joined #openstack-swift | 14:00 | |
| *** annegent_ has quit IRC | 14:01 | |
| *** jyoti_ranjan has quit IRC | 14:06 | |
| *** kenhui has joined #openstack-swift | 14:12 | |
| *** kris_h has joined #openstack-swift | 14:13 | |
| *** kris_h has quit IRC | 14:24 | |
| notmyname | peluse_: almost certainly | 14:31 |
| dANOKELOFF | hi , when i type the swift command for list all containers i have this error message : https://gist.github.com/anonymous/017bdaaa67da02f02c5a | 14:44 |
| *** mkollaro1 has quit IRC | 14:46 | |
| *** mkollaro has joined #openstack-swift | 14:47 | |
| *** foexle has quit IRC | 14:48 | |
| *** nacim has quit IRC | 14:51 | |
| *** nshaikh has left #openstack-swift | 14:52 | |
| *** joeljwright has quit IRC | 14:54 | |
| *** jamiehannaford has quit IRC | 14:55 | |
| *** rpedde has quit IRC | 14:56 | |
| *** miqui_ has joined #openstack-swift | 14:57 | |
| *** miqui has quit IRC | 14:58 | |
| acoles | dANOKELOFF: for your command to work your proxy-server.conf would need a line 'user_test_test = test .admin' under the tempauth section. | 14:58 |
| dANOKELOFF | acoles: already done | 14:58 |
| *** elambert has joined #openstack-swift | 14:59 | |
| *** rpedde has joined #openstack-swift | 14:59 | |
| *** bvandenh has quit IRC | 15:00 | |
| dANOKELOFF | i think the problem is the result write two times the port 8080 | 15:01 |
| *** tdasilva has joined #openstack-swift | 15:01 | |
| dANOKELOFF | see : Account GET failed: http://10.9.9.9:8080:8080/v1/.... | 15:01 |
| *** jkelly has joined #openstack-swift | 15:03 | |
| *** kevinc_ has joined #openstack-swift | 15:05 | |
| *** byeager has joined #openstack-swift | 15:06 | |
| peluse_ | notmyname: OK, wrt ceil sstuff is that something one of us does or do we ping someone over there or what? | 15:08 |
| tdasilva | dANOKELOFF: are you forgetting to close the quote on your password? swift -A http://10.9.9.9:8080/auth/v1.0 -U ‘test:test’ | 15:10 |
| tdasilva | -K ‘test list | 15:10 |
| dANOKELOFF | tdasilva: same problem | 15:11 |
| *** miqui_ is now known as miqui | 15:11 | |
| dANOKELOFF | tdasilva: if i use the command curl , for take the token that work , but if i use the command curl for inspect the contenair with token that doesnt work :( | 15:13 |
| dANOKELOFF | i really dont understand that bug. | 15:13 |
| tdasilva | dANOKELOFF: are you using a saio setup or a multinode cluster? can you post your proxy-server.conf to a gist? | 15:15 |
| acoles | dANOKELOFF: I think the double port number may be a fixed bug https://bugs.launchpad.net/swift/+bug/1326321 (a bug in the client exception report) | 15:16 |
| dANOKELOFF | tdasilva: I follow this guide : http://docs.openstack.org/developer/swift/howto_installmultinode.html and my proxy-server.conf = https://gist.github.com/anonymous/4dd8a92bf586c3f00acf | 15:16 |
| dANOKELOFF | acoles: How i can update my swift version ? because i'm in 2.0.3 | 15:18 |
| acoles | dANOKELOFF: the conf file you just posted doesn't have a tempauth entry for test:test | 15:19 |
| dANOKELOFF | acoles: yeah, sorry but for my test i use dano:authot authot | 15:19 |
| *** psharma has joined #openstack-swift | 15:22 | |
| dANOKELOFF | acoles: How i can upgrade my swift version to 2.1.0 ? | 15:22 |
| *** mmcardle has joined #openstack-swift | 15:22 | |
| *** byeager has quit IRC | 15:28 | |
| *** byeager has joined #openstack-swift | 15:29 | |
| *** mmcardle has quit IRC | 15:32 | |
| *** haomaiwa_ has joined #openstack-swift | 15:34 | |
| *** kenhui has quit IRC | 15:35 | |
| *** mmcardle has joined #openstack-swift | 15:35 | |
| *** kenhui has joined #openstack-swift | 15:35 | |
| dANOKELOFF | How i can upgrade swift 2.0.3 to 2.1.0 ? | 15:37 |
| *** gyee has joined #openstack-swift | 15:40 | |
| tdasilva | dANOKELOFF: how did you install it? | 15:40 |
| dANOKELOFF | with the guide , sudo apt-get install swift python-swiftclient | 15:40 |
| dANOKELOFF | but if i do sudo apt-get upgrade swift python-swiftclient | 15:41 |
| dANOKELOFF | i have no update | 15:41 |
| tdasilva | can you un-install and do a git clone? | 15:42 |
| acoles | dANOKELOFF: (sorry, stepped away) you can grab from here https://github.com/openstack/python-swiftclient/tree/2.1.0 if not via your distro | 15:42 |
| dANOKELOFF | acoles: ok , thank's | 15:43 |
| *** dANOKELOFF has quit IRC | 15:44 | |
| *** dmsimard has quit IRC | 15:46 | |
| *** dmsimard has joined #openstack-swift | 15:47 | |
| *** byeager has quit IRC | 15:54 | |
| *** byeager has joined #openstack-swift | 15:54 | |
| notmyname | good morning, world | 15:55 |
| *** NM1 has quit IRC | 15:56 | |
| *** NM has joined #openstack-swift | 15:59 | |
| notmyname | FYI, I'm being asked to provide the list of "designated sections" of Swift by the end of the week. guidelines are at http://git.openstack.org/cgit/openstack/governance/tree/resolutions/20140402-defcore-designated-sections-guidelines.rst | 15:59 |
| acoles | notmyname: morning | 16:00 |
| *** jkelly has quit IRC | 16:00 | |
| notmyname | As I build the list, I'll probably be asking around for input and review | 16:00 |
| *** annegent_ has joined #openstack-swift | 16:00 | |
| acoles | notmyname: clayg: i put some thoughts re. fast-POST on a wiki (as promised at last week's meeting) https://wiki.openstack.org/wiki/Swift/FastPost | 16:01 |
| swifterdarrell | notmyname: what's the granularity on that? per-file? per-class? | 16:01 |
| *** annegent_ has quit IRC | 16:01 | |
| notmyname | swifterdarrell: isnt' that a good question? :-) | 16:01 |
| swifterdarrell | notmyname: isn't there a good answer? ;) | 16:01 |
| *** annegent_ has joined #openstack-swift | 16:01 | |
| acoles | gholt: redbo: creiht: would appreciate your wisdom on this too https://wiki.openstack.org/wiki/Swift/FastPost | 16:02 |
| acoles | and anyone else... | 16:02 |
| notmyname | swifterdarrell: not really. it could be down to class-level. generally, it's to give people something to check when they make the assertion that they are running openstack code | 16:02 |
| creiht | if only we had a repo where we could propose such specs.... | 16:02 |
| notmyname | creiht: should be soon. mordred was getting that together yesterday afternoon | 16:03 |
| creiht | :) | 16:03 |
| notmyname | acoles: thanks. I'll look it over later today | 16:03 |
| acoles | notmyname: thanks. i have to finish early now anyway. | 16:04 |
| peluse_ | notmyname: fyi, swift priority reviews needs updating (still lists all the SP patches) | 16:04 |
| notmyname | peluse_: correct | 16:04 |
| *** bboris has quit IRC | 16:04 | |
| * notmyname needs to write all this down ;-) | 16:04 | |
| *** byeager has quit IRC | 16:05 | |
| peluse_ | notmyname: curious on the DefCore link above - they are asking you for a list of functions or modules or what? | 16:07 |
| notmyname | peluse_: are you looking in to the ceilometer updates for SP account-level reporting, or would you like me to look in to it? | 16:07 |
| *** jkelly has joined #openstack-swift | 16:07 | |
| notmyname | peluse_: and like swifterdarrell, you're getting into the ambiguity of the defcore process. as I understand it, the designated sections are the parts of the openstack project code that a deployer asserts they are running. combined with the results of functional tests, this gives them the right to use the openstack trademark. | 16:08 |
| notmyname | which means that the designated sections need to be specific enough to follow for a deployer, but also not so specific that eg RAX's non-pbr setup.py file means it's not swift. IMO the middleware and the daemon processes are where it gets tricky | 16:10 |
| gholt | acoles: re: fast post: Is there consideration of container-sync and getting updated info to the far side? | 16:10 |
| peluse_ | notmyname: ugh... sorry, missed the dup'd question. Prety poor wording they're using there "Sections of code should generally be DESIGNATED when:" | 16:16 |
| notmyname | peluse_: don't worry. you aren't the first and won't be the last to ask and get an ambiguous answer :-) | 16:16 |
| *** kenhui has quit IRC | 16:17 | |
| *** bvandenh has joined #openstack-swift | 16:18 | |
| *** judd7_ has joined #openstack-swift | 16:24 | |
| *** jyoti_ranjan has joined #openstack-swift | 16:28 | |
| *** jyoti_ranjan2 has joined #openstack-swift | 16:31 | |
| *** jyoti_ranjan has quit IRC | 16:32 | |
| *** jyoti_ranjan has joined #openstack-swift | 16:35 | |
| *** pberis has quit IRC | 16:35 | |
| *** byeager has joined #openstack-swift | 16:37 | |
| *** jyoti_ranjan2 has quit IRC | 16:37 | |
| notmyname | `pbpaste | wc` is terribly handy | 16:45 |
| notmyname | ell really just pbpaste and pbcopy overall are very nice | 16:45 |
| notmyname | for non mac users: "pbcopy(1), pbpaste(1) - provide copying and pasting to the pasteboard (the Clipboard) from command line" | 16:46 |
| notmyname | is there a similar tool in linux userland? | 16:46 |
| peluse_ | notmyname: I putty in to all my linux VMs so can just highlight to copy and right click to paste... | 16:51 |
| notmyname | peluse_: in this case I was writing a short blog post and needed a word count. the app I was using didn't have it so `pbpaste | wc -w` was my solution | 16:52 |
| peluse_ | way cool... | 16:53 |
| *** jkelly has quit IRC | 16:56 | |
| *** byeager has quit IRC | 16:57 | |
| *** CaioBrentano has quit IRC | 17:01 | |
| *** CaioBrentano has joined #openstack-swift | 17:04 | |
| notmyname | "storage-policy-historical" tag was pushed to gerrit this morning at the HEAD of feature/ec. Now we'll delete feature/ec and recreate it against current master to give the actual EC patches a place to land | 17:11 |
| *** NM has quit IRC | 17:11 | |
| notmyname | peluse_: ^^ | 17:11 |
| *** mkollaro has quit IRC | 17:11 | |
| peluse_ | cool | 17:11 |
| *** nsquare has joined #openstack-swift | 17:11 | |
| *** NM has joined #openstack-swift | 17:13 | |
| *** Kbee has joined #openstack-swift | 17:14 | |
| *** zaitcev has joined #openstack-swift | 17:14 | |
| *** ChanServ sets mode: +v zaitcev | 17:14 | |
| *** mlipchuk has joined #openstack-swift | 17:15 | |
| *** mkollaro has joined #openstack-swift | 17:16 | |
| *** haomaiwa_ has quit IRC | 17:16 | |
| *** NM has quit IRC | 17:16 | |
| *** NM has joined #openstack-swift | 17:19 | |
| *** shri has joined #openstack-swift | 17:20 | |
| *** jyoti_ranjan has quit IRC | 17:23 | |
| *** mmcardle has quit IRC | 17:28 | |
| notmyname | peluse_: clayg: in order to delete the current feature/ec branch I need to abandon all changes that are currently proposed against it | 17:31 |
| *** byeager has joined #openstack-swift | 17:33 | |
| notmyname | peluse_: but there is an important note here | 17:33 |
| notmyname | peluse_: that includes the current WIP patches you, tushar, and Yuan have proposed | 17:33 |
| notmyname | peluse_: you three will need to repropose these patches to master or the new feature/ec | 17:33 |
| clayg | gholt: acoles main work on fast-post is getting it to work with container-sync | 17:34 |
| clayg | notmyname: that's got to be fine, they need to be rebased against the new state of the world anyway | 17:35 |
| *** openstackgerrit has quit IRC | 17:35 | |
| *** openstackgerrit has joined #openstack-swift | 17:35 | |
| notmyname | peluse_: importantly those WIP patches, when reproposed will need a new change-id in the commit message so gerrit doesn't get all weird. I think you'll get an error if you try to use the same id, so we're sorta protected | 17:36 |
| openstackgerrit | Samuel Merritt proposed a change to openstack/swift: Fix some spelling in comments https://review.openstack.org/102297 | 17:37 |
| *** kevinc_ has quit IRC | 17:38 | |
| *** byeager has quit IRC | 17:38 | |
| *** byeager has joined #openstack-swift | 17:40 | |
| notmyname | clayg: mordred isn't yet active on IRC today, so I haven't talked to him, but current consensus from clarkb and jeblair is to abandon the currently open SP patches in gerrit. I'll keep looking for mordred and hold out hope they can be marked "merged", but either way, it will be resolved today. | 17:41 |
| notmyname | clayg: and I'll take care of the tedium. don't worry about marking them all abandoned | 17:41 |
| *** jkelly has joined #openstack-swift | 17:42 | |
| clayg | so allegedly the pbr TypeError dist must be a Distribution is fixed? https://review.openstack.org/#/c/101785/ | 17:48 |
| clayg | notmyname: ok | 17:48 |
| notmyname | clayg: that's good | 17:49 |
| *** kevinc_ has joined #openstack-swift | 17:52 | |
| *** shri1 has joined #openstack-swift | 17:59 | |
| *** jkelly has quit IRC | 18:00 | |
| *** shri has quit IRC | 18:00 | |
| *** shri1 has quit IRC | 18:00 | |
| openstackgerrit | Samuel Merritt proposed a change to openstack/swift: Don't count 412 or 416 as errors in stats https://review.openstack.org/98530 | 18:05 |
| notmyname | yuan: you had some old draft reviews on the feature/ec branch that were preventing us from moving the feature/ec branch. as you can see from your email, those have been published and abandoned | 18:09 |
| *** mjseger has joined #openstack-swift | 18:10 | |
| notmyname | all: the feature/ec branch has now been deleted and then recreated off of current master. this is where EC dev work will go until it's ready to be merged to master | 18:10 |
| notmyname | consider feature/ec unfrozen and open for patches | 18:11 |
| elambert | notmyname: peluse_ any chance one of you can add me a member of the ec trello board? | 18:11 |
| *** psharma has quit IRC | 18:11 | |
| notmyname | elambert: yup. what's your trello id or email to use? | 18:11 |
| *** mkollaro has quit IRC | 18:11 | |
| elambert | ericlambert1 | 18:12 |
| notmyname | elambert: done | 18:12 |
| elambert | thanks! | 18:12 |
| mjseger | I'd like to share some of the work I've been doing with statsd and see what people think | 18:13 |
| notmyname | mjseger: patches or prose? | 18:13 |
| mjseger | prose | 18:13 |
| mjseger | basically I'm collecting data and writing to a file of rolling counters similar to /proc | 18:13 |
| mjseger | there are currently 2 formats, one with headers and one without, see this: | 18:14 |
| notmyname | mjseger: sending it to the openstack-dev mailing list with the [Swift] topic in the subject would also let people who aren't here now see it | 18:14 |
| mjseger | http://paste.openstack.org/show/84813/ | 18:14 |
| mjseger | ok, I'll do that, even though I just gave you a preview ;) | 18:14 |
| mjseger | stay tuned | 18:15 |
| notmyname | that's good. talking in here is fine, but the ML is a slightly wider net. here is more back-and-forth, though, depending on who's online and what they're doing | 18:15 |
| notmyname | mjseger: eg, I'd expect swifterdarrell to be interested, but I think he's working on some swiftstack product stuff right now. he may be able to peruse a ML message later though | 18:16 |
| mjseger | gottcha | 18:16 |
| mjseger | thx | 18:16 |
| *** CaioBrentano has quit IRC | 18:17 | |
| *** psharma has joined #openstack-swift | 18:18 | |
| swifterdarrell | mjseger: love it | 18:18 |
| *** Kbee has quit IRC | 18:21 | |
| *** NM has quit IRC | 18:22 | |
| *** NM has joined #openstack-swift | 18:23 | |
| *** mlipchuk has quit IRC | 18:25 | |
| *** wasmum is now known as zz_wasmum | 18:29 | |
| *** CaioBrentano has joined #openstack-swift | 18:32 | |
| *** zz_wasmum is now known as wasmum | 18:33 | |
| *** NM has quit IRC | 18:33 | |
| *** psharma has quit IRC | 18:33 | |
| notmyname | going offline for a bit while I do the PTL webinar thingy | 18:41 |
| mjseger | notmyname: swifterdarrell: email to the list on the way | 18:41 |
| notmyname | mjseger: thanks | 18:41 |
| *** pconstantine has quit IRC | 18:41 | |
| *** wasmum is now known as zz_wasmum | 18:42 | |
| *** pconstantine_ has quit IRC | 18:44 | |
| peluse_ | notmyname: got all the above notes on feature/ec, sounds great - thanks! | 18:47 |
| *** swat30 has quit IRC | 18:48 | |
| *** swat30 has joined #openstack-swift | 18:51 | |
| *** zz_wasmum is now known as wasmum | 18:52 | |
| *** pconstantine_ has joined #openstack-swift | 18:52 | |
| *** joeljwright has joined #openstack-swift | 19:06 | |
| goodes | clayg: any reason that you only have date in VM name for vagrant-saio (stops me from running multiple standalone instances at same time as it complains they have same name)? | 19:25 |
| *** cm2 has joined #openstack-swift | 19:25 | |
| clayg | goodes: are you sure you've got the latest patches? | 19:26 |
| goodes | clayg: checking | 19:27 |
| clayg | goodes: oh hrmm... i thought the resolution on that got higher, but github says now | 19:27 |
| clayg | goodes: you could try the extra_vms option, it's not super fleshed out, but generally nah i think it'd be fine if you wanted to add yymmdd-time/hash to the vmname | 19:28 |
| clayg | goodes: send a PR! | 19:28 |
| goodes | clayg: I made my own patch - adding -%H%M%S") to it and then moving it out so that extra_vms all have the same name | 19:28 |
| *** jergerber has joined #openstack-swift | 19:29 | |
| notmyname | what distro is "SLES"? | 19:29 |
| goodes | clayg: will also send a PR to allow you to override the git URLS for swift, swiftclient etc to use a fork or branch | 19:30 |
| clayg | goodes: cool! | 19:31 |
| zackmdavis | notmyname, "SUSE Linux Enterprise Server"? | 19:33 |
| notmyname | zackmdavis: thanks :-) | 19:33 |
| notmyname | zackmdavis: thanks google zack :-) | 19:33 |
| *** jasondotstar has quit IRC | 19:33 | |
| *** byeager has quit IRC | 19:33 | |
| tdasilva | clayg: goodes: if I wanted to add a 5th object server using the vagrant-saio VM, do I have to use the same .d dir structure? | 19:35 |
| tdasilva | for the configuration file? | 19:35 |
| *** wasmum is now known as zz_wasmum | 19:37 | |
| *** jkelly has joined #openstack-swift | 19:42 | |
| *** zz_wasmum is now known as wasmum | 19:45 | |
| *** kenhui has joined #openstack-swift | 19:48 | |
| *** kevinc_ has quit IRC | 19:52 | |
| *** baojg has quit IRC | 19:53 | |
| *** wasmum is now known as zz_wasmum | 19:54 | |
| *** jamiehannaford has joined #openstack-swift | 19:56 | |
| *** byeager has joined #openstack-swift | 20:02 | |
| *** joeljwright has quit IRC | 20:09 | |
| *** redbo has quit IRC | 20:11 | |
| *** redbo has joined #openstack-swift | 20:12 | |
| *** ChanServ sets mode: +v redbo | 20:12 | |
| *** byeager has quit IRC | 20:13 | |
| *** zz_wasmum is now known as wasmum | 20:15 | |
| *** byeager has joined #openstack-swift | 20:18 | |
| *** kevinc_ has joined #openstack-swift | 20:20 | |
| *** fifieldt_ has joined #openstack-swift | 20:22 | |
| zaitcev | At least he's not asking about RHEL or LTS | 20:23 |
| mjseger | am I the only one who finds help in the swift client difficult? for example the only way I can see to get help on switch options is to enter an invalid number of options, or am I missing something? | 20:24 |
| goodes | mjseger: +1 | 20:25 |
| mjseger | I'd like to be able to type "swift command help or --help and get context specific help, not a long blast of all the options. kind of like this: | 20:25 |
| mjseger | http://paste.openstack.org/show/84820/ | 20:25 |
| *** fifieldt has quit IRC | 20:25 | |
| mjseger | I'd be happy to submit a patch, but wanted to get a sense of what people thought first | 20:26 |
| goodes | tdasilva: not sure what you're asking | 20:26 |
| *** jergerber has quit IRC | 20:26 | |
| clayg | tdasilva: *have* to, no you just need a .conf file or .conf.d dir in /etc/swift/object-server | 20:27 |
| clayg | prolly like a 5.conf or a 5.conf.d | 20:27 |
| clayg | tdasilva: i had some notion that might be configurable in localrc, something like "NODES" - but that probably effects all the servers not just objects (if that matters) | 20:27 |
| zackmdavis | mjseger, +1 to --help patch also | 20:33 |
| *** kevinc_ has quit IRC | 20:34 | |
| *** byeager has quit IRC | 20:35 | |
| *** shri has joined #openstack-swift | 20:35 | |
| *** kevinc_ has joined #openstack-swift | 20:36 | |
| goodes | clayg: looks like we may be hitting limit of localrc - as people start wanted to do all kinds of scenarios including multiple nodes/vms and replication you may need to move to a configuration template rather then the flat localrc model | 20:36 |
| *** byeager has joined #openstack-swift | 20:38 | |
| goodes | clayg: i.e. specify configuration per node (i.e. IPs, server type and count), and per profile ec or replication count | 20:38 |
| notmyname | mjseger: yes! | 20:47 |
| mjseger | ok, I'll submit a patch shortly | 20:48 |
| clayg | goodes: well, as simple as we can keep it and simpler i suppose - not absolutly everything concieveable has to be expressable either, once the things provisioned it's just a box - you can do additional configuration out of band | 20:49 |
| clayg | goodes: I currently has a .scratch dir in the root of the vagrant shared file system that I stuff things in that I want to run from inside my saio | 20:50 |
| clayg | goodes: I had some idea that we could make something along the lines of an autoexec.bat that gets run post provision so folks could do additional automated provisioning on bring up | 20:51 |
| notmyname | anyone played with running swift in docker? http://serverascode.com/2014/06/12/run-swift-in-docker.html | 20:52 |
| clayg | goodes: i'm not above a contrib dir that contains .patch files that appy to cookbooks dir either | 20:52 |
| clayg | notmyname: that seems quite a bit more sane than the last swift-in-docker setup I saw which did a bunch of loopback filesystem non-sense, using the docker volumes seems much more interesting | 20:55 |
| *** kenhui has quit IRC | 20:57 | |
| *** jamiehannaford has quit IRC | 21:00 | |
| *** wasmum is now known as zz_wasmum | 21:04 | |
| *** bvandenh has quit IRC | 21:06 | |
| creiht | notmyname: I did it once, and it was very ugly | 21:07 |
| creiht | but I was doing a full saio | 21:08 |
| *** byeager has quit IRC | 21:10 | |
| *** jasond has joined #openstack-swift | 21:19 | |
| tdasilva | goodes, clayg: thanks, i was trying to configure another storage policy and needed a 5th object-server to be running | 21:25 |
| *** tdasilva has quit IRC | 21:28 | |
| *** pberis has joined #openstack-swift | 21:38 | |
| *** kevinc_ has quit IRC | 21:41 | |
| *** kevinc_ has joined #openstack-swift | 21:50 | |
| *** jamiehannaford has joined #openstack-swift | 21:50 | |
| *** jamiehannaford has quit IRC | 21:51 | |
| *** jkelly has quit IRC | 21:53 | |
| *** annegent_ has quit IRC | 21:57 | |
| openstackgerrit | Longgeek proposed a change to openstack/python-swiftclient: Updated from global requirements https://review.openstack.org/102365 | 22:08 |
| *** pberis has quit IRC | 22:11 | |
| *** miqui has quit IRC | 22:11 | |
| mattoliverau | Morning | 22:14 |
| peluse_ | notmyname: are all of your avoton systems up and running again? | 22:18 |
| mattoliverau | notmyname: how'd your presentation go earlier? | 22:20 |
| *** dmsimard is now known as dmsimard_away | 22:20 | |
| notmyname | mattoliverau: went well | 22:26 |
| notmyname | peluse_: close to it, I think. | 22:26 |
| notmyname | peluse_: hugokuo ran some benchmarks last night (yesterday his time) and I'm planning on comparing those to the same test against the 2.0 RC | 22:27 |
| notmyname | mattoliverau: video should be up by end of week | 22:27 |
| mattoliverau | notmyname: cool, I look forward to seeing it :) | 22:28 |
| peluse_ | notmyname: cool, we're having a few issues with our stuff, Bill is going to ping Martin here shortly about some network settings or something - we're going to re-run all that stuff we reveiwed several weeks ago on an avoton cluster and compare | 22:30 |
| notmyname | peluse_: great. I think the "close to it" has to do with some kernel/OS issue. thanks for checking (/cc mlanner) | 22:31 |
| *** CaioBrentano has quit IRC | 22:34 | |
| shri | Hey all… I was running some experiments against my single node swift cluster. Observed something interesting: http://www.gossamer-threads.com/lists/openstack/dev/39260 | 22:38 |
| torgomatic | shri: yeah, the guy that replied to you was right | 22:39 |
| torgomatic | the proxy streams the object data from client --> object server, so higher client latency will slow down small PUTs | 22:39 |
| torgomatic | there's only minimal buffering on the proxy side of things (like a few KiB) | 22:40 |
| openstackgerrit | A change was merged to openstack/swift: Replace POLICY and POLICY_INDEX with string literals https://review.openstack.org/101991 | 22:40 |
| openstackgerrit | A change was merged to openstack/swift: Minor Updates to Storage Policy Docs https://review.openstack.org/101705 | 22:40 |
| shri | I see .. is there a config option controlling this? | 22:40 |
| torgomatic | shri: nope; upload streaming is always on | 22:40 |
| jroll | hey y'all, thoughts on something for generating temp urls in python-swiftclient? | 22:41 |
| torgomatic | if you really want to measure internal latency, try zero-byte objects | 22:41 |
| jroll | (we're happy to write the code, just want to see if it would be accepted) | 22:41 |
| shri | torgomatic: yeah… I tried a 1 byte object and it did not have that problem | 22:42 |
| torgomatic | jroll: could be useful to have; I'm neutral on the topic (but I've given it only a second or two of thought) | 22:42 |
| torgomatic | shri: yep, well, now you know :) | 22:42 |
| jroll | torgomatic: yeah. my thought is: I'm going to implement it. If you folks don't want it in the client, I'll probably toss it over to oslo. My inclination is that it belongs in the client, but I'm not a core here, so :) | 22:43 |
| torgomatic | jroll: fair enough :) | 22:43 |
| shri | torgomatic: but then when I look at the obj/server.py code, I see that data is read self.network_chunk_size bytes at a time | 22:43 |
| shri | If I update the network_chunk_size is the object-server.conf, shouldn't that take effect here? | 22:44 |
| torgomatic | shri: well, two things... first, you have to do that in the proxy configs, not just object configs | 22:44 |
| shri | so I add network_chunk_size in the proxy-server.conf and object-server.conf? | 22:45 |
| torgomatic | second, that's only for reading the request body, which isn't started until after the connections to the object servers have been created | 22:45 |
| notmyname | jroll: you had me at "we're happy to write the code" ;-) | 22:46 |
| torgomatic | IOW, the proxy (a) reads request headers, (b) opens connections to backends, (c) reads a chunk, (d) writes a chunk, (e) repeats c-d until body is done, (f) streams response to clients | 22:46 |
| jroll | notmyname: ha, it's not much code :) | 22:46 |
| jroll | notmyname: will post reviews soon, thanks | 22:46 |
| torgomatic | so if you're measuring latency from end-of-(a) until start-of-(f), then you've still got some streaming from the client going on | 22:47 |
| notmyname | jroll: and the existing bin script in swift is a great start | 22:47 |
| jroll | notmyname: yeah, we already have code based on that in an ironic review, but decided it should go elsewhere | 22:47 |
| torgomatic | the reason it goes away with 1-byte objects is that the response body can fit in the same TCP frame as the end of the headers, so when the proxy goes to read it, it's already sitting in a kernel buffer (or possibly a user-space buffer; not sure) | 22:47 |
| jroll | notmyname: for reference if you're interested: https://review.openstack.org/#/c/81391/38/ironic/common/glance_service/v2/image_service.py | 22:48 |
| notmyname | jroll: wow that has 38 patch sets | 22:49 |
| jroll | yeah, well :| | 22:49 |
| jroll | I didn't say I'm happy about that review :) | 22:49 |
| jroll | it's also been open for over 3 months | 22:50 |
| notmyname | jroll: I think a cool think about doing it in swift client is that you could (optionally) fetch the info from swift (the key and the tempurl limits from /info) | 22:50 |
| jroll | indeed | 22:50 |
| torgomatic | suggestion, though: make that two different functions | 22:50 |
| torgomatic | I'd want one that just does local computation; if I get some surprise network IO, it'll make me cranky | 22:51 |
| jroll | indeed | 22:52 |
| jroll | I'll see what I can do and we can talk more in code review :) | 22:52 |
| torgomatic | works for me :) | 22:52 |
| * torgomatic goes back to banging on some benchmarks | 22:52 | |
| jroll | thanks :) | 22:53 |
| *** lpabon has quit IRC | 22:56 | |
| *** mwstorer has joined #openstack-swift | 23:01 | |
| shri | torgomatic: sorry… was away from the desk for a bit. I like the explanation of steps a through f that you've mentioned. | 23:09 |
| torgomatic | shri: hope it helps | 23:09 |
| shri | Indeed I'm measuring the time from end of a through start of f. | 23:10 |
| shri | But I'm surprised that there isn't a way to configure the chunk size for steps c and d | 23:10 |
| torgomatic | shri: there is; it's the proxy's chunk sizes | 23:10 |
| shri | The proxy has object_chunk_size and object_chunk_size. Are you talking about these? | 23:12 |
| torgomatic | yep | 23:12 |
| shri | Aah… I see. | 23:12 |
| shri | So I can run an expt with these values bumped up to 256K in the proxy-server.conf and that should (in theory) do the trick. | 23:12 |
| shri | Let me try that.. | 23:12 |
| torgomatic | I don't think tuning that will give you the results you want to see, though. Zero chunks are buffered prior to opening backend connections, so the size of the chunks won't matter | 23:13 |
| shri | but bumping up the chunk size will reduce the number of RTTs between swift and the client, right? | 23:14 |
| torgomatic | not necessarily; the TCP frame size is going to be determined by the network MTU, not the size of the buffer used in a recvfrom() syscall | 23:15 |
| shri | oh.. that's right! | 23:15 |
| notmyname | I'm going to go through gerrit and abandon the 28 bp/storage-policies patches now (with a link to the merge commit in the message) | 23:16 |
| shri | torgomatic: thanks for the help. I'll run some more expts and get back to you.. Hopefully with a "Yay!! That worked"…but if not, maybe with more questions. Thanks! | 23:20 |
| torgomatic | shri: good luck! | 23:20 |
| peluse_ | shri: kinda curious to hear about your results as well | 23:27 |
| notmyname | hey look at that! all the gerrit patches for storage policies are cleaned up now | 23:28 |
| shri | sure… I'm going to play around with the object_chunk_size, client_chunk_size on the proxy-server side and disk_chunk_size, network_chunk_size on the object-server side. | 23:29 |
| *** elambert has quit IRC | 23:44 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!
