| portante | 1.8 is grizzly, 1.10 is havana (for swift at least, right notmyname?) | 00:00 |
|---|---|---|
| themadcanudist | 1 proxy node, 32 workers on the same server | 00:00 |
| themadcanudist | that seems to be fine | 00:00 |
| notmyname | portante: correct | 00:00 |
| portante | themadcanudist: run | 00:00 |
| portante | away | 00:00 |
| portante | :) | 00:00 |
| themadcanudist | lol | 00:00 |
| notmyname | and a newer version of swift will still be compatible | 00:00 |
| portante | sorry, just kidding | 00:00 |
| themadcanudist | hehe, i know | 00:00 |
| portante | so my guess is that the traffic of lo is high? | 00:01 |
| portante | can I assume you have separate storage nodes? | 00:01 |
| themadcanudist | yep | 00:01 |
| themadcanudist | for both | 00:01 |
| themadcanudist | high, busy, but like 30 mbits | 00:01 |
| themadcanudist | 60 mbits at peack | 00:01 |
| themadcanudist | on lo | 00:02 |
| themadcanudist | but still | 00:02 |
| themadcanudist | that's not *crazy* | 00:02 |
| notmyname | themadcanudist: so first off, yes. what you are seeing is known. the bad news is that there are 2 basic ways around it. 1) add more keystone servers (this is what mercado libre did). 2) don't use keystone. this is what most swift deployments do (because of what you see right now) | 00:02 |
| notmyname | (just getting that out there so everything else from now on can be good news for you) | 00:02 |
| themadcanudist | hmmm, add more keystone servers? as in workers that the proxy servers can contact in some round robin fashion? | 00:03 |
| notmyname | "don't use keystone" probably is a non-starter for you, so let's figure out a way to do more with what you have | 00:03 |
| themadcanudist | exactly ;) i must use it | 00:03 |
| themadcanudist | is this "fixed" in future versions? | 00:03 |
| notmyname | sort of. well, it's a little better. I don't know that "fixed" is the right word. keystone was able to use swift's memcache ring strategy starting in grizzly. that's a good improvement. anything else as to their indexing or whatever they are doing internally, I don't knwo | 00:04 |
| themadcanudist | ok. the "multi-keyston server strategy" how does that look? how do the proxies utilize it? | 00:05 |
| notmyname | hmmm....that's a good question. /me is open to suggestions here | 00:05 |
| notmyname | unfortunately, according to instagram, the mercado libre guys are out salsa dancing right now ;-) | 00:06 |
| *** rando has joined #openstack-swift | 00:06 | |
| *** mdonohoe has quit IRC | 00:08 | |
| *** shri has quit IRC | 00:18 | |
| openstackgerrit | John Dickinson proposed a change to openstack/swift: Added docs about the swift_source log field https://review.openstack.org/71163 | 00:23 |
| notmyname | git st | 00:23 |
| notmyname | I don't suppose that works here | 00:24 |
| torgomatic | I can tell you about my local changes if you want | 00:24 |
| torgomatic | they're mostly rubbish | 00:24 |
| *** matsuhashi has joined #openstack-swift | 00:25 | |
| notmyname | themadcanudist: I'm not really sure where to tell you to look for optimizations in keystone, because I don't know where it's internal bottlenecks are. you may try asking ayoung in #openstack-dev, but he may simply tell you to upgrade (I'm hoping it can be something more helpful than that) | 00:26 |
| * themadcanudist nods | 00:26 | |
| themadcanudist | thank you! | 00:26 |
| notmyname | themadcanudist: I want you to have a happy swift cluster. you're always free to ask questions in here about swift :-) | 00:27 |
| *** shri has joined #openstack-swift | 00:27 | |
| jokke__ | notmyname: \o | 00:27 |
| jokke__ | how are you? Keeping busy it seems | 00:27 |
| notmyname | swifterdarrell: https://review.openstack.org/71163 should address your comment on the log docs, and it adds some more info about swift_source | 00:27 |
| notmyname | jokke__: indeed (aren't we all?) :-) | 00:28 |
| themadcanudist | i *very much* appreciate it! | 00:28 |
| *** jokke__ is now known as jokke_ | 00:30 | |
| notmyname | themadcanudist: sorry you got those responses :-( | 00:30 |
| themadcanudist | no problem | 00:31 |
| themadcanudist | they're good responses | 00:31 |
| themadcanudist | sometimes things are just contingent | 00:31 |
| themadcanudist | =) | 00:31 |
| *** Dieterbe has joined #openstack-swift | 00:34 | |
| notmyname | Dieterbe: you aren't using keystone are you? | 00:36 |
| notmyname | on your swift cluster | 00:36 |
| Dieterbe | right | 00:36 |
| Dieterbe | tempauth | 00:36 |
| notmyname | ah ok | 00:36 |
| Dieterbe | why? | 00:37 |
| notmyname | Dieterbe: right before you came in, themadcanudist was asking about some scaling issues with keystone in front of swift | 00:37 |
| Dieterbe | ha. i've had my fair share of issues with swift, but nothing keystone related :] | 00:38 |
| Dieterbe | afk* | 00:38 |
| notmyname | themadcanudist: if you end up upgrading swift too, then you can take a look at https://swiftstack.com/blog/2013/12/20/upgrade-openstack-swift-no-downtime/ | 00:39 |
| themadcanudist | wonderful! | 00:39 |
| notmyname | I'm out for most of the evening. later, everyone | 00:41 |
| *** mkollaro has quit IRC | 00:41 | |
| jokke_ | take care | 00:41 |
| themadcanudist | peac | 00:42 |
| *** jokke_ has quit IRC | 00:43 | |
| *** jokke_ has joined #openstack-swift | 00:44 | |
| *** sungju has joined #openstack-swift | 00:47 | |
| *** jokke_ has quit IRC | 00:47 | |
| *** jokke_ has joined #openstack-swift | 00:48 | |
| openstackgerrit | Steve Kowalik proposed a change to openstack/swift: Drop nose, switching to testtools and testr https://review.openstack.org/65014 | 00:55 |
| *** nosnos has joined #openstack-swift | 00:57 | |
| *** nosnos has quit IRC | 01:01 | |
| *** nosnos has joined #openstack-swift | 01:01 | |
| *** dmorita has joined #openstack-swift | 01:26 | |
| *** themadcanudist has left #openstack-swift | 01:28 | |
| *** shri has quit IRC | 01:42 | |
| openstackgerrit | A change was merged to openstack/swift: Let swift-object-info skip etag verification https://review.openstack.org/67296 | 01:55 |
| *** jasondotstar has joined #openstack-swift | 02:22 | |
| zaitcev | Mother of god | 02:53 |
| zaitcev | portante: I started writing tests for the code that I copied verbatim from Gluster and its FULL OF BUGS. | 02:53 |
| portante | zaitcev: gluster is full of bugs? | 03:11 |
| portante | or i should say the gluster-swift code is full of bugs? | 03:11 |
| portante | can you say more? | 03:11 |
| zaitcev | portante: no, just the UFO | 03:11 |
| portante | so what kind of bugs are you seeing? | 03:12 |
| zaitcev | Although truth to be told Swift Devguide is kind of undewhelming, too. | 03:12 |
| portante | swift devguide in the gluster-swift tree? | 03:12 |
| zaitcev | portante: the first was the limit. You know, GET /v1/a/c?limit=3 | 03:12 |
| portante | okay, and does that fail in gluster-swift unmodified? | 03:13 |
| zaitcev | if c/ only contains 2 objects, the filter_limit() tracebacks with index outside of array | 03:13 |
| portante | perhaps we should discuss this in the gluster-swift channel instead? | 03:13 |
| zaitcev | Well. I need to refresh my git trees for UFO, see maybe Kaleb or whoever already had all it patched up. | 03:14 |
| portante | what is the tree you are using? | 03:14 |
| portante | https://github.com/gluster/gluster-swift.git | 03:15 |
| portante | ? | 03:15 |
| zaitcev | url = git://github.com/gluster/gluster-swift.git | 03:15 |
| portante | nobody uses the UFO name anymore, it has no meaning, btw | 03:15 |
| portante | so did you run all the functional tests in your tree? | 03:16 |
| portante | or is this something that is not tested? | 03:16 |
| portante | and can we switch to the #gluster-swift channel in freenode so the folks in Bangalore can join the discussion? | 03:17 |
| portante | zaitcev? | 03:17 |
| zaitcev | No, I did not run functional tests. | 03:17 |
| zaitcev | Not against the in-memory back-ends, I mean. | 03:17 |
| portante | zaitcev: can we talk about the base gluster-swift issues in the #gluster-swift channel and the in-memory ones here? | 03:18 |
| zaitcev | portante: sure. But I'll probably need to take it all back. << >>. I'm looking at the semi-recent tree and the code is nothing like what I picked up. | 03:19 |
| zaitcev | portante: My point was that your idea to have tests was brilliant. | 03:19 |
| portante | thanks | 03:22 |
| portante | let's hope it makes a material difference | 03:22 |
| *** themadcanudist has joined #openstack-swift | 03:50 | |
| openstackgerrit | Pete Zaitcev proposed a change to openstack/swift: Pluggable Back-ends for account and container servers https://review.openstack.org/47713 | 04:15 |
| *** jd__ has quit IRC | 04:21 | |
| *** jd__ has joined #openstack-swift | 04:22 | |
| *** byeager has joined #openstack-swift | 04:48 | |
| *** marcusvrn has quit IRC | 04:48 | |
| *** ppai has joined #openstack-swift | 04:54 | |
| *** nshaikh has joined #openstack-swift | 05:16 | |
| *** igor__ has joined #openstack-swift | 05:37 | |
| *** byeager has quit IRC | 05:58 | |
| *** krast has joined #openstack-swift | 06:14 | |
| *** igor__ has quit IRC | 06:21 | |
| *** jasondotstar has quit IRC | 06:22 | |
| *** mmcardle has joined #openstack-swift | 06:22 | |
| *** zaitcev has quit IRC | 06:23 | |
| *** mmcardle has quit IRC | 06:26 | |
| *** nshaikh has quit IRC | 06:27 | |
| *** odyssey4me has joined #openstack-swift | 06:39 | |
| *** SkyRocknRoll__ has quit IRC | 07:55 | |
| *** haomaiwa_ has quit IRC | 08:28 | |
| *** haomaiwang has joined #openstack-swift | 08:28 | |
| *** haomai___ has joined #openstack-swift | 08:32 | |
| *** haomaiwang has quit IRC | 08:36 | |
| *** krtaylor has quit IRC | 08:47 | |
| *** mlipchuk has joined #openstack-swift | 08:49 | |
| *** krtaylor has joined #openstack-swift | 08:49 | |
| openstackgerrit | Christian Schwede proposed a change to openstack/swift: Add tests for swift-ring-builder https://review.openstack.org/71106 | 08:52 |
| *** matsuhashi has quit IRC | 08:54 | |
| *** krtaylor has quit IRC | 08:55 | |
| *** matsuhashi has joined #openstack-swift | 08:55 | |
| *** nacim has joined #openstack-swift | 08:55 | |
| *** krtaylor has joined #openstack-swift | 08:56 | |
| *** psharma has joined #openstack-swift | 08:57 | |
| *** mkollaro has joined #openstack-swift | 09:05 | |
| *** fbo_away is now known as fbo | 09:13 | |
| *** thomaschaaf has joined #openstack-swift | 09:19 | |
| *** odyssey4me has quit IRC | 09:25 | |
| *** mmcardle has joined #openstack-swift | 09:31 | |
| thomaschaaf | Hey I am running 1.10.0 on all my servers (5 total) and the network traffic between them being mostly idle is almost 10 MB/s how could I debug this? Is this expected behaviour? | 09:32 |
| thomaschaaf | 10 MB/s per server | 09:32 |
| thomaschaaf | this is the recon output: https://gist.github.com/thomaschaaf/d8367a1ad8865c634a49 | 09:33 |
| *** odyssey4me has joined #openstack-swift | 09:34 | |
| *** grapsus_1 has quit IRC | 09:52 | |
| *** grapsus_ has joined #openstack-swift | 09:54 | |
| *** xga has joined #openstack-swift | 09:57 | |
| *** themadcanudist has quit IRC | 10:00 | |
| *** jasondotstar has joined #openstack-swift | 10:00 | |
| *** sphoorti has joined #openstack-swift | 10:08 | |
| sphoorti | I have swift-keystone running on a vm. I am trying to run this command "swift -v -V 2.0 -A http://192.168.1.16:5000/v2.0/ -U admin -K password stat" but i get the following error : "Endpoint for object-store not found - have you specified a region?" . This command however works fine locally from the vm. What possibly could be going wrong? | 10:15 |
| saurabh_ | sphoorti: execute 'keystone catalog' is there any entry for Service: object-store | 10:18 |
| openstackgerrit | Chmouel Boudjnah proposed a change to openstack/python-swiftclient: Add missing backslash. https://review.openstack.org/71224 | 10:19 |
| chmouel | the easiest commit to review ever ^ | 10:19 |
| sphoorti | saurabh_: i see only service :- identity | 10:21 |
| sphoorti | and object store too saschpe_ | 10:22 |
| sphoorti | saurabh_: | 10:22 |
| openstackgerrit | Christian Schwede proposed a change to openstack/swift: Add tests for swift-ring-builder https://review.openstack.org/71106 | 10:23 |
| saurabh_ | sphoorti: execute 'keystone endpoint-list' is there any entry for object-stor 's IP and port | 10:24 |
| sphoorti | saurabh_: I do see Auth urls | 10:25 |
| saurabh_ | sphoorti: any entry for swift endpoint | 10:26 |
| saurabh_ | ? | 10:26 |
| saurabh_ | like http://IP:8080 | 10:26 |
| saurabh_ | v1/AUTH_$(tenant_id)s | 10:26 |
| sphoorti | yes saurabh_ | 10:27 |
| openstackgerrit | Christian Schwede proposed a change to openstack/swift: Add tests for swift-ring-builder https://review.openstack.org/71106 | 10:31 |
| sphoorti | saurabh_: ? | 10:41 |
| *** nosnos has quit IRC | 10:53 | |
| *** saurabh_ has quit IRC | 10:59 | |
| *** jjw has joined #openstack-swift | 11:00 | |
| *** jasondotstar has quit IRC | 11:00 | |
| *** jjw is now known as Guest96079 | 11:01 | |
| *** ben_roeder has joined #openstack-swift | 11:02 | |
| *** joeljwright has joined #openstack-swift | 11:07 | |
| *** saurabh_ has joined #openstack-swift | 11:10 | |
| *** saurabh_ has joined #openstack-swift | 11:10 | |
| *** jasondotstar has joined #openstack-swift | 11:13 | |
| chmouel | hahah the one char commit failed | 11:13 |
| openstackgerrit | Chmouel Boudjnah proposed a change to openstack/python-swiftclient: Add missing backslash. https://review.openstack.org/71224 | 11:15 |
| *** ben_roeder has quit IRC | 11:27 | |
| *** mmcardle has quit IRC | 11:32 | |
| *** mmcardle has joined #openstack-swift | 11:36 | |
| *** ben_roeder has joined #openstack-swift | 11:48 | |
| *** dmorita has quit IRC | 11:55 | |
| *** igor_ has joined #openstack-swift | 11:57 | |
| *** ppai has quit IRC | 12:04 | |
| *** ben_roeder has quit IRC | 12:08 | |
| *** mkollaro has quit IRC | 12:11 | |
| *** bada has joined #openstack-swift | 12:22 | |
| *** bada_ has quit IRC | 12:24 | |
| *** bada_ has joined #openstack-swift | 12:30 | |
| *** bada has quit IRC | 12:31 | |
| *** jasondotstar has quit IRC | 12:41 | |
| *** saurabh_ has quit IRC | 12:44 | |
| *** mkerrin has joined #openstack-swift | 12:44 | |
| *** saurabh_ has joined #openstack-swift | 12:44 | |
| *** saurabh_ has joined #openstack-swift | 12:44 | |
| *** mmcardle has quit IRC | 12:50 | |
| *** Guest96079 has quit IRC | 12:51 | |
| *** xga has quit IRC | 13:02 | |
| *** thomaschaaf has quit IRC | 13:03 | |
| *** ppai has joined #openstack-swift | 13:04 | |
| *** xga has joined #openstack-swift | 13:08 | |
| *** ppai has quit IRC | 13:09 | |
| *** Trixboxer has joined #openstack-swift | 13:13 | |
| *** ppai has joined #openstack-swift | 13:26 | |
| *** igor_ has quit IRC | 13:27 | |
| *** mkollaro has joined #openstack-swift | 13:27 | |
| *** igor_ has joined #openstack-swift | 13:28 | |
| *** mmcardle has joined #openstack-swift | 13:29 | |
| openstackgerrit | Christian Schwede proposed a change to openstack/swift: Add tests for swift-ring-builder https://review.openstack.org/71106 | 13:31 |
| *** igor_ has quit IRC | 13:33 | |
| *** ppai has quit IRC | 13:46 | |
| *** zigo_ is now known as zigo | 13:50 | |
| *** igor_ has joined #openstack-swift | 13:58 | |
| *** igor__ has joined #openstack-swift | 14:00 | |
| *** igor_ has quit IRC | 14:03 | |
| *** sphoorti has quit IRC | 14:05 | |
| *** jasondotstar has joined #openstack-swift | 14:14 | |
| *** tongli has joined #openstack-swift | 14:16 | |
| *** ccorrigan has joined #openstack-swift | 14:18 | |
| ccorrigan | notmyname: ot - eventual consistency explained - - http://www.youtube.com/watch?v=s6EaoPMANQM | 14:23 |
| *** psharma has quit IRC | 14:31 | |
| joeljwright | Hi, I'm trying to develop a middleware for swift that needs to read the entire body of a PUT request | 14:39 |
| joeljwright | obviously I want to avoid reading (and buffering) the wsgi.input multiple times | 14:40 |
| joeljwright | so is there some best practice for doing this? | 14:40 |
| joeljwright | currently I'm wrapping the wsgi.input and performing my calculations when the proxy-server performs its reads, and using a WSGIContext to catch the response and perform actions on the result | 14:42 |
| joeljwright | any advice would be greatly appreciated | 14:42 |
| *** zaitcev has joined #openstack-swift | 14:48 | |
| *** ChanServ sets mode: +v zaitcev | 14:49 | |
| *** mlipchuk has quit IRC | 14:51 | |
| *** xga has quit IRC | 15:03 | |
| *** jergerber has joined #openstack-swift | 15:09 | |
| *** jergerber has quit IRC | 15:12 | |
| *** mlipchuk has joined #openstack-swift | 15:12 | |
| *** nacim_ has joined #openstack-swift | 15:13 | |
| *** nacim has quit IRC | 15:14 | |
| *** matsuhashi has quit IRC | 15:17 | |
| *** matsuhashi has joined #openstack-swift | 15:18 | |
| *** bvandenh has joined #openstack-swift | 15:19 | |
| *** peluse has quit IRC | 15:22 | |
| *** peluse has joined #openstack-swift | 15:22 | |
| *** xga has joined #openstack-swift | 15:26 | |
| *** byeager has joined #openstack-swift | 15:31 | |
| notmyname | joeljwright: do you need to read the whole body before you can pass it on to the underlying app? or can you simply do your calculation as you stream it through the middleware? | 15:33 |
| *** matsuhashi has quit IRC | 15:34 | |
| joeljwright | I only want to add a header, I don't want to allow/deny the request | 15:35 |
| dfg | joeljwright: to me sounds like you're on the right track. you can look at swift/common/middleware/proxy_logging.py as an example if you want | 15:36 |
| notmyname | joeljwright: do you have to add the header based on the full contents of the body? | 15:37 |
| notmyname | joeljwright: dfg: proxy_logging is a good place to look, especially to take advantage of eventlet posthooks for callbacks. but for a very simply middleware example that adds a header, take a look at catch_errors | 15:38 |
| joeljwright | notmyname: yes, I need to read the full content - I'm just calculating SHA1 and CRC32 atm to add them as headers | 15:38 |
| joeljwright | dfg: thanks for the hint, I'll take a look | 15:40 |
| notmyname | joeljwright: so that means you'll have to read the full body in the middleware, reset the stream (ie you'll have it spooled), then send it on with your new headers. that seems not good | 15:42 |
| dfg | isn't he just setting headers on the response? just like proxy_logging is? | 15:43 |
| notmyname | I thought it was being saved with the object | 15:45 |
| joeljwright | actually, I'm adding the headers afterwards with a separate POST | 15:45 |
| notmyname | so what is your middleware doing? | 15:46 |
| dfg | oh- turn post_as_copy off :) | 15:46 |
| joeljwright | I'm wrapping the wsgi.input so that the sha1 and crc32 are calculated ad the proxy server does its reads | 15:47 |
| notmyname | reads from what? the client or the cluster? | 15:47 |
| joeljwright | and using a WSGIContext to get the final result of the put | 15:47 |
| notmyname | and what do you do with the hashes once you have them? | 15:48 |
| joeljwright | an extra post from within the middleware to add the headers | 15:48 |
| joeljwright | I got the idea from the undelete middleware here https://github.com/swiftstack/swift_undelete | 15:49 |
| joeljwright | using a wsgicontext to keep hold of access to the wrapped wsgi.input and intercept the final response form the put | 15:49 |
| joeljwright | once the put has succeeded I know the data has all been read and the checksums are calculated, so I can make another request before returning the result of the put to the client | 15:50 |
| joeljwright | the second request is a post to add the headers to the object | 15:50 |
| notmyname | dfg: any chance you'll be working on unifying the POST data flows into just the fast-POST version, like we talked about last fall in Austin? | 15:51 |
| notmyname | joeljwright: ya, like dfg said, you'll want to turn off post_as_copy | 15:52 |
| *** pscheie has quit IRC | 15:54 | |
| notmyname | joeljwright: just curious, but why are you calculating the hashes server side? what are they used for and why does the client not have them? | 15:56 |
| *** mmcardle1 has joined #openstack-swift | 15:56 | |
| *** igor__ has quit IRC | 15:57 | |
| joeljwright | notmyname: we're generally interested in having the hashes, but don't have control over the clients used to upload the data | 15:58 |
| *** mmcardle has quit IRC | 15:59 | |
| joeljwright | we've been asked if we can add the data by some clients, but they still want to continue using apps like cloudberry/cyberduck to upload | 16:00 |
| *** mmcardle has joined #openstack-swift | 16:01 | |
| dfg | notmyname: like- me doing all the typing and stuff? no prob not :) we have a bunch of other stuff to do- somebody should do it though... | 16:01 |
| joeljwright | to be honest, this was as much an investigation of writing middlewares for me as it was a requirement | 16:01 |
| notmyname | I wonder if having a DiskFile to calculate and save those would be more efficient (just thinking aloud; paging portante). you'd do the math on each replica, but you'd also have it spread out to every storage node without having to coordinate a 2nd request from the proxy | 16:01 |
| *** mmcardle1 has quit IRC | 16:01 | |
| *** ChanServ sets mode: +v chmouel | 16:01 | |
| notmyname | dfg: ok. just curious | 16:02 |
| notmyname | dfg: and ya. somebody should do all that typing ;-) | 16:02 |
| *** igor_ has joined #openstack-swift | 16:03 | |
| *** mkollaro has quit IRC | 16:07 | |
| *** mmcardle has quit IRC | 16:08 | |
| *** igor_ has quit IRC | 16:08 | |
| *** mmcardle has joined #openstack-swift | 16:10 | |
| *** byeager has quit IRC | 16:13 | |
| joeljwright | notmyname: I'm not sure I understand that last suggestion | 16:15 |
| *** mmcardle has quit IRC | 16:20 | |
| openstackgerrit | Christian Schwede proposed a change to openstack/swift: Add tests for swift-ring-builder https://review.openstack.org/71106 | 16:20 |
| *** jasondotstar has quit IRC | 16:21 | |
| zaitcev | pushing the calcs out to DiskFile means that it's redundant (3x) but then it unloads proxy which is what we want most of the time | 16:25 |
| joeljwright | I think I'm starting to understand the DiskFile suggestion | 16:26 |
| joeljwright | thanks for all the help, I'll have a dig about in the code | 16:27 |
| chmouel | cschwede: about the test ringbuilder tests, wouldn't that work to just create the tempfile in create_sample_ring() directly? | 16:31 |
| *** jasondotstar has joined #openstack-swift | 16:31 | |
| *** mmcardle has joined #openstack-swift | 16:32 | |
| cschwede | chmouel: well, than we need to collect all the filenames to delete them in tearDown i think? | 16:33 |
| chmouel | cschwede: ah yeah right | 16:33 |
| chmouel | cschwede: it did come to mind at certain point and i'm not sure why i thought it may have been a good idea :) | 16:33 |
| cschwede | chmouel: well there are two ways to do it, and i think both are ok :) | 16:33 |
| cschwede | chmouel: i can change that also if you prefer ;-) | 16:34 |
| chmouel | cschwede: ah no nervermind, i don't want to nitpick :) | 16:34 |
| *** tdasilva has left #openstack-swift | 16:43 | |
| *** bvandenh has quit IRC | 16:46 | |
| *** nacim_ has quit IRC | 16:46 | |
| notmyname | torgomatic: are dynamic large objects setting the swift_source for any of the subrequests generated? | 16:50 |
| *** gyee has joined #openstack-swift | 16:56 | |
| *** hurricanerix has joined #openstack-swift | 16:59 | |
| *** tdasilva has joined #openstack-swift | 17:04 | |
| *** igor_ has joined #openstack-swift | 17:04 | |
| gholt | swifterdarrell: All the error log lines with no txn: are probably because of that bug you filed on background coroutines, right? | 17:05 |
| gholt | I mean, it's because you filed the bug, but due do that bug, and... oh nm | 17:05 |
| *** CoilDomain has joined #openstack-swift | 17:07 | |
| *** CoilDomain has left #openstack-swift | 17:07 | |
| *** igor_ has quit IRC | 17:09 | |
| briancline | morning -- is 1.12.0 currently planned to be the release to coincide with the 2014.1/Icehouse release prior to summit? | 17:10 |
| briancline | sorry, 1.13.0 | 17:10 |
| portante | notmyname: what's up, something about using DiskFile to calculate hashes, it does that on PUT already, not sure I followed the above well | 17:20 |
| *** odyssey4me2 has joined #openstack-swift | 17:23 | |
| *** odyssey4me has quit IRC | 17:24 | |
| *** odyssey4me2 is now known as odyssey4me | 17:24 | |
| *** mlipchuk has quit IRC | 17:37 | |
| torgomatic | notmyname: I have no idea, but I'd guess not | 17:44 |
| torgomatic | if the original code did, then mine will too, but otherwise... maybe? | 17:44 |
| *** bsdkurt has joined #openstack-swift | 17:45 | |
| *** jasondotstar has quit IRC | 17:46 | |
| notmyname | portante: joeljwright wants to compute some extra stuff on a write and save it with the data. seems that doing that in a diskfile could be an interesting avenue to explore | 17:51 |
| notmyname | torgomatic: I didn't see anything in the DLO patch, but I wanted to ask in case I missed something | 17:51 |
| notmyname | torgomatic: it probably should, but I'm noticing it now because I submitted a patch yesterday to document all that | 17:51 |
| notmyname | torgomatic: but I wouldn't (won't) hold up your current patch because of that. just something that should probably be added | 17:52 |
| portante | notmyname: yes | 17:54 |
| notmyname | portante: and I have the ulterior motive of pushing for that kind of functionality since I think that will result in better config designation of a particular DiskFile (like clayg's pluggable daemons method) ;-) | 17:55 |
| portante | =) | 17:57 |
| swifterdarrell | gholt: haha! I'm not sure... that could be a perverse side-effect of letting those laggard greenthreads "finish". I didn't specifically investigate that | 17:57 |
| * portante thinks, "smile and wave boys, just smile and wave." | 17:57 | |
| *** xga has quit IRC | 17:57 | |
| portante | o/ | 17:58 |
| * portante forgot the wave | 17:58 | |
| *** shri has joined #openstack-swift | 17:59 | |
| swifterdarrell | gholt: I went back and checked one of my smoking-gun log-grabs and sure enough, the final log message from the laggard greenthread, which complains about hitting Timeout (10s) even though that 3rd req cleared the object-server in ~1.8s, does not have a txid | 18:03 |
| *** gyee has quit IRC | 18:05 | |
| *** jasondotstar has joined #openstack-swift | 18:05 | |
| *** kris_h has joined #openstack-swift | 18:05 | |
| *** igor__ has joined #openstack-swift | 18:05 | |
| *** kris_h has quit IRC | 18:05 | |
| *** kris_h has joined #openstack-swift | 18:06 | |
| gholt | swifterdarrell: Oh good, I was hoping it was that and not something more sinister. | 18:07 |
| *** igor__ has quit IRC | 18:10 | |
| *** jokke__ has joined #openstack-swift | 18:11 | |
| *** jokke__ has quit IRC | 18:12 | |
| *** CoilDomain has joined #openstack-swift | 18:14 | |
| *** gvernik has joined #openstack-swift | 18:28 | |
| *** CoilDomain has left #openstack-swift | 18:31 | |
| *** bvandenh has joined #openstack-swift | 18:31 | |
| *** zul has quit IRC | 18:33 | |
| *** zul has joined #openstack-swift | 18:37 | |
| *** lpabon has joined #openstack-swift | 18:47 | |
| notmyname | swift team meeting in 10 minutes in #openstack-swift https://wiki.openstack.org/wiki/Meetings/Swift | 18:50 |
| notmyname | er..in #openstack-meeting | 18:50 |
| *** bvandenh has quit IRC | 18:51 | |
| *** mmcardle has quit IRC | 18:51 | |
| *** mmcardle1 has joined #openstack-swift | 18:51 | |
| *** joeljwright has quit IRC | 18:53 | |
| *** joeljwright has joined #openstack-swift | 18:53 | |
| notmyname | peluse: torgomatic: portante: I want to talk about the storage policy/EC status first up in the meeting | 18:53 |
| peluse | OK | 18:55 |
| portante | notmyname: okay | 18:55 |
| notmyname | creiht: did you add the item to the agenda about the asyncio lib for python2? or maybe portante? | 18:56 |
| portante | I did not add it, but I think it is worth talking about | 18:56 |
| notmyname | ya, absolutely. it was listed on the agenda under the swiftclient topic. I think that was wrong and moved it under creiht's topic of py3k support | 18:57 |
| *** dirk has joined #openstack-swift | 18:57 | |
| portante | k | 18:57 |
| notmyname | looks like we have quite a bit (potentially) to cover today | 18:58 |
| notmyname | swift meeting time in #openstack-meeting | 19:00 |
| *** tgohad has joined #openstack-swift | 19:02 | |
| *** bvandenh has joined #openstack-swift | 19:05 | |
| *** igor__ has joined #openstack-swift | 19:06 | |
| *** igor__ has quit IRC | 19:10 | |
| *** mfactor has joined #openstack-swift | 19:17 | |
| *** byeager has joined #openstack-swift | 19:26 | |
| *** hurrican_ has joined #openstack-swift | 19:28 | |
| *** hurrican_ has quit IRC | 19:29 | |
| *** mmcardle1 has quit IRC | 19:29 | |
| *** hurricanerix has quit IRC | 19:31 | |
| *** igor___ has joined #openstack-swift | 19:32 | |
| *** hurricanerix has joined #openstack-swift | 19:35 | |
| *** igor___ has quit IRC | 19:37 | |
| *** odyssey4me has quit IRC | 19:40 | |
| *** mfactor has quit IRC | 19:41 | |
| *** mfactor has joined #openstack-swift | 19:42 | |
| openstackgerrit | Tristan Cacqueray proposed a change to openstack/python-swiftclient: Port to python-requests https://review.openstack.org/69187 | 19:43 |
| portante | clayg: I should have put a smiley face on the end of that /me comment | 19:47 |
| clayg | portante: i could tell you were kidding around - it was funny | 19:50 |
| portante | :) | 19:51 |
| *** davidhadas has joined #openstack-swift | 19:56 | |
| *** joeljwright has quit IRC | 19:59 | |
| notmyname | bah. time limits | 20:01 |
| portante | clayg: I am thinking that a core dev sponsor will help with folks who are not cores to get their patches considered | 20:01 |
| portante | so a core dev could review it sufficiently to help notmyname prioritize it | 20:01 |
| chmouel | i think we need hyperthreading on sam | 20:01 |
| peluse | good meeting today notmyname | 20:01 |
| *** gvernik has quit IRC | 20:02 | |
| clayg | yeah I mean I think of it as a formalization of jump in IRC and be like "hey yeah so, i've been thinking this was a good idea -> link to review" | 20:02 |
| portante | torgomatic: how about we give you more machines? | 20:02 |
| notmyname | peluse: the meetings are as good as the participation. thanks for taking ec and storage policies updates :-) | 20:02 |
| torgomatic | portante: coffee machines? | 20:02 |
| portante | :) | 20:02 |
| portante | yes | 20:02 |
| portante | ! | 20:02 |
| portante | lol | 20:02 |
| notmyname | actually, I''m working on getting torgomatic some more machines (not for coffee) :-) | 20:02 |
| *** gvernik_ has joined #openstack-swift | 20:03 | |
| clayg | I think we should do meetings every week (notmyname is really the only one that has to ALL of them) - and he can dole out "sponsers" to patches that no-one has looked at | 20:03 |
| notmyname | I got a pile of hard drives to put in them yesterday | 20:03 |
| *** igor___ has joined #openstack-swift | 20:03 | |
| torgomatic | I don't think the sponsorship thing is particularly official right now; it's just an emergent behavior | 20:03 |
| zaitcev | It was v.informative meeting today. I did not know about gevent. | 20:03 |
| notmyname | torgomatic: ya | 20:03 |
| briancline | are there not folks currently who focus primarily on specific parts of swift? or is it currently a free-for-all? that may be part of the problem | 20:04 |
| notmyname | clayg: the same time slot on the other weeks is currently taken by "Neutron Advanced Services' Common requirements team meeting" | 20:04 |
| briancline | core folks, that is | 20:04 |
| portante | clayg: for example, I'd like to be able to say that before others spend time on a patch, it is verified to pass unit/func/probe and all tox stuff independent of jenkins | 20:04 |
| torgomatic | if some other core reviewer has looked at a patch, then it's probably easier to review for me because it'll at least have had some basic sanity checks | 20:04 |
| clayg | isn't there an openstack-meeting-alt? | 20:04 |
| notmyname | briancline: although there is some degree of specialization, there isn't too much siloing in swift | 20:04 |
| torgomatic | and if everyone observes the same thing, then we get the current situation where patches take a long time for the first +2, and then ZOOM they're done | 20:04 |
| *** dirk has quit IRC | 20:04 | |
| notmyname | clayg: and now also #openstack-meeting-3 | 20:04 |
| *** igor____ has joined #openstack-swift | 20:05 | |
| torgomatic | wait, so meeting, meeting-alt, and meeting-3? Can we have meeting-delta next? | 20:05 |
| briancline | I've heard #whatdayisit is also available | 20:05 |
| notmyname | clayg: oh, wait, that other meeting is in -alt | 20:05 |
| clayg | x-temp-url-key and x-temp-url-key-2 - these things happen man | 20:05 |
| briancline | notmyname: well, I'd hate to think of it as siloing, as that tends to have a negative connotation | 20:06 |
| torgomatic | clayg: yeah, but I'm not adding temp-url-key-three and -key-cuatro next :) | 20:06 |
| clayg | yes! | 20:06 |
| notmyname | briancline: ya, I know. | 20:06 |
| * notmyname will look into moving to a weekly meeting | 20:07 | |
| *** igor___ has quit IRC | 20:07 | |
| chmouel | torgomatic: :) | 20:08 |
| notmyname | /lunch | 20:09 |
| *** hurricanerix has quit IRC | 20:09 | |
| chmouel | dinner time! | 20:09 |
| briancline | as a somewhat separate topic, to be honest I've seen non-core folks with legitimate concerns over patches (echoing what I would have written in some cases) that get hurried in and merged because of specific attention from core folks | 20:09 |
| briancline | certainly not the case for the majority | 20:09 |
| briancline | just another thing to consider as we think about this | 20:09 |
| briancline | perhaps it's not *because* of core attention, but it gives the appearance that voices get lost in the shuffle or in the dash to get something implemented | 20:12 |
| *** kragniz has quit IRC | 20:16 | |
| zaitcev | Bah, I think the only time it went pear shaped was when Alex busted DB cursors and thus Havana GA went out with the "DB Locked" tracebacks under any decent load. | 20:17 |
| zaitcev | Otherwise it's just the SJF that any printer spooler does. | 20:17 |
| zaitcev | I learned to be philosophical about it while sitting on 1 patch for a few months. | 20:18 |
| *** gvernik_ has quit IRC | 20:18 | |
| briancline | just a thought. luckily we can be smarter than the print spooler ;) | 20:21 |
| *** mfactor has quit IRC | 20:21 | |
| *** hurricanerix has joined #openstack-swift | 20:28 | |
| *** joeljwright has joined #openstack-swift | 20:30 | |
| *** dirk has joined #openstack-swift | 20:30 | |
| portante | briancline: core devs are human beings, and have preferences and leanings, just like the folks submitting patches | 20:31 |
| portante | it is the nature of the community | 20:31 |
| portante | no community is fair | 20:32 |
| portante | we are fooling ourselves to think that we are any different | 20:32 |
| portante | we just need to acknowledge that, and point things out, be as open as possible, and available for folks to make their case for their patches | 20:32 |
| portante | notmyname, clayg, torgomatic, chmouel, zaitcev, briancline: not that I don't see anybody else in #openstack-meeting | 20:34 |
| * zaitcev is very confused by the double negative | 20:37 | |
| *** joeljwright has quit IRC | 20:38 | |
| torgomatic | yeah, it's true... if it's a patch I have a use case for, I'm much (much!) more likely to help it along than if it's something I'm indifferent to | 20:41 |
| glange | some patches address production issues and thus get more immediate attention | 20:44 |
| *** _bluev has joined #openstack-swift | 20:45 | |
| *** _bluev has quit IRC | 20:46 | |
| *** shri has quit IRC | 20:56 | |
| *** gvernik_ has joined #openstack-swift | 21:02 | |
| peluse | zaitcev: wow, that patch you mentioed is a monster :) I will indeed take a look for what its worth though (there are some areas there I'm not super familiar with) | 21:11 |
| zaitcev | peluse: I started by splitting it, but then some complained that it's hard to review without seeing how all parts fit together. | 21:12 |
| peluse | yup, I can see that view for sure | 21:12 |
| *** ben_roeder has joined #openstack-swift | 21:15 | |
| *** tongli has quit IRC | 21:18 | |
| *** _bluev has joined #openstack-swift | 21:19 | |
| *** bvandenh has quit IRC | 21:21 | |
| *** kris_h has quit IRC | 21:30 | |
| *** kris_h has joined #openstack-swift | 21:30 | |
| *** kris_h has quit IRC | 21:31 | |
| *** joeljwright has joined #openstack-swift | 21:35 | |
| *** joeljwright has quit IRC | 21:39 | |
| *** lpabon has quit IRC | 21:40 | |
| *** wer has joined #openstack-swift | 21:43 | |
| notmyname | chmouel: are you seeing https://twitter.com/chmouel/status/431178123649249280 in swift? if so, can you document some examples for me (maybe send privately)? I don't want this to happen in swift | 21:44 |
| * notmyname is worried about it | 21:44 | |
| *** ben_roeder has quit IRC | 21:48 | |
| *** _bluev has quit IRC | 21:50 | |
| *** _bluev has joined #openstack-swift | 21:52 | |
| *** jasondotstar has quit IRC | 21:56 | |
| *** jasondotstar has joined #openstack-swift | 21:57 | |
| *** gvernik_ has quit IRC | 21:58 | |
| *** nacim has joined #openstack-swift | 22:03 | |
| *** nacim has quit IRC | 22:07 | |
| *** nacim has joined #openstack-swift | 22:08 | |
| *** nacim has quit IRC | 22:12 | |
| *** tgohad has quit IRC | 22:12 | |
| *** dirk has quit IRC | 22:15 | |
| openstackgerrit | A change was merged to openstack/swift-bench: Sync with global requirements https://review.openstack.org/67064 | 22:16 |
| *** Trixboxer has quit IRC | 22:23 | |
| *** jergerber has joined #openstack-swift | 22:25 | |
| *** shri has joined #openstack-swift | 22:25 | |
| creiht | notmyname: sorry I'm on vacation this week... chmouel added it for me | 22:25 |
| notmyname | creiht: ah, ok. have fun on your vacation :-) | 22:30 |
| creiht | hehe | 22:35 |
| creiht | fwiw, I also think there is a lot more to the question of python3 support than just eventlet | 22:35 |
| creiht | but maybe we should just have a session about it at the summit | 22:36 |
| creiht | or we just choose to port to go instead ;) | 22:36 |
| *** joeljwright has joined #openstack-swift | 22:36 | |
| *** joeljwright has quit IRC | 22:40 | |
| chmouel | notmyname: oh not at all | 22:40 |
| chmouel | i think we are doing a very good job about being relax | 22:41 |
| portante | creiht: there was some talk about possibly switching to gevent first, and then ride that to py3 | 22:41 |
| creiht | portante: well that assumes that gevent ever actually gets to py3 compat | 22:42 |
| creiht | it has been stalled for quite a while | 22:42 |
| creiht | at least the last time that I checked | 22:42 |
| creiht | in my mind, the biggest concern is are we shooting for py2.x and 3 compat, or just moving to 3 | 22:42 |
| chmouel | creiht: this one? https://github.com/fantix/gevent | 22:42 |
| creiht | shooting for both is very hairy in my estimation | 22:42 |
| chmouel | creiht: what do you think about asyncio/tulip? | 22:43 |
| portante | my take is that we can't leap to py3 and support existing distributions | 22:43 |
| portante | we need a transition period | 22:43 |
| * creiht needs to get back to vacationing | 22:44 | |
| creiht | but I'll submit a proposal for the next summit | 22:44 |
| portante | good for you, and the creepers | 22:44 |
| chmouel | creiht: proposals for python3 ? | 22:44 |
| creiht | because I think it warrants a much deeper discussion | 22:44 |
| creiht | chmouel: yeah | 22:44 |
| creiht | I also would like to have a focused plan rather than these drive by python3 patches that may or may not help down the road | 22:45 |
| chmouel | it would be a nice thing to have both but I think what victor is trying to do is to easy the migration with the trollius project | 22:45 |
| creiht | chmouel: yeah I read some of that, but I'm also cautious | 22:45 |
| portante | creiht: agreed | 22:45 |
| creiht | chmouel: last I heard quido talk about tulip, is that it would never work on python 2.x | 22:45 |
| portante | it is still going to be a fair amount of work to do that transition | 22:46 |
| chmouel | creiht: ah he changed his opinions on it | 22:46 |
| creiht | and having a different syntax isn't really compat | 22:46 |
| chmouel | he was endorsing it in his talks | 22:46 |
| creiht | chmouel: ahh well that is good to know | 22:46 |
| chmouel | and may even plan to make it easier for p python2 | 22:46 |
| * creiht goes back to vacation now :) | 22:46 | |
| chmouel | ah | 22:46 |
| chmouel | i go to sleep :) | 22:46 |
| creiht | portante: oh, I have moved on to bigger and better things :) | 22:47 |
| chmouel | creiht: thanks for jumping in | 22:47 |
| creiht | like warhammer 40k figures :) | 22:47 |
| creiht | oh but the bigger issue will be all the unicode handling and changing to bytestreams | 22:47 |
| creiht | that seems to be the most difficult to me | 22:47 |
| creiht | or at a minimum, the least understood | 22:47 |
| *** jasondotstar has quit IRC | 22:48 | |
| chmouel | yeah that part is going to be a PITA | 22:48 |
| portante | creiht: if you have creepers walking down the streets of Austin, we are NOT going to have the hackathon there! | 22:49 |
| peluse | creepers? | 22:51 |
| creiht | lol | 22:51 |
| creiht | http://minecraft.gamepedia.com/Creeper | 22:51 |
| creiht | peluse: -^ | 22:51 |
| peluse | ahhh | 22:51 |
| portante | and yes, I am a scaredy cat | 22:51 |
| openstackgerrit | Greg Lange proposed a change to openstack/swift: Fix a couple ifs in account quota middleware. https://review.openstack.org/71405 | 22:52 |
| *** _bluev has quit IRC | 22:54 | |
| openstackgerrit | A change was merged to openstack/swift: Move all DLO functionality to middleware https://review.openstack.org/63326 | 22:59 |
| openstackgerrit | Zack Feldstein proposed a change to openstack/swift: Implements checking ring file existence on recon https://review.openstack.org/71408 | 22:59 |
| *** _bluev has joined #openstack-swift | 23:00 | |
| openstackgerrit | Zack Feldstein proposed a change to openstack/swift: Implements checking ring file existence on recon https://review.openstack.org/71408 | 23:03 |
| *** byeager has quit IRC | 23:07 | |
| *** zackf has joined #openstack-swift | 23:13 | |
| *** byeager has joined #openstack-swift | 23:13 | |
| openstackgerrit | Samuel Merritt proposed a change to openstack/swift: Make ranges_for_length return (first, last) pairs https://review.openstack.org/63327 | 23:15 |
| *** byeager has quit IRC | 23:19 | |
| *** zackf has quit IRC | 23:19 | |
| openstackgerrit | Samuel Merritt proposed a change to openstack/swift: Ensure swift.source is set for DLO/SLO requests https://review.openstack.org/71415 | 23:24 |
| *** kragniz has joined #openstack-swift | 23:31 | |
| *** joeljwright has joined #openstack-swift | 23:37 | |
| *** joeljwright has quit IRC | 23:42 | |
| clayg | so setuptools entry_points for console_scripts were "too slow" -> https://github.com/openstack-dev/pbr/commit/8e58c2fa58fd1aa6f9985dcb4e210508a73e1df7 | 23:46 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!