| notmyname | still here? | 00:03 |
|---|---|---|
| zaitcev | Why is it that I suspect you mean clayg | 00:04 |
| notmyname | nope. I meant that for someone in a whole different server | 00:04 |
| zaitcev | I am laughing. | 00:04 |
| *** jrichli has joined #openstack-swift | 00:04 | |
| notmyname | :-) | 00:04 |
| *** rmcall has quit IRC | 00:05 | |
| notmyname | so I just finished a rough draft of a blog post, and I'm looking for one of our marketing people | 00:05 |
| notmyname | whew. now that's off my plate, I can do some of the other stuff that's been piling up | 00:06 |
| *** breitz has quit IRC | 00:08 | |
| *** breitz has joined #openstack-swift | 00:08 | |
| *** rmcall has joined #openstack-swift | 00:09 | |
| *** zaitcev has left #openstack-swift | 00:10 | |
| *** rmcall has quit IRC | 00:10 | |
| *** zaitcev has joined #openstack-swift | 00:10 | |
| *** ChanServ sets mode: +v zaitcev | 00:10 | |
| *** thumpba has joined #openstack-swift | 00:20 | |
| *** rmcall has joined #openstack-swift | 00:21 | |
| *** thumpba_ has joined #openstack-swift | 00:21 | |
| *** asettle has quit IRC | 00:24 | |
| *** thumpba has quit IRC | 00:24 | |
| *** david-lyle has quit IRC | 00:32 | |
| *** theanalyst has quit IRC | 00:39 | |
| *** theanalyst has joined #openstack-swift | 00:43 | |
| *** zhill has quit IRC | 00:56 | |
| openstackgerrit | Janie Richling proposed openstack/swift: WIP - working on the encryption feature. https://review.openstack.org/157907 | 01:09 |
| *** fanyaohong has quit IRC | 01:09 | |
| *** asettle has joined #openstack-swift | 01:18 | |
| asettle | Hey notmyname - could you relink me to the design summit page you guys had up? :) | 01:18 |
| notmyname | asettle: the etherpad of ideas for summit sessions? | 01:19 |
| asettle | That's the one! | 01:19 |
| notmyname | it's in the topic | 01:19 |
| asettle | Ha. You guys are organised. | 01:19 |
| asettle | Most people just have illiterate comments in the topic | 01:19 |
| *** david-lyle has joined #openstack-swift | 01:20 | |
| zaitcev | Looks like we fixed swift --os-cacert after 2.1.0 some time. BUT now the requests says: | 01:34 |
| zaitcev | /usr/lib/python2.7/site-packages/requests/packages/urllib3/connection.py:251: SecurityWarning: Certificate has no `subjectAltName`, falling back to check for a `commonName` for now. This feature is being removed by major browsers and deprecated by RFC 2818. | 01:34 |
| zaitcev | KILL ALL HUMANS | 01:35 |
| *** rmcall has quit IRC | 01:38 | |
| openstackgerrit | Tim Burke proposed openstack/python-swiftclient: Stop throwing UnicodeEncodeErrors when deleting https://review.openstack.org/176945 | 01:46 |
| *** jrichli has quit IRC | 01:52 | |
| *** asettle has quit IRC | 01:57 | |
| *** wer has joined #openstack-swift | 02:04 | |
| *** pberis has joined #openstack-swift | 02:27 | |
| *** david-lyle has quit IRC | 02:30 | |
| *** david-lyle has joined #openstack-swift | 02:31 | |
| *** david-lyle has quit IRC | 02:36 | |
| *** gyee has quit IRC | 02:53 | |
| *** pberis has quit IRC | 02:58 | |
| *** david-lyle has joined #openstack-swift | 03:32 | |
| *** asettle has joined #openstack-swift | 03:32 | |
| *** ho has joined #openstack-swift | 04:00 | |
| *** thumpba_ has quit IRC | 04:10 | |
| *** asettle has quit IRC | 04:30 | |
| *** itlinux has quit IRC | 04:40 | |
| openstackgerrit | Prashanth Pai proposed openstack/swift: WIP: new attempt at single-process https://review.openstack.org/159285 | 04:40 |
| notmyname | I added the release notes for Swift https://wiki.openstack.org/wiki/ReleaseNotes/Kilo | 04:58 |
| *** Trixboxer has quit IRC | 05:00 | |
| notmyname | meeting tomorrow will be to go over the summit sessions. I'll start formally scheduling them later this week | 05:00 |
| notmyname | I've rearranged them on the etherpad into "fishbowl" and "working" sessions (with the current overflow listed at the bottom) | 05:01 |
| notmyname | link is in the channel topic | 05:01 |
| zaitcev | ok | 05:07 |
| *** zaitcev has quit IRC | 05:07 | |
| ho | notmyname: thanks for the info! | 05:12 |
| *** SkyRocknRoll has joined #openstack-swift | 05:17 | |
| mattoliverau | notmyname: nice work :) | 05:19 |
| notmyname | just my job :-) | 05:20 |
| *** asettle has joined #openstack-swift | 05:20 | |
| *** yuan has quit IRC | 05:35 | |
| *** silor has joined #openstack-swift | 05:42 | |
| *** ppai has joined #openstack-swift | 06:07 | |
| *** pcaruana has joined #openstack-swift | 06:09 | |
| *** ho has quit IRC | 06:11 | |
| *** zul has joined #openstack-swift | 06:17 | |
| *** proteusguy has quit IRC | 06:24 | |
| *** zul has quit IRC | 06:24 | |
| *** zul has joined #openstack-swift | 06:25 | |
| *** proteusguy has joined #openstack-swift | 06:49 | |
| openstackgerrit | Prashanth Pai proposed openstack/swift: Fix incorrect passing of file object to fsync() https://review.openstack.org/178528 | 06:53 |
| *** knl has joined #openstack-swift | 07:06 | |
| *** jordanP has joined #openstack-swift | 07:19 | |
| *** zul has quit IRC | 07:32 | |
| *** zul has joined #openstack-swift | 07:32 | |
| *** proteusguy_ has joined #openstack-swift | 07:54 | |
| *** proteusguy_ has quit IRC | 07:55 | |
| *** jeblair has quit IRC | 07:56 | |
| *** proteusguy has quit IRC | 07:56 | |
| *** jeblair has joined #openstack-swift | 07:56 | |
| *** jistr has joined #openstack-swift | 07:58 | |
| *** zhill has joined #openstack-swift | 08:08 | |
| *** zhill has quit IRC | 08:09 | |
| *** zul has quit IRC | 08:09 | |
| *** jbeusen_ has joined #openstack-swift | 08:23 | |
| *** jbeusen_ has quit IRC | 08:24 | |
| *** jbeusen_ has joined #openstack-swift | 08:25 | |
| *** zul has joined #openstack-swift | 08:25 | |
| *** acoles_away is now known as acoles | 08:26 | |
| *** jeblair has quit IRC | 08:33 | |
| *** jbeusen_ has quit IRC | 08:43 | |
| *** jbeusen_ has joined #openstack-swift | 08:44 | |
| *** jbeusen_ has quit IRC | 08:48 | |
| *** jbeusen_ has joined #openstack-swift | 08:49 | |
| *** jeblair has joined #openstack-swift | 08:51 | |
| *** jbeusen_ has quit IRC | 08:53 | |
| *** zul has quit IRC | 09:27 | |
| *** krykowski has joined #openstack-swift | 09:35 | |
| *** zul has joined #openstack-swift | 09:51 | |
| *** minwoob has quit IRC | 09:53 | |
| *** ppai has quit IRC | 09:57 | |
| *** jeblair has quit IRC | 10:03 | |
| *** jeblair has joined #openstack-swift | 10:03 | |
| *** mtreinish has quit IRC | 10:07 | |
| *** mwheckmann has quit IRC | 10:13 | |
| *** zul has quit IRC | 10:19 | |
| *** plindgre has joined #openstack-swift | 10:53 | |
| plindgre | hi guys, total noob here. just having a question regarding object storage | 10:53 |
| plindgre | im wondering, in an object storage in an open stack environment | 10:54 |
| plindgre | would a users cloud machines local storage be one object in the object storage | 10:54 |
| plindgre | or would all his individual files be objects in the object storage? | 10:55 |
| plindgre | cause from what i understand object strage is basically a database with blob files with an "orm" / "dao" wrapped around it. im just wondering how the implementation would look like | 10:56 |
| plindgre | would there be one object representing one users drive or would there be several individual objects representing the users individual files? | 10:56 |
| *** a1|away is now known as JelleB | 10:58 | |
| *** krykowski has quit IRC | 11:03 | |
| *** krykowski_ has joined #openstack-swift | 11:04 | |
| *** knl has quit IRC | 11:11 | |
| *** geaaru has joined #openstack-swift | 11:20 | |
| *** lpabon has joined #openstack-swift | 11:32 | |
| *** aix has joined #openstack-swift | 11:35 | |
| *** krykowski_ has quit IRC | 11:38 | |
| *** cdelatte has joined #openstack-swift | 11:42 | |
| *** delattec has joined #openstack-swift | 11:42 | |
| *** delatte has joined #openstack-swift | 11:44 | |
| *** krykowski has joined #openstack-swift | 11:45 | |
| *** cdelatte has quit IRC | 11:47 | |
| *** delattec has quit IRC | 11:47 | |
| *** cdelatte has joined #openstack-swift | 11:47 | |
| *** delatte has quit IRC | 11:52 | |
| *** cdelatte has quit IRC | 11:52 | |
| *** haomaiwa_ has quit IRC | 11:52 | |
| *** zul has joined #openstack-swift | 11:53 | |
| *** krykowski has quit IRC | 12:03 | |
| *** krykowski has joined #openstack-swift | 12:03 | |
| *** openstackgerrit has quit IRC | 12:07 | |
| *** openstackgerrit has joined #openstack-swift | 12:07 | |
| *** krykowski has quit IRC | 12:08 | |
| *** dencaval has joined #openstack-swift | 12:11 | |
| *** knl has joined #openstack-swift | 12:11 | |
| *** EmilienM is now known as EmilienM|afk | 12:41 | |
| *** chlong has joined #openstack-swift | 12:44 | |
| *** rodrigods has joined #openstack-swift | 12:44 | |
| *** krykowski has joined #openstack-swift | 12:45 | |
| *** krykowski has quit IRC | 12:52 | |
| *** SkyRocknRoll has quit IRC | 12:56 | |
| *** sandywalsh has joined #openstack-swift | 13:00 | |
| *** ir2ivps8_ has quit IRC | 13:01 | |
| *** annegentle has joined #openstack-swift | 13:03 | |
| *** krykowski has joined #openstack-swift | 13:06 | |
| glange | plindgre: do you mean something like block storage instead of object storage? https://wiki.openstack.org/wiki/Cinder | 13:12 |
| plindgre | what? | 13:13 |
| plindgre | im trying to understand object storage | 13:13 |
| plindgre | like i understand its objects with uids instead of filesystem with hierarchies | 13:14 |
| plindgre | but a practical example is what im lacking | 13:14 |
| plindgre | something like when is it useful to use it | 13:14 |
| plindgre | is it used at all with IaaS? | 13:14 |
| plindgre | for example | 13:14 |
| ctennis | the most common use case for IaaS would be as a place to send the snapshot of the VM | 13:15 |
| glange | plindgre: it's not a file system | 13:16 |
| plindgre | so in IaaS, the actual file systems that are running on the cloud machines is running in a block storage? | 13:16 |
| glange | plindgre: and objects aren't stored in a database | 13:16 |
| ctennis | some kind of block storage yes, either something ephemeral or attached as a long term storage option | 13:17 |
| glange | objects don't have uids, they have urls | 13:20 |
| glange | and md5sums/etags | 13:21 |
| *** minwoob has joined #openstack-swift | 13:25 | |
| *** ir2ivps8_ has joined #openstack-swift | 13:32 | |
| *** EmilienM|afk is now known as EmilienM | 13:37 | |
| *** esker has quit IRC | 13:41 | |
| plindgre | So Object Storage is mostly used for snapshots? | 13:53 |
| *** mtreinish has joined #openstack-swift | 13:54 | |
| glange | I guess in IaaS, but there are lots of use cases beyond that, basically the cost efficent and highly available storage of lots of data | 13:55 |
| *** ujjain- is now known as ujjain | 13:57 | |
| *** mtreinish has quit IRC | 14:05 | |
| *** vinsh has quit IRC | 14:05 | |
| *** jrichli has joined #openstack-swift | 14:06 | |
| *** silor has quit IRC | 14:06 | |
| *** vinsh has joined #openstack-swift | 14:06 | |
| -openstackstatus- NOTICE: gerrit has been restarted to clear a stuck events queue. any change events between 13:29-14:05 utc should be rechecked or have their approval votes reapplied to trigger jobs | 14:06 | |
| *** haomaiwang has joined #openstack-swift | 14:10 | |
| *** vinsh has quit IRC | 14:10 | |
| *** annegentle has quit IRC | 14:12 | |
| *** annegentle has joined #openstack-swift | 14:17 | |
| *** mahatic has joined #openstack-swift | 14:23 | |
| *** vinsh has joined #openstack-swift | 14:25 | |
| *** esker has joined #openstack-swift | 14:27 | |
| *** plindgre has quit IRC | 14:34 | |
| *** itlinux has joined #openstack-swift | 14:37 | |
| *** bkopilov has quit IRC | 14:42 | |
| *** mtreinish has joined #openstack-swift | 14:45 | |
| *** bkopilov has joined #openstack-swift | 14:45 | |
| *** marcusvrn_ has joined #openstack-swift | 14:49 | |
| *** mtreinish has quit IRC | 14:49 | |
| *** minwoob has quit IRC | 14:58 | |
| *** itlinux has quit IRC | 15:15 | |
| *** itlinux has joined #openstack-swift | 15:28 | |
| *** minwoob has joined #openstack-swift | 15:31 | |
| *** doxavore has joined #openstack-swift | 15:36 | |
| notmyname | good morning | 15:42 |
| acoles | notmyname: good morning! | 15:42 |
| *** mtreinish has joined #openstack-swift | 15:45 | |
| *** zhill has joined #openstack-swift | 15:47 | |
| peluse | good morning | 15:48 |
| notmyname | super preliminary initial results from EC testing (so super preliminary that I'm not going to give you numbers yet) | 15:50 |
| notmyname | small objects are much faster for read/write with repl over EC. but somewhere in the 2-25MB range (and especially as you get to the 100MB+ range), EC is actually faster for read/write | 15:51 |
| notmyname | and that surprised me so much that I wanted to share that :-) | 15:51 |
| peluse | cool | 15:51 |
| peluse | not totally unexpected | 15:51 |
| notmyname | maybe for you, mr cpu man | 15:51 |
| peluse | small slowdown is obvious I think - assume we're getting some speedup from parallelism over so many more nodes | 15:51 |
| peluse | or we're really not transferring any real data :) | 15:52 |
| peluse | or its just those killer CPUs you have in your test cluster!! | 15:52 |
| notmyname | nothing says "fast" like a low-power avoton! | 15:53 |
| *** annegentle has quit IRC | 15:53 | |
| notmyname | it's also your test cluster. just happens to physically be in my office ;-) | 15:53 |
| peluse | rock on | 15:53 |
| peluse | those are snappy this SOCs for sure | 15:53 |
| acoles | ah so its altitude enhanced | 15:53 |
| peluse | this=little | 15:53 |
| peluse | no idea how I typed that | 15:54 |
| notmyname | FYI https://wiki.openstack.org/wiki/Liberty_Release_Schedule | 15:54 |
| peluse | so on our end, we have our 15 node xeon cluster almost ready and our 15 node avoton cluster is ready. once baselined w/confidence we'll be able to run 3x vs repl with various CPUs, mem and network config | 15:55 |
| notmyname | the important date is October 15 for the OpenStack Liberty release | 15:55 |
| notmyname | cool | 15:55 |
| peluse | we should find time to chat about this at the summit since its both our companies working on it | 15:55 |
| notmyname | yeah, I'd love to see xeon proxies + avoton storage nodes | 15:55 |
| peluse | once it "works" we can come up with some cool experiments | 15:55 |
| peluse | once the equipment is ready and the summit is out of the way we're looking forward to hosting caleb, yes! | 15:56 |
| peluse | BTW, xeon proxy and avoton nodes is what I menat when I said the avoton cluster | 15:56 |
| peluse | they both have xeon proxies, one has avoton sn and the other is all xeon | 15:56 |
| peluse | something was goofed in out LB though (haproxy) and kiling perf so we need to straighten that out | 15:57 |
| * peluse off to a meeting for an hr, be back shortly... | 15:57 | |
| *** zaitcev has joined #openstack-swift | 15:57 | |
| *** ChanServ sets mode: +v zaitcev | 15:57 | |
| notmyname | the test I* am running now has a lot of dimensions for repl vs ec at various sizes. (*not really me) | 15:57 |
| notmyname | and we're testing at various cluster fullness levels and then will test again with failures in the cluster | 15:58 |
| peluse | cool - yeah we have 7 or 8 workload configs and also plan to test steady state, during rebalance, etc. Will review all that with ctennis before he comes out | 16:00 |
| peluse | all of our nodes are instrumented with zabbix too | 16:01 |
| *** jistr has quit IRC | 16:11 | |
| *** zul has quit IRC | 16:15 | |
| *** chlong has quit IRC | 16:19 | |
| *** pcaruana has quit IRC | 16:22 | |
| *** rmcall has joined #openstack-swift | 16:25 | |
| *** gyee has joined #openstack-swift | 16:25 | |
| *** mwheckmann has joined #openstack-swift | 16:27 | |
| mahatic | clayg, cool video on EC! I wanted to understand and it helped. | 16:28 |
| mahatic | I was at the openstack India user group meetup over the last weekend, I also met Kavit Munshi there and notmyname you're apparently "that really nice person" :) | 16:31 |
| notmyname | lol | 16:31 |
| mahatic | :) | 16:32 |
| *** SkyRocknRoll has joined #openstack-swift | 16:37 | |
| *** SkyRocknRoll has quit IRC | 16:38 | |
| *** ajiang has joined #openstack-swift | 16:41 | |
| *** itlinux has quit IRC | 16:48 | |
| *** annegentle has joined #openstack-swift | 16:54 | |
| *** marcusvrn_ is now known as marcusvrn | 16:55 | |
| *** thumpba has joined #openstack-swift | 16:57 | |
| *** thumpba has quit IRC | 16:57 | |
| *** knl has quit IRC | 16:57 | |
| *** thumpba has joined #openstack-swift | 16:58 | |
| *** thumpba_ has joined #openstack-swift | 16:59 | |
| *** annegentle has quit IRC | 17:01 | |
| *** thumpba has quit IRC | 17:03 | |
| *** annegentle has joined #openstack-swift | 17:03 | |
| *** aix has quit IRC | 17:10 | |
| *** tellesnobrega_ has joined #openstack-swift | 17:16 | |
| *** marcusvrn has quit IRC | 17:19 | |
| *** tellesnobrega_ has quit IRC | 17:21 | |
| acoles | notmyname: can you see any reason why this review is not being submitted to gate https://review.openstack.org/#/c/159530/ ? | 17:26 |
| acoles | i tried reverify which i thought did just that | 17:27 |
| *** jordanP has quit IRC | 17:27 | |
| notmyname | yeah, me too | 17:28 |
| notmyname | acoles: ok, I added "recheck" and I see it in zuul's status page now | 17:29 |
| *** mahatic has quit IRC | 17:29 | |
| notmyname | looks like it's going through the check queue again. I think I saw some messages about gerrit having issues. maybe it just didn't get picked up for the gate queue | 17:29 |
| acoles | thx. i watched it go through check but then not get to gate, lets see what happens this time | 17:30 |
| *** itlinux has joined #openstack-swift | 17:41 | |
| *** jkremer has joined #openstack-swift | 17:52 | |
| *** geaaru has quit IRC | 17:55 | |
| hurricanerix | morning | 17:58 |
| *** ajiang has quit IRC | 17:58 | |
| *** wbhuber has joined #openstack-swift | 17:58 | |
| *** ajiang has joined #openstack-swift | 18:00 | |
| *** ajiang has quit IRC | 18:01 | |
| *** krykowski has quit IRC | 18:04 | |
| notmyname | reminder that there's a meeting in about 50 minutes. today's agenda is to go over the proposed sessions on the etherpad. I want to start pushing the schedule later this week | 18:08 |
| notmyname | I've organized the etherpad to be fishbowl, working, and overflow | 18:09 |
| notmyname | link is in the topic | 18:09 |
| *** acoles is now known as acoles_away | 18:21 | |
| *** gvernik has joined #openstack-swift | 18:24 | |
| openstackgerrit | Tim Burke proposed openstack/python-swiftclient: Actually function under python 3 https://review.openstack.org/178793 | 18:26 |
| *** jordanP has joined #openstack-swift | 18:27 | |
| *** cutforth has joined #openstack-swift | 18:33 | |
| *** annegentle has quit IRC | 18:36 | |
| *** ho has joined #openstack-swift | 18:48 | |
| *** rmcall has quit IRC | 18:48 | |
| ho | good morning guys! | 18:48 |
| notmyname | good morning ho | 18:49 |
| *** rmcall has joined #openstack-swift | 18:49 | |
| *** annegentle has joined #openstack-swift | 18:49 | |
| ho | notmyname: hello | 18:49 |
| *** annegentle has quit IRC | 18:52 | |
| *** annegentle has joined #openstack-swift | 18:53 | |
| *** dmorita has joined #openstack-swift | 18:56 | |
| *** kota_ has joined #openstack-swift | 18:56 | |
| kota_ | morning | 18:57 |
| ho | kota_: morning! | 18:57 |
| notmyname | hello kota_ | 18:58 |
| mattoliverau | morning | 18:58 |
| *** erlon has joined #openstack-swift | 18:58 | |
| ho | mattoliverau: morning! | 18:58 |
| notmyname | the night side of the earth is here. must be time for a swift team meeting :-) | 18:59 |
| *** tsg_ has joined #openstack-swift | 18:59 | |
| mattoliverau | that it is :) | 18:59 |
| notmyname | meeting time in #openstack-meeting | 18:59 |
| *** riybsmh has joined #openstack-swift | 19:02 | |
| *** riybsmh has left #openstack-swift | 19:02 | |
| *** rmcall has quit IRC | 19:07 | |
| *** rmcall has joined #openstack-swift | 19:07 | |
| openstackgerrit | Merged openstack/swift: Make the reaper use same timestamp for replica deletes https://review.openstack.org/174381 | 19:08 |
| *** vinsh has quit IRC | 19:18 | |
| *** vinsh has joined #openstack-swift | 19:34 | |
| doxavore | i'm trying to delete all the objects/containers in a particular account and it's painfully slow, in large part because we have a lot of containers used during testing which are making our container-replicators work constantly. is there a way to speed up a `swift delete --all` or is that my best bet for deleting them all out (while leaving other accounts in place)? | 19:39 |
| *** david-lyle has quit IRC | 19:44 | |
| *** david-lyle has joined #openstack-swift | 19:44 | |
| *** rmcall has quit IRC | 19:49 | |
| *** dmorita has quit IRC | 19:53 | |
| *** kota_ has quit IRC | 19:55 | |
| mattoliverau | I'm going back to bed, see y'all back here later. :) | 19:55 |
| mattoliverau | as I assuming they are ^^ (kota and dmorita) :P | 19:55 |
| notmyname | and ho | 19:55 |
| cschwede | torgomatic: notmyname: rhel6 will get extended support until 30. November 2023. but: | 19:56 |
| *** gvernik has quit IRC | 19:58 | |
| *** dencaval has quit IRC | 19:58 | |
| cschwede | rhel osp 6 (our openstack juno product) requires rhel7, and rhel7 includes python2.7. so dropping py2.6 for the next releases should be fine? i don’t expect people running upstream swift on rhel6 | 19:58 |
| *** itlinux has quit IRC | 19:58 | |
| notmyname | good to kno | 19:58 |
| cschwede | zaitcev: what do you think about dropping py2.6 support in swift? | 19:59 |
| cschwede | to me it makes sense. | 19:59 |
| notmyname | I knwo that officially openstack doesn't support py26, but we've kept compatibility because there have been reasonable assumptions that deployers are running on an LTS-style distro with py26 | 19:59 |
| notmyname | but that's changed now | 19:59 |
| notmyname | torgomatic: I'm writing the email now | 19:59 |
| zaitcev | cschwede: gimme a sec | 20:00 |
| torgomatic | so the rest of openstack dropped 2.6 support a little while ago, and we held off until Ubuntu Lucid died? is that accurate? | 20:00 |
| notmyname | close enough | 20:01 |
| *** jordanP has quit IRC | 20:01 | |
| zaitcev | cschwede: I agree. | 20:01 |
| notmyname | I think it's also valid to point out that rh/cent6 support is important for people deploying swift without other openstack components | 20:01 |
| notmyname | ie not using RDO | 20:02 |
| cschwede | zaitcev: thanks Pete! | 20:02 |
| zaitcev | Although I do in fact run latest Swift on RHEL 6 with Python 2.6, it's not in production and I can migrate easily. Probably should have migrated long time ago. | 20:02 |
| notmyname | anyone know about suse? (my stereotypes say that you'd know cschwede ;-) | 20:02 |
| cschwede | notmyname: i have to disappoint you - i have no clue ;) | 20:03 |
| notmyname | heh ok | 20:03 |
| zaitcev | My personal issue is rhev-a24c-01.mpc.lab.eng.bos.redhat.com, which is a DRAC 5 system. I had to install Fedora 10 with 32-bit Firefox in order to access that shit. | 20:04 |
| zaitcev | What a pain | 20:04 |
| zaitcev | It's like everything in my life is a lie | 20:06 |
| zaitcev | The vendor (Dell) does not want to support VNC, because that's icky open source thing that you have to maintain blah blah. So, the controller serves a Java applet, straight into your browser. That ought to be compatible forever, but it's a LIE. You must have a 32-bit Java and an obsolete version too. | 20:07 |
| zaitcev | so... Firefox in 32-bit VM and ssh port forwarding is the answer | 20:08 |
| *** esker has quit IRC | 20:08 | |
| notmyname | http://lists.openstack.org/pipermail/openstack-operators/2015-April/006844.html | 20:08 |
| ho | I added "role-based access control" in the etherpad. is it necessary to do "save operation"? I could not find it. | 20:09 |
| notmyname | ho: no, it's automatically saved | 20:09 |
| ho | notmyname: thanks! | 20:09 |
| *** mwheckmann has quit IRC | 20:09 | |
| notmyname | ho: I see it. thanks | 20:10 |
| *** mwheckmann has joined #openstack-swift | 20:11 | |
| wbhuber | notmyname: good post on the backward compatibility when dropping py26 support | 20:13 |
| *** annegentle has quit IRC | 20:15 | |
| *** annegentle has joined #openstack-swift | 20:23 | |
| notmyname | torgomatic: also, for all your social media needs, https://twitter.com/notmyname/status/593510962277142529 | 20:24 |
| torgomatic | notmyname: hooray for tweetbook+ | 20:24 |
| *** itlinux has joined #openstack-swift | 20:37 | |
| *** mtreinish_ has joined #openstack-swift | 20:45 | |
| *** mtreinish has quit IRC | 20:51 | |
| *** mtreinish_ is now known as mtreinish | 20:51 | |
| *** esker has joined #openstack-swift | 21:04 | |
| dmsimard | I'm trying to find general documentation on improving Swift performance, any pointers ? | 21:05 |
| *** lpabon has quit IRC | 21:05 | |
| notmyname | dmsimard: specifically what are you trying to improve? | 21:06 |
| *** shri has joined #openstack-swift | 21:08 | |
| dmsimard | notmyname: Two use cases. | 21:08 |
| dmsimard | I have a cluster that has dual 10 Gbps connectivity across the board and I'm having a hard time pushing the Swift proxy nodes to pull more than 2 Gbps for large file uploads (multiple uploads from multiple locations simultaneously) | 21:08 |
| dmsimard | And the other use case is a large amount of smaller files (although I recognize this might be client side) | 21:09 |
| dmsimard | I just wanted to read on some general tips to see if I was doing something wrong out of the box | 21:09 |
| notmyname | dmsimard: are you using ssl? if so, where it it being terminated? could that be a bottleneck? | 21:09 |
| dmsimard | notmyname: Load balancers are active-active (DNS load balancers) SSL is terminated by haproxy 1.5 with swift-proxy on the same host (other swift-proxies are "backup" in the haproxy backend config) | 21:10 |
| dmsimard | er, DNS load balanced* | 21:11 |
| dmsimard | Unfortunately, still running Swift 1.13 but with plans to upgrade it soon to the juno/kilo releases | 21:11 |
| notmyname | where are you currently seeing the bottleneck? are you cpu limited? network limited? | 21:11 |
| dmsimard | CPU is idling, there's no i/owait anywhere - network gear and ports are not saturated either | 21:12 |
| dmsimard | using 3 replications across three zones | 21:12 |
| ctennis | how many concurrent connections hitting hte proxy at one time? | 21:13 |
| dmsimard | Not that many, perhaps 50 at most - forking something like "while true swift upload bigfile && swift delete bigfile" on a couple 1 Gbps machines and one 10 Gbps | 21:14 |
| dmsimard | Let me check what haproxy has to tell me | 21:14 |
| *** itlinux has quit IRC | 21:14 | |
| ctennis | if you're not in iowait then I'd add more clients. | 21:15 |
| ctennis | if you're just doing load gen you might want to use ss-bench, it will be much more deterministic and let you hit max throughput for a given benchmarking machine | 21:15 |
| dmsimard | I'll look into that, never heard of it - thanks | 21:16 |
| ctennis | https://github.com/swiftstack/ssbench | 21:16 |
| ctennis | intel also has one called cosbench that's more user friendly (GUI based) but may take a little more work to setup | 21:16 |
| dmsimard | glanced at ssbench, looks good enough - I'll try that | 21:17 |
| *** esker has quit IRC | 21:39 | |
| peluse | dmsimard, so we recently had something similar. 15 node cluster w/2proxies and haproxy for lb w/plenty of workload generation and couldn't get more thank like 3Gbps or something. Turned out to be something w/haproxy (not solved yet) and | 21:40 |
| peluse | some other thought of various mixture of kernels in the nodes (but I think that latter thing may not be all that real) | 21:40 |
| *** esker has joined #openstack-swift | 21:40 | |
| peluse | but anyway, you might try just using one proxy and bypassing your lb for a quick/easy test if you can | 21:41 |
| dmsimard | peluse: Yeah, I was trying to look at general tweaks I could do.. you know, like tcp_tw_recycle or tcp_tw_reuse, etc. | 21:42 |
| dmsimard | I have 5 ssbench running simultaneously right now (on 5 different machines that aren't pushovers) and I'm trying to push my LBs over 2 Gbps, pretty hard | 21:44 |
| openstackgerrit | Michael Barton proposed openstack/swift: initial import of go code https://review.openstack.org/178851 | 21:45 |
| *** theanalyst has quit IRC | 21:46 | |
| notmyname | redbo: well look at that :-) | 21:47 |
| *** theanalyst has joined #openstack-swift | 21:49 | |
| *** jrichli has quit IRC | 21:51 | |
| dmsimard | peluse: So you isolated the issue to something around haproxy ? | 21:51 |
| *** annegentle has quit IRC | 21:51 | |
| peluse | dmsimard, its a team I work with so I'm not familiar with the first hand debug but yeah that's what they tell me | 21:52 |
| peluse | dmsimard, don't believe they have it figured out yet - something to do with how they bonded the ports i believe | 21:52 |
| dmsimard | peluse: It's not a bad idea, i'll investigate that route | 21:53 |
| peluse | dmsimard, yeah, its a quick/easy way to rule out the LB assuming your numbers are low enough now that you would expect them to rise significantly going through one proxy (sounds like that is the case for you) | 21:54 |
| peluse | redbo, rock on man! | 21:54 |
| *** esker has quit IRC | 21:55 | |
| *** tsg_ has quit IRC | 22:00 | |
| *** tsg_ has joined #openstack-swift | 22:01 | |
| mattoliverau | Morning | 22:01 |
| ho | mattoliverau: morning again! | 22:04 |
| *** doxavore has quit IRC | 22:05 | |
| *** jkremer has quit IRC | 22:15 | |
| *** tsg_ has quit IRC | 22:18 | |
| *** itlinux has joined #openstack-swift | 22:22 | |
| *** itlinux has left #openstack-swift | 22:24 | |
| *** esker has joined #openstack-swift | 22:27 | |
| clayg | whooo exciting day! | 22:43 |
| mattoliverau | it's clayg! | 22:52 |
| *** km has joined #openstack-swift | 22:54 | |
| *** tsg has joined #openstack-swift | 23:03 | |
| openstackgerrit | Michael Barton proposed openstack/swift: initial import of go code https://review.openstack.org/178851 | 23:08 |
| *** otoolee has quit IRC | 23:10 | |
| notmyname | redbo: thanks | 23:10 |
| *** bkopilov has quit IRC | 23:10 | |
| clayg | it's like redbo can see my review comments before I posted them!? | 23:15 |
| *** thumpba_ has quit IRC | 23:15 | |
| clayg | I'm still having problems with installation tho | 23:16 |
| redbo | which part? | 23:16 |
| notmyname | the RAckspace copyright and the inaccessible githib repos ;-) | 23:16 |
| redbo | I don't understand copyrights, but which inacessible github repos? | 23:17 |
| notmyname | redbo: the ones you took out of the readme. the rackerlabs/hummingbird one | 23:17 |
| notmyname | oh, I was talking about the stuff I noticed. not the install | 23:18 |
| notmyname | I don't know what issues clayg is having | 23:18 |
| dmsimard | peluse: Signing off, haven't found the exact issue just yet. I'll let you know if I find it :) | 23:19 |
| dmsimard | Removing the load balancer is a luxury that's hard to do on a production infrastructure :p | 23:19 |
| dmsimard | Involves changing/distributing the keystone catalog to move off of https, etc. etc. | 23:19 |
| redbo | Yeah, I forgot those were there. I'll have to remove the readme and do an .rst eventually anyway I suppose. | 23:20 |
| notmyname | redbo: I'm going to look in to what's possible for the CI jobs on a particular branch. I suppose there's no rush, but it seems silly to run pep8 on that branch ;-) | 23:20 |
| clayg | what version of golang are you using? | 23:21 |
| redbo | I'm using 1.4.2 | 23:21 |
| clayg | first the symlink trick didn't work - so I needed to cp -a my swift dir into my GOPATH - now i'm getting name errors - or undefined or whatever for stuff that's probably not in 1.2's stdlib | 23:22 |
| clayg | redbo: k cool! | 23:22 |
| *** bkopilov has joined #openstack-swift | 23:25 | |
| redbo | hmm, the symlink works for me. But yeah you can just check it out to $GOPATH/src/openstack/swift or copy it instead. | 23:25 |
| notmyname | mailing list message with the subject "[Openstack] [Swift] tmp directory causing Swift slowdown" is interesting | 23:26 |
| notmyname | shri: is that yours? ^ | 23:26 |
| clayg | redbo: yeah with trusty's golang I got "go install: no install location for directory outside gopath" - which led to https://groups.google.com/forum/#!topic/golang-nuts/f5ZYztyHK5I | 23:26 |
| swifterdarrell | clayg: trusty's golang?! #buildfromsourc | 23:27 |
| swifterdarrell | clayg: trusty's golang?! #buildfromsource | 23:27 |
| swifterdarrell | ;) | 23:27 |
| notmyname | redbo: what distro are you running on? | 23:28 |
| redbo | ummm I don't know, precise? | 23:28 |
| redbo | but I downloaded go from golang.org | 23:28 |
| redbo | nobody builds go from source, because you need go to build it :) | 23:31 |
| clayg | bah how strange, go install: no install location for directory /vagrant/swift/hummingbird outside GOPATH | 23:31 |
| clayg | trusty | 23:31 |
| *** otoolee has joined #openstack-swift | 23:31 | |
| clayg | swifterdarrell: I was curious how the distro's are keeping up with the new hawtness, i'm not *surpirsed* but there was also a question of how quickly the language is moving around - like if we'll need to build and distribute a new golang every swift release | 23:32 |
| clayg | redbo: yeah I pulled down the linux binary tarball from https://golang.org/dl/ and extracted into $HOME on my vagrant-swift-all-in-one - then `export GOPATH=${HOME}/go` and followed your instructions. | 23:32 |
| redbo | I think if the binary distrubition isn't at /usr/local/go, you need to set GOROOT to its location too. | 23:34 |
| swifterdarrell | clayg: ya, i was just being stupid; surely you just set a minimum version of golang that any particular drop of hummingbird code requires and the deployer (or gate job, haha!) is responsible for making that available | 23:35 |
| clayg | oh i think i had some stupid trusty shit still holding me back | 23:36 |
| clayg | swifterdarrell: yeah right on! | 23:37 |
| redbo | I think the language is mostly frozen for now and they're just working on improving the GC and stuff anyway. | 23:37 |
| clayg | redbo: oic, yeah it was complaining about the GOROOT==GOPATH having no effect | 23:37 |
| redbo | as soon as there's a goslo, I'm rewriting it in nim or rust | 23:38 |
| redbo | ok I have to go to a hockey game and will be back later! | 23:39 |
| shri | notmyname: yes! | 23:43 |
| notmyname | shri: interesting idea (to move the tmp dir lower in the fs tree) | 23:44 |
| *** zhill has quit IRC | 23:44 | |
| shri | does it have other repercussions? replicator? | 23:45 |
| notmyname | shri: IIRC the daemons will filter directories that aren't actual hash suffixes. and "tmp" isn't a valid hash suffix name. so it might work :-) | 23:45 |
| shri | cool! | 23:46 |
| shri | I have that change in my local tree… breaks a few unit tests :-) | 23:46 |
| notmyname | shri: it's certainly something worth exploring | 23:47 |
| *** wbhuber has quit IRC | 23:47 | |
| shri | I'll take a look. | 23:47 |
| shri | actually one more option would be to create the tmp file in the final datadir with O_TMPFILE and then use linkat. But not sure if python2.7 has O_TMPFILE. And it surely does not have linkat. | 23:49 |
| *** chlong has joined #openstack-swift | 23:50 | |
| notmyname | shri: isn't that a recent kernel thing? sounds familiar. I think someone may have a patch for that | 23:50 |
| shri | python 3.4 has it I believe: https://bugs.python.org/issue18673 | 23:52 |
| *** km__ has joined #openstack-swift | 23:52 | |
| notmyname | clayg: https://gist.github.com/notmyname/45ba9a889a4e315e630c what am I missing? | 23:53 |
| *** km has quit IRC | 23:53 | |
| notmyname | clayg: nm | 23:54 |
| notmyname | got it | 23:54 |
| notmyname | well, no | 23:56 |
| notmyname | https://gist.github.com/notmyname/4eed6f719c092b05393e | 23:56 |
| clayg | notmyname: follow this https://golang.org/doc/install#tarball | 23:56 |
| clayg | notmyname: i posted some notes on the review | 23:57 |
| notmyname | yeah, I got it working | 23:58 |
| notmyname | I had GOROOT set incorrectly | 23:58 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!