| *** chlong has joined #openstack-swift | 00:15 | |
| *** mwheckmann has quit IRC | 00:31 | |
| *** marzif has quit IRC | 01:00 | |
| *** tellesnobrega has quit IRC | 01:16 | |
| *** petertr7 is now known as petertr7_away | 01:25 | |
| *** logan2 has joined #openstack-swift | 01:25 | |
| *** tellesnobrega has joined #openstack-swift | 01:29 | |
| *** CaioBrentano has joined #openstack-swift | 02:16 | |
| *** bkopilov_wfh has quit IRC | 02:18 | |
| *** CaioBrentano has quit IRC | 02:20 | |
| *** vjujjuri has joined #openstack-swift | 02:34 | |
| *** kei_yama has quit IRC | 03:11 | |
| *** notmyname has joined #openstack-swift | 03:13 | |
| *** ChanServ sets mode: +v notmyname | 03:13 | |
| *** patchbot has joined #openstack-swift | 03:14 | |
| *** kei_yama has joined #openstack-swift | 03:15 | |
| *** vjujjuri has quit IRC | 03:45 | |
| *** fanyaohong has joined #openstack-swift | 03:46 | |
| *** kota_ has joined #openstack-swift | 03:47 | |
| *** bkopilov_wfh has joined #openstack-swift | 04:13 | |
| *** asettle has joined #openstack-swift | 04:24 | |
| *** links has joined #openstack-swift | 04:27 | |
| *** kota_ has quit IRC | 05:01 | |
| *** logan2 has quit IRC | 05:08 | |
| *** petertr7_away is now known as petertr7 | 05:19 | |
| *** petertr7 is now known as petertr7_away | 05:28 | |
| *** yuan has quit IRC | 05:29 | |
| *** proteusguy has quit IRC | 05:37 | |
| *** krykowski has joined #openstack-swift | 05:43 | |
| *** asettle has quit IRC | 05:50 | |
| *** proteusguy has joined #openstack-swift | 05:53 | |
| *** fanyaohong has quit IRC | 06:03 | |
| *** SkyRocknRoll has joined #openstack-swift | 06:10 | |
| *** SkyRocknRoll has joined #openstack-swift | 06:10 | |
| *** chlong has quit IRC | 06:30 | |
| *** vjujjuri has joined #openstack-swift | 06:35 | |
| *** kei_yama has quit IRC | 06:39 | |
| *** kei_yama has joined #openstack-swift | 06:45 | |
| *** vjujjuri has quit IRC | 06:45 | |
| *** admin6 has joined #openstack-swift | 07:06 | |
| *** logan2 has joined #openstack-swift | 07:06 | |
| *** petertr7_away is now known as petertr7 | 07:07 | |
| *** jordanP has joined #openstack-swift | 07:09 | |
| *** ppai has joined #openstack-swift | 07:13 | |
| *** logan2 has quit IRC | 07:17 | |
| *** petertr7 is now known as petertr7_away | 07:17 | |
| *** bkopilov_wfh is now known as bkopilov | 07:24 | |
| *** krykowski has quit IRC | 07:25 | |
| *** ppai has quit IRC | 07:29 | |
| *** jistr has joined #openstack-swift | 07:39 | |
| openstackgerrit | Merged openstack/swift: Add Swift Inspector to assoicated projects https://review.openstack.org/184661 | 07:41 |
|---|---|---|
| *** SkyRocknRoll has quit IRC | 07:45 | |
| *** SkyRocknRoll has joined #openstack-swift | 07:48 | |
| *** SkyRocknRoll has quit IRC | 07:48 | |
| *** SkyRocknRoll has joined #openstack-swift | 07:48 | |
| *** silor has joined #openstack-swift | 08:01 | |
| openstackgerrit | Michael MATUR proposed openstack/swift: fixup! Patch of "parse_content_disposition" method to meet RFC 2183 https://review.openstack.org/185389 | 08:05 |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/python-swiftclient: Updated from global requirements https://review.openstack.org/89250 | 08:09 |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/swift: Updated from global requirements https://review.openstack.org/88736 | 08:10 |
| *** flaper87 has joined #openstack-swift | 08:13 | |
| *** mmcardle has joined #openstack-swift | 08:24 | |
| *** joeljwright has joined #openstack-swift | 08:24 | |
| admin6 | Hi all. I’m just about to test Erasure coding with v2.3. I’m getting this error on the proxy : | 08:37 |
| admin6 | ERROR: Invalid Storage Policy Configuration in /etc/swift/swift.conf (Invalid option 'ec_num_parity_fragments' in storage-policy section, for index '4') | 08:37 |
| admin6 | :-) Sorry for the buzz, I was simply missing option : policy_type = erasure_coding | 08:41 |
| *** hseipp has joined #openstack-swift | 08:55 | |
| *** petertr7_away is now known as petertr7 | 08:56 | |
| ho | admin6: No problem :-) You solved it before we recoginzed your message. | 08:59 |
| *** logan2 has joined #openstack-swift | 09:02 | |
| *** leopoldj has joined #openstack-swift | 09:04 | |
| *** petertr7 is now known as petertr7_away | 09:06 | |
| *** aix has joined #openstack-swift | 09:08 | |
| *** hseipp has quit IRC | 09:15 | |
| *** hseipp has joined #openstack-swift | 09:15 | |
| *** haomaiwa_ has quit IRC | 09:17 | |
| openstackgerrit | Michael MATUR proposed openstack/swift: fixup!Patch of "parse_content_disposition" method to meet RFC2183 https://review.openstack.org/185389 | 09:24 |
| *** haomaiwang has joined #openstack-swift | 10:06 | |
| *** proteusguy has quit IRC | 10:15 | |
| *** goodes has quit IRC | 10:29 | |
| *** acoles_away is now known as acoles | 10:30 | |
| *** marzif has joined #openstack-swift | 10:39 | |
| *** proteusguy has joined #openstack-swift | 10:42 | |
| *** petertr7_away is now known as petertr7 | 10:45 | |
| *** acoles is now known as acoles_away | 10:45 | |
| *** petertr7 is now known as petertr7_away | 10:55 | |
| *** acoles_away is now known as acoles | 11:00 | |
| *** marcusvrn has joined #openstack-swift | 11:08 | |
| *** aix has quit IRC | 11:14 | |
| *** aix has joined #openstack-swift | 11:15 | |
| *** ho has quit IRC | 11:30 | |
| *** jistr is now known as jistr|class | 11:33 | |
| *** haomaiwang has quit IRC | 11:40 | |
| *** ppai has joined #openstack-swift | 11:47 | |
| *** ChanServ sets mode: +v cschwede | 12:00 | |
| *** km has quit IRC | 12:10 | |
| *** logan2 has quit IRC | 12:12 | |
| *** ekarlso has quit IRC | 12:16 | |
| *** ekarlso has joined #openstack-swift | 12:16 | |
| *** samueldmq has joined #openstack-swift | 12:17 | |
| samueldmq | hi, good morning | 12:18 |
| samueldmq | swift doesn't use rbac for authn, does it ? | 12:18 |
| samueldmq | like other projects in the openstack ecosystem | 12:19 |
| *** ppai has quit IRC | 12:21 | |
| *** SkyRocknRoll has quit IRC | 12:23 | |
| *** petertr7_away is now known as petertr7 | 12:34 | |
| *** bkopilov has quit IRC | 12:34 | |
| *** petertr7 is now known as petertr7_away | 12:43 | |
| *** jistr|class is now known as jistr | 13:00 | |
| *** kei_yama has quit IRC | 13:08 | |
| *** CaioBrentano has joined #openstack-swift | 13:11 | |
| *** mwheckmann has joined #openstack-swift | 13:14 | |
| *** logan2 has joined #openstack-swift | 13:21 | |
| *** haypo has joined #openstack-swift | 13:21 | |
| haypo | hi. i write a first patch serie to port swift to python3: https://review.openstack.org/#/q/status:open+project:openstack/swift+branch:master+topic:py3,n,z | 13:21 |
| haypo | tell me if you need more information on them, or on python 3 in general | 13:22 |
| *** links has quit IRC | 13:29 | |
| *** acoles is now known as acoles_away | 13:31 | |
| *** zynisch_o7 has joined #openstack-swift | 13:43 | |
| *** zynisch_o7 has quit IRC | 13:45 | |
| *** logan2 has quit IRC | 13:54 | |
| *** zynisch_o7 has joined #openstack-swift | 13:56 | |
| *** zynisch_o7 has quit IRC | 13:57 | |
| *** zynisch_o7 has joined #openstack-swift | 13:58 | |
| *** annegentle has joined #openstack-swift | 13:59 | |
| *** zynisch__ has joined #openstack-swift | 14:01 | |
| *** zynisch_o7 has quit IRC | 14:01 | |
| *** zynisch__ has quit IRC | 14:03 | |
| *** minwoob has joined #openstack-swift | 14:04 | |
| *** zynisch_o7 has joined #openstack-swift | 14:04 | |
| *** zynisch_o7 has quit IRC | 14:11 | |
| *** zynisch_o7 has joined #openstack-swift | 14:13 | |
| *** zynisch_o7 has quit IRC | 14:18 | |
| *** bkopilov has joined #openstack-swift | 14:18 | |
| *** annegentle has quit IRC | 14:19 | |
| *** annegentle has joined #openstack-swift | 14:19 | |
| *** breitz has joined #openstack-swift | 14:20 | |
| *** petertr7_away is now known as petertr7 | 14:23 | |
| *** petertr7 is now known as petertr7_away | 14:32 | |
| *** annegentle has quit IRC | 14:38 | |
| *** zynisch_o7 has joined #openstack-swift | 14:44 | |
| *** proteusguy has quit IRC | 14:47 | |
| *** jrichli has joined #openstack-swift | 14:50 | |
| *** annegentle has joined #openstack-swift | 14:50 | |
| *** proteusguy has joined #openstack-swift | 14:59 | |
| *** geaaru has joined #openstack-swift | 15:02 | |
| *** haomaiwang has joined #openstack-swift | 15:04 | |
| *** annegentle has quit IRC | 15:11 | |
| *** alejandrito has joined #openstack-swift | 15:14 | |
| *** petertr7_away is now known as petertr7 | 15:23 | |
| *** petertr7 is now known as petertr7_away | 15:33 | |
| *** zaitcev has joined #openstack-swift | 15:35 | |
| *** ChanServ sets mode: +v zaitcev | 15:35 | |
| notmyname | good morning | 15:37 |
| *** haomaiwang has quit IRC | 15:37 | |
| *** haomaiwang has joined #openstack-swift | 15:37 | |
| *** jistr has quit IRC | 15:40 | |
| *** gyee has joined #openstack-swift | 15:42 | |
| *** vjujjuri has joined #openstack-swift | 15:44 | |
| tdasilva | zaitcev: looking for me? | 15:44 |
| *** leopoldj has quit IRC | 15:45 | |
| zaitcev | tdasilva: if you could look over Pai's fixes to your versioned writes middleware, it would be great | 15:45 |
| zaitcev | tdasilva: other than that, keep up the good work :-) | 15:45 |
| zaitcev | https://review.openstack.org/134347 | 15:47 |
| tdasilva | zaitcev: sounds good. I'm hoping we can put some focus back on versioning and copy middleware | 15:47 |
| *** slavisa has joined #openstack-swift | 15:58 | |
| *** zynisch_o7 has quit IRC | 15:58 | |
| *** zynisch_o7 has joined #openstack-swift | 15:58 | |
| *** slavisa has quit IRC | 16:01 | |
| *** slavisa has joined #openstack-swift | 16:01 | |
| *** slavisa has quit IRC | 16:06 | |
| vjujjuri | Hi | 16:07 |
| vjujjuri | I am not getting good perf numbers on swift. What am I doing wrong? | 16:07 |
| vjujjuri | doing 1k writes; I have a 4 node clsuter. | 16:07 |
| vjujjuri | 3 storage nodes + 4th one for keystone and proxy; and this is where I am running my tests | 16:08 |
| vjujjuri | Last-byte latency: 0.042 - 0.061 0.051 ( 0.005) 0.059 ( tiny objs) | 16:08 |
| vjujjuri | Average latency is 51ms for 1k objects | 16:08 |
| vjujjuri | my storage nodes are SSDs and my network is 1gig | 16:08 |
| vjujjuri | obviously I think I am doing something wrong | 16:09 |
| *** petertr7_away is now known as petertr7 | 16:09 | |
| vjujjuri | direct writes to ssds I get 1ms | 16:09 |
| vjujjuri | I am using ssbench | 16:09 |
| vjujjuri | anyone here experimented small objects on SSD with swift? | 16:09 |
| vjujjuri | I am using icehouse | 16:09 |
| swifterdarrell | vjujjuri: the most common mistake when using ssbench is to start with some arbitrarily-chosen concurrency level that is too high for the hardware, saturating something, and getting poor results; what concurrency were you using? did you make sure the server running ssbench wasn't the bottleneck? did you try concurrency of 1 for comparison? | 16:11 |
| vjujjuri | -p 2 -u 2 | 16:11 |
| vjujjuri | and objects are 10,000 | 16:11 |
| vjujjuri | I mean -u2 --workers 2 | 16:12 |
| vjujjuri | I tried upto 16 parallel | 16:12 |
| vjujjuri | results did not change much | 16:12 |
| vjujjuri | swifterdarrell: I even tried -u 1 | 16:12 |
| vjujjuri | and similar numbers | 16:13 |
| vjujjuri | bench is on the proxy machine | 16:13 |
| *** haypo has quit IRC | 16:13 | |
| swifterdarrell | vjujjuri: one tack would be to use something else to corroborate the results... like what kind of latencies do you see with PUTs from curl? a single-stream swift-bench? | 16:14 |
| swifterdarrell | vjujjuri: i think ssbench might also record the worst txid, so grepping that in the logs might reveal something | 16:15 |
| *** vjujjuri_ has joined #openstack-swift | 16:16 | |
| *** vjujjuri has quit IRC | 16:17 | |
| *** vjujjuri_ is now known as vjujjuri | 16:17 | |
| *** admin6 has quit IRC | 16:17 | |
| vjujjuri | I tried swift upload even that is not giving great numbers | 16:19 |
| vjujjuri | sh-4.2# ls -l jv1k | 16:19 |
| vjujjuri | -rw------- 1 root root 1024 May 25 09:25 jv1k | 16:19 |
| vjujjuri | sh-4.2# set -o vi | 16:19 |
| vjujjuri | sh-4.2# date; swift upload myfiles jv1k; date | 16:19 |
| vjujjuri | Tue May 26 09:18:41 PDT 2015 | 16:19 |
| vjujjuri | jv1k | 16:19 |
| vjujjuri | Tue May 26 09:18:42 PDT 2015 | 16:19 |
| vjujjuri | I put this in a loop | 16:19 |
| vjujjuri | to put say 100 objects | 16:19 |
| vjujjuri | it is taking around 50-60 ms | 16:19 |
| swifterdarrell | vjujjuri: swift CLI tool authenticates every ClI invocation, so that's a poor tool for this | 16:19 |
| swifterdarrell | vjujjuri: getting an auth token and then using curl in a loop would be better | 16:20 |
| *** annegentle has joined #openstack-swift | 16:20 | |
| vjujjuri | I don't know how to use authtoken | 16:20 |
| vjujjuri | whatever swift stat gives? | 16:20 |
| swifterdarrell | vjujjuri: ya... "swift stat -v" | 16:21 |
| vjujjuri | ok; I got a multi like auth token | 16:21 |
| swifterdarrell | vjujjuri: then include that in your curl with -H "X-Auth-Token: <...>" | 16:21 |
| vjujjuri | and how do I use it? | 16:21 |
| swifterdarrell | vjujjuri: oh... v2/keystone or something? | 16:21 |
| vjujjuri | let me put it in pastebin | 16:22 |
| vjujjuri | http://pastebin.com/nM26fc8Y | 16:23 |
| vjujjuri | swifterdarrell: ^ | 16:23 |
| zackmdavis | that is an admirably high-entropy auth token | 16:24 |
| vjujjuri | I did not do anything special | 16:25 |
| glange | you don't get an auth token like that without something special having been done :) | 16:25 |
| vjujjuri | glange: may be I did but not intentional | 16:26 |
| vjujjuri | I just followed directions in the swift page | 16:26 |
| glange | ok | 16:27 |
| vjujjuri | so you guys think my auth is causing the latency? | 16:27 |
| vjujjuri | anyway I can avoid or reduce the overhead? | 16:27 |
| notmyname | any chance you can use keystone's uuid tokens? | 16:28 |
| vjujjuri | notmyname: anything is fine. Security is not a great concern I am looking for latency numbers | 16:28 |
| vjujjuri | but I need to know how to get that | 16:28 |
| vjujjuri | new here :) | 16:29 |
| vjujjuri | notmyname: if uuid auth saves sure.. can you tell me how to do that? | 16:31 |
| *** haomaiwang has quit IRC | 16:31 | |
| notmyname | uuid tokens are 32 bytes long, so the per-request transfer overhead is less | 16:31 |
| glange | if it's authing each request, that is the main cause of bad performance? | 16:32 |
| *** hseipp has left #openstack-swift | 16:32 | |
| notmyname | vjujjuri: http://docs.openstack.org/developer/keystone/configuration.html#token-provider looks like it has some info on configuring keystone for that | 16:33 |
| vjujjuri | notmyname: if it is just overhead it is negligible ; unless as glange said if auth is taking each request. | 16:33 |
| *** hseipp has joined #openstack-swift | 16:33 | |
| glange | vjujjuri: you could try just doing curl requests in a loop as described here: http://docs.openstack.org/developer/swift/development_saio.html | 16:33 |
| vjujjuri | may be I just disable auth? | 16:33 |
| glange | you get the auth token once and then pass it in on the command line | 16:33 |
| notmyname | pki tokens aren't cacheable, so it has to go back to keystone every time (AIUI) | 16:33 |
| glange | maybe by setting an env variable | 16:33 |
| *** hseipp has left #openstack-swift | 16:34 | |
| *** acoles_away is now known as acoles | 16:37 | |
| notmyname | vjujjuri: if you want to test to see if it is actually auth that's slowing you down or not, then take keystone out of the proxy server pipeline (in /etc/swift/proxy-server.conf), restart the proxy, and run the test again | 16:38 |
| *** haomaiwang has joined #openstack-swift | 16:38 | |
| *** zynisch_o7 has quit IRC | 16:39 | |
| *** zynisch_o7 has joined #openstack-swift | 16:42 | |
| *** zynisch_o7 has quit IRC | 16:42 | |
| vjujjuri | notmyname: let me try that | 16:47 |
| vjujjuri | [pipeline:main] | 16:48 |
| vjujjuri | #pipeline = healthcheck cache authtoken keystoneauth proxy-server | 16:48 |
| vjujjuri | pipeline = healthcheck cache proxy-server | 16:48 |
| vjujjuri | notmyname: is this good? | 16:48 |
| vjujjuri | I took out authtoken and keystonauth both out | 16:48 |
| notmyname | vjujjuri: what version of swift do you havE? | 16:48 |
| notmyname | vjujjuri: yeah, that will work | 16:49 |
| vjujjuri | 2.0.3 | 16:49 |
| notmyname | hmm..I don't think that's right (there's never been a 2.0.3 version of the server-side swift). I think that's the version of the CLI you have | 16:50 |
| notmyname | vjujjuri: do this: `curl http://<swift endpoint>/info | python -m json.tool` | 16:50 |
| notmyname | vjujjuri: that will give you a formatted json output of the capabilities of the cluster, including the version | 16:51 |
| vjujjuri | notmyname I have two end points one with controller:5000 and other with controller:8080 | 16:52 |
| notmyname | :8080 is likely the swift one. the :5000 one is probably keystone | 16:52 |
| vjujjuri | notmyname: sorry for basic questions. | 16:54 |
| vjujjuri | http://controller:8080/v1/AUTH_%(tenant_id)s | 16:54 |
| vjujjuri | how do I know what is my tenatnt_id is? | 16:54 |
| vjujjuri | the above one is listed | 16:54 |
| vjujjuri | with endpoint-list | 16:54 |
| vjujjuri | BTW even after restarting my proxy with keystone deleted from pipeline I still get same numbers | 16:54 |
| *** acoles is now known as acoles_away | 17:06 | |
| *** harlowja has joined #openstack-swift | 17:07 | |
| *** jordanP has quit IRC | 17:07 | |
| *** annegentle has quit IRC | 17:09 | |
| openstackgerrit | Merged openstack/swift: Cleanup and extend end to end ssync tests https://review.openstack.org/179379 | 17:12 |
| notmyname | vjujjuri: ah, good question | 17:12 |
| notmyname | vjujjuri: if you don't have any auth, I think the tenant id could be anything | 17:13 |
| notmyname | vjujjuri: for the sake of argument, just use "AUTH_foo" or something like that | 17:13 |
| vjujjuri | sh-4.2# curl http://controller:8080/v1/AUTH_foo/info | 17:18 |
| vjujjuri | <html><h1>Not Found</h1><p>The resource could not be found.</p></html>sh-4.2# | 17:18 |
| vjujjuri | notmyname: ^ | 17:18 |
| notmyname | http://controller:8080/info | 17:18 |
| vjujjuri | ok | 17:23 |
| vjujjuri | that says | 17:23 |
| vjujjuri | version": "1.13.1" | 17:23 |
| notmyname | that's the one :-) | 17:23 |
| vjujjuri | ok; so basically disabling auth in pipeline did not improve numbers | 17:24 |
| vjujjuri | you guys also see these are bad numbers right? | 17:25 |
| openstackgerrit | Samuel Merritt proposed openstack/swift: Handle empty suffix dirs the same for EC and replication https://review.openstack.org/185679 | 17:25 |
| clayg | google calander - why do you hate UTC so much!? | 17:31 |
| notmyname | clayg: use iceland | 17:31 |
| clayg | it's so stupid, I found GMT (no daylight savings) and I put it to 9:00pm and click save | 17:32 |
| notmyname | clayg: reykjavik is UTC with no DST | 17:33 |
| clayg | THEN it moves it forward one hour? why does it move it!? | 17:33 |
| clayg | yeah i tried iceland - same thing | 17:33 |
| notmyname | clayg: yeah, it's really terrible | 17:33 |
| *** annegentle has joined #openstack-swift | 17:33 | |
| vjujjuri | notmyname: Somehow it is still looking for auth; Just for experiment I unset OS_PASSWORD | 17:33 |
| clayg | i had to set it 8pm so that it would move it 9pm after save - it makes no sense | 17:33 |
| vjujjuri | and now I cant upload | 17:33 |
| vjujjuri | uth version 1.0 requires ST_AUTH, ST_USER, and ST_KEY environment variables | 17:33 |
| vjujjuri | to be set or overridden with -A, -U, or -K. | 17:33 |
| notmyname | vjujjuri: ok. let's back up. | 17:34 |
| notmyname | vjujjuri: also, I'm sorry. i know this must be frustrating | 17:34 |
| clayg | i don't get it - google is a big company - and I know they use this app - they must have a workflow that makes sense | 17:34 |
| notmyname | vjujjuri: go ahead and put keystone back in. ideally, you should switch it to uuid tokens. they are much better for swift thank the pki tokens | 17:35 |
| notmyname | vjujjuri: so the problem you're seeing is a 50ms latency on uploading a 1k object? | 17:35 |
| vjujjuri | ok; I will put it back; but what surprises me is; when you took it off of backend why is it still asking for OS auth; may be I am doing something wrong in disabling auth | 17:35 |
| vjujjuri | notmyname: yep 50ms on ssd is pretty high | 17:36 |
| vjujjuri | for 1k objects | 17:36 |
| vjujjuri | I should be seeing 5-6 ms | 17:36 |
| notmyname | vjujjuri: in order to see where the slowdown is, you'll have to do some investigation. | 17:37 |
| vjujjuri | yeah; first step I tough is disabling auth. but I guess not having success there | 17:38 |
| notmyname | I don't run swift on SSDs, so I can't validate what you're seeing | 17:38 |
| vjujjuri | if I run on regular disks | 17:39 |
| vjujjuri | these are directories | 17:39 |
| notmyname | vjujjuri: but with smaller objects, the latency will be dominated by the various network hops in the cluster | 17:39 |
| vjujjuri | I get around 100-150s | 17:39 |
| vjujjuri | ms | 17:39 |
| notmyname | you've got 3 servers? how many drives? | 17:39 |
| vjujjuri | 3 storage servers. But I have created directories on XFS | 17:39 |
| vjujjuri | as i don't have individual spindles | 17:40 |
| notmyname | I don't follow. what do you mean by not having individual spindles? | 17:41 |
| *** HenryG has quit IRC | 17:42 | |
| vjujjuri | notmyname: I have only one HDD and one SSD on each of my machines; Swift's recommendation is to use entire disk as one mount point. | 17:42 |
| notmyname | right | 17:42 |
| vjujjuri | Since I have only one HDD on the machine I creaed directories | 17:42 |
| vjujjuri | So my object ring contains | 17:43 |
| vjujjuri | 9 entries | 17:43 |
| vjujjuri | 3 machines; and 3 directories on each machine | 17:43 |
| vjujjuri | and I put mount_check=false | 17:43 |
| notmyname | ah. so you have multiple mount points on each server? | 17:43 |
| vjujjuri | in the configuration machine | 17:43 |
| notmyname | ok | 17:43 |
| notmyname | don't do that ;-) | 17:43 |
| vjujjuri | I understand that is not recommended but that is the best I can do with the hardware I have :( | 17:44 |
| vjujjuri | and I will be surprised if that can cause huge perf issue; | 17:44 |
| notmyname | on reason is that swift is placing data across what you tell it in the ring. so it may place replicas across "drives" that are actually the same physical device. so you lose one device and then you're toast | 17:44 |
| notmyname | it's not a problem to have only one mount point per server | 17:45 |
| notmyname | so you'd have 3 in the cluster total in your 3 servers | 17:45 |
| vjujjuri | notmyname: this is just perf testing | 17:45 |
| vjujjuri | setup in my lab | 17:45 |
| *** HenryG has joined #openstack-swift | 17:46 | |
| notmyname | ok, but deploying it a way that isn't what you'd see in a prod setup isn't any way to do perf testing | 17:47 |
| *** thumpba_ has quit IRC | 17:47 | |
| vjujjuri | I agree; trying to workout with what is available. don;t have access to production machines. | 17:49 |
| vjujjuri | but just the basic tests | 17:49 |
| vjujjuri | shows something wrong may be in my setup | 17:49 |
| vjujjuri | as I get 1ms local writes to ssds | 17:49 |
| vjujjuri | and my network is 1gige | 17:49 |
| *** fthiagogv has joined #openstack-swift | 17:50 | |
| *** fthiagogv has quit IRC | 17:50 | |
| *** fthiagogv has joined #openstack-swift | 17:50 | |
| *** cloudm2 has joined #openstack-swift | 17:51 | |
| *** haomaiw__ has joined #openstack-swift | 17:51 | |
| *** haomaiwang has quit IRC | 17:52 | |
| notmyname | vjujjuri: there are 2 things that are next. ie 2 things I'd do if I were you. first, set it up again to only use one mount point per drive. second, you'll have to start measuring various parts of the cluster | 17:52 |
| *** mmcardle has quit IRC | 17:52 | |
| notmyname | for the second one, I think I heard of some tools last week that people have to do that. maybe hurricanerix_ has something | 17:53 |
| *** annegentle has quit IRC | 17:58 | |
| *** annegentle has joined #openstack-swift | 17:59 | |
| *** HenryG has quit IRC | 18:01 | |
| *** HenryG has joined #openstack-swift | 18:01 | |
| tdasilva | clayg: I used Ghana (Accra) this morning and it seems to have worked | 18:05 |
| *** harlowja_ has joined #openstack-swift | 18:06 | |
| *** harlowja has quit IRC | 18:06 | |
| *** aix has quit IRC | 18:25 | |
| *** fthiagogv has quit IRC | 18:32 | |
| *** fthiagogv has joined #openstack-swift | 18:33 | |
| *** gyee has quit IRC | 18:37 | |
| notmyname | reminder to everyone: check the channel topic. our meeting time has changed to 2100UTC (this is 2 hours after the previous time) | 18:38 |
| *** alejandrito has quit IRC | 18:38 | |
| openstackgerrit | Samuel Merritt proposed openstack/swift: EC: don't 503 on marginally-successful PUT https://review.openstack.org/180795 | 18:39 |
| *** logan2 has joined #openstack-swift | 18:47 | |
| clayg | what'd I do? | 18:52 |
| openstackgerrit | Victor Stinner proposed openstack/swift: Replace dict.iteritems() with six.iteritems(dict) https://review.openstack.org/185451 | 18:56 |
| clayg | whoooo python3! | 18:58 |
| *** silor has quit IRC | 19:02 | |
| *** annegentle has quit IRC | 19:15 | |
| openstackgerrit | Michael Barton proposed openstack/swift: go: log 499 on client early disconnect https://review.openstack.org/183577 | 19:16 |
| *** barra204 has joined #openstack-swift | 19:16 | |
| *** annegentle has joined #openstack-swift | 19:23 | |
| portante | redbo: do you have any performance numbers yet comparing hummingbird to regular swift? | 19:24 |
| peluse_ | clayg, yeah I saw that. Note that if we move to py3 then there's a good chance an Intel team (compiler/interpreter team) will toss us some cpython updates to test perf improvements... | 19:39 |
| peluse_ | clayg, headsup I may ned to pick your brain a bit on putting together some sane tests for the EC GET cases we talked about in vancouver, looking for a decent way now so will continue for a while then punt if I have to | 19:41 |
| openstackgerrit | paul luse proposed openstack/swift: Exclude local_dev from sync partners on failure https://review.openstack.org/175076 | 19:52 |
| *** marcusvrn has quit IRC | 19:59 | |
| *** mcnully has joined #openstack-swift | 20:02 | |
| *** mcnully has quit IRC | 20:06 | |
| *** acampbell has joined #openstack-swift | 20:08 | |
| openstackgerrit | Merged openstack/python-swiftclient: be more explicit in the --version CLI option https://review.openstack.org/182892 | 20:09 |
| openstackgerrit | Merged openstack/swift: Add swift-durability-calculator line to docs https://review.openstack.org/185313 | 20:09 |
| *** wbhuber has joined #openstack-swift | 20:11 | |
| *** fthiagogv has quit IRC | 20:13 | |
| *** samueldmq has quit IRC | 20:15 | |
| *** gyee has joined #openstack-swift | 20:28 | |
| *** acampbell has quit IRC | 20:30 | |
| *** setmason has joined #openstack-swift | 20:31 | |
| redbo | portante: no, dfg has some better numbers but he's out today. I'll see if I can figure out where they live. | 20:36 |
| portante | redbo: thanks | 20:36 |
| *** vjujjuri has quit IRC | 20:40 | |
| *** HenryG has quit IRC | 21:04 | |
| clayg | peluse_: check out test.unit.proxy.controller.test_obj | 21:07 |
| *** HenryG has joined #openstack-swift | 21:07 | |
| clayg | peluse_: but at leaast part of it is strategy for the moment - it would have been great to run through a matrix high-bandwidth on Friday and look for some general strategies to apply | 21:08 |
| clayg | it's unclear to me what to do with a primary connection when we have to throw it out for a duplicate or because it missed a durable - there may be some hidden knowledge on the node we could expose to the proxy with x-backend headers about "other available fragments" - and esspecially since we already have a socket - it may be cheaper than a new connection to a handoff node | 21:10 |
| peluse_ | clayg, yeah, I was looking there and also looking at something similar to TestECMismatchedFA and agree wrt strategies, etc. Right now just trying to get a base test together that allows me to dictate the sceanrios we white boarded. Have a few ideas so will work on them and ping ya maybe tomorrow sometime if I get stuck | 21:29 |
| notmyname | has anyone thought about implementing CORS as middleware instead of in the proxy? I don't remember if there's anything in there that requires it to be in the proxy | 21:30 |
| notmyname | there's a current cross-project spec to have a common cors middleware | 21:31 |
| peluse_ | or maybe even a smaller version, CORS light (couldnt resist) | 21:31 |
| *** peluse_ is now known as peluse | 21:31 | |
| *** ChanServ sets mode: +v peluse | 21:31 | |
| notmyname | nice :-) | 21:32 |
| notmyname | we already had keystone and keystone light | 21:32 |
| notmyname | and PBR | 21:32 |
| notmyname | and now Fernet tokens in keystone | 21:32 |
| *** swdweeb has quit IRC | 21:49 | |
| *** mwheckmann has quit IRC | 21:52 | |
| *** geaaru has quit IRC | 21:53 | |
| *** joeljwright has left #openstack-swift | 21:54 | |
| openstackgerrit | Mauricio Lima proposed openstack/swift: Uncomment [filter: keystoneauth] and [filter: authtoken] sessions https://review.openstack.org/185755 | 22:00 |
| *** annegentle has quit IRC | 22:13 | |
| *** kutija has quit IRC | 22:14 | |
| *** wbhuber has quit IRC | 22:21 | |
| clayg | CORS light - lawl | 22:23 |
| morganfainberg | notmyname: Fernets! | 22:26 |
| notmyname | morganfainberg: if I didn't know better, I'd say that openstack likes to drink ;-) | 22:27 |
| morganfainberg | notmyname: well i mean, we are moving to flask for keystone so we have somewhere to put the fernet | 22:28 |
| mattoliverau | morning all! | 22:29 |
| clayg | morning mattoliverau | 22:32 |
| notmyname | yo | 22:34 |
| *** annegentle has joined #openstack-swift | 22:36 | |
| peluse | mornin' | 22:40 |
| *** km has joined #openstack-swift | 23:04 | |
| *** chlong has joined #openstack-swift | 23:09 | |
| *** annegentle has quit IRC | 23:11 | |
| *** asettle has joined #openstack-swift | 23:12 | |
| *** chlong has quit IRC | 23:15 | |
| openstackgerrit | Merged openstack/swift: EC: don't 503 on marginally-successful PUT https://review.openstack.org/180795 | 23:16 |
| *** annegentle has joined #openstack-swift | 23:17 | |
| *** jrichli has quit IRC | 23:18 | |
| *** kei_yama has joined #openstack-swift | 23:18 | |
| *** annegentle has quit IRC | 23:29 | |
| *** annegentle has joined #openstack-swift | 23:33 | |
| *** jamielennox is now known as jamielennox|away | 23:42 | |
| *** chlong has joined #openstack-swift | 23:44 | |
| *** annegentle has quit IRC | 23:45 | |
| *** annegentle has joined #openstack-swift | 23:50 | |
| *** ho has joined #openstack-swift | 23:50 | |
| notmyname | clayg: turns out that doing the spawn_n for container listing updates works. until you run tests. then you get a whole lotta fun | 23:56 |
| mattoliverau | notmyname: cool your giving that a go, awesome! One take away from summit.. container listings or object listings, I thnk I need to change my vocab :) | 23:59 |
| notmyname | mattoliverau: heh | 23:59 |
| notmyname | mattoliverau: it's whichever one I mean | 23:59 |
| clayg | lol | 23:59 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!