15:30:05 #startmeeting Performance Team 15:30:06 Meeting started Tue Aug 29 15:30:05 2017 UTC and is due to finish in 60 minutes. The chair is DinaBelova. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:30:07 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:30:09 The meeting name has been set to 'performance_team' 15:30:39 hello ilyashakhat o/ 15:30:54 tovin07 o/ 15:30:54 hi! 15:31:00 o/ 15:31:19 okay, so there are three of us so far :) 15:31:33 hi all o/ 15:31:38 rajulk o/ 15:31:47 hi rajulk 15:31:55 okay, so let's probably get started 15:31:59 #topic Action Items 15:32:10 first of all I want to say sorry for my action item 15:32:24 i was not able to review current patches in osprofiler 15:32:41 due to the high load on other internal tasks 15:32:47 so lemme keep it for the next time as well 15:33:14 #action DinaBelova review all patches in osprofiler 15:33:46 tovin07 - any updates from you side? On reviewing osprofiler bugs on the LP ? 15:33:59 there are a few :D 15:34:20 mostly on bugs and patches of ilyashakhat 15:34:21 o/ 15:34:28 hi rcherrueau 15:34:36 long time :D 15:34:48 hi tovin07, yep 15:35:28 #link https://bugs.launchpad.net/osprofiler/+bug/1706299 15:35:30 Launchpad bug 1706299 in osprofiler "OSProfiler does not catch DB error events" [Undecided,Confirmed] 15:35:58 (hi all :) ) 15:36:05 msimonin o/ 15:36:07 o/ 15:36:23 tovin07 please go ahead, sorry 15:36:31 and this one 15:36:34 #link https://review.openstack.org/#/c/492154/ 15:36:34 1706299 requires some work, we are waiting for oslo.db release 15:36:46 hi msimonin 15:37:14 #link https://review.openstack.org/#/c/492498/ 15:37:31 besides 15:37:39 there’a a patch from me 15:37:52 #link Filter for OSprofiler output https://review.openstack.org/#/c/496543/ 15:38:05 msimonin, you should take a look at it 15:38:41 thanks, I just added myself :) 15:38:49 496543 is a very first PoC for your case 15:39:31 that’s all from me for reviewing 15:39:59 tovin07 thanks for the update 15:40:00 #topic Current status 15:40:01 tovin07 let's proceed with current status 15:40:09 tovin07 please start :) 15:40:25 hmm, not much from last week 15:40:48 I publish a patch for filtration 15:41:25 working on monasca log API as an OSprofiler driver (forgot to publish it, will publish soon) 15:41:47 rajulk, do you have any news from your side? 15:41:58 tovin07 ack 15:42:21 for your blueprints? 15:42:22 nothing much. still working to merge the changes to osprofiler for sync 15:42:40 async call* 15:42:50 rajulk thanks 15:43:03 ilyashakhat anything to share? 15:43:28 posted some patches on code removal :) 15:43:52 now osprofiler is independent from oslo.messaging and doesn't require os-auth vars 15:44:04 ilyashakhat yeah, sorry :( this week was a bit crazy after returning from sick leave - will review for sure 15:44:37 they are merged already 15:44:45 thanks to tovin07 and Vipin 15:44:47 ilyashakhat well, you have more on review :D 15:45:51 that's all I have for now :) 15:46:05 msimonin rcherrueau okay, so today we have you folks :) anu update from your side? Any news on distributed messaging routing you're working on? 15:46:47 yes 15:46:59 There are two directions actually 15:47:07 distributed messaging and distributed databases 15:47:47 regarding distributed messaging we started a test plan for this 15:48:00 #link https://review.openstack.org/#/c/491818/ 15:48:09 it's still WiP 15:48:43 but feel free to check it (especially if you have some ideas regarding the resilience test cases) 15:48:45 :) 15:49:21 msimonin gotcha 15:49:43 In parallel, we need to code some specifix tool to experiment with qpid dispatch 15:49:58 the same it's progressing 15:50:16 rcherrueau: want to speak about distributing the database? 15:50:23 yep 15:50:49 On my side, I am working (with an intern) on making OpenStack supports CockroachDB 15:51:12 CockroachDB is a *New*SQL Database 15:51:39 i.e., A database that scales and still offers ACID transactions. 15:52:08 We have a POC that deploys Keystone over CockroachDB 15:52:20 #link https://github.com/BeyondTheClouds/openstack-cockroachdb-dev 15:52:34 You can try it. You only need vagrant 15:52:48 then it deploys cockroacdb and keystone using devstack 15:53:04 I will make an announcement on the ml soon 15:53:32 Next step is doing performance analysis on CockroachDB vs Galera 15:54:00 as a spoiler - did you face any issues on making Keystone work with it? 15:54:30 ilyashakhat: not that much. We have to patch one file in oslo.db 15:54:51 sounds cool :) 15:54:54 Especially because that contains code that is backend specific 15:55:20 thanks rcherrueau msimonin 15:55:25 there a lots of error-handling specific things 15:55:27 and we also have to add some python decorators into keystone, but only few 15:55:48 This will be fun to deploy at scale ;) 15:56:04 true :D 15:56:55 okay, it looks like we can proceed to the open discussion 15:57:00 #topic Open Discussion 15:57:20 ilyashakhat msimonin rcherrueau tovin07 rajulk - anything to share, that was not yet mentioned? 15:58:04 yep 15:58:06 I have a question - do we plan to have sort of plan for the next release? 15:58:13 ilyashakhat: +1 15:59:12 i’m thinking of separating error traces (traces that contain a span with error/exception information) from normal trace. 15:59:20 in OSprofiler 15:59:58 ilyashakhat msimonin I suspect we do not have specific items so far - talking about osprofiler or test plans. Let's probably gather them to one place till next meeting? 16:00:07 DinaBelova, could you make an etherpad to list our ideas 16:00:09 yep 16:00:16 yep, one sec 16:00:23 that is a small change in idea, however, need to modify drivers 16:00:26 ilyashakhat: +1 too 16:00:28 tovin07 can you share the idea behind doing that 16:01:58 #action ilyashakhat msimonin rcherrueau tovin07 rajulk DinaBelova add ideas to https://etherpad.openstack.org/p/performance-team-q-release-ideas 16:02:42 rajulk: the idea is quite simple, we can use OSprofiler to trouble-shoot errors in our OpenStack services. I just want to separate error traces out of normal trace and store it in other place —> easy for query. i.e normal traces in redis db 0, error trace in redis db 1 16:02:50 DinaBelova: thanks 16:03:41 tovin07 got it. thanks 16:03:59 tovin07 I'm ok with this idea, but only if it's going to be an optional - not all people are ready to have multiple DBs for profiling and not everybody needs it 16:04:16 yes 16:04:33 it should be an option only 16:04:44 cool, thanks 16:05:31 okay, anything else? 16:05:45 none :d 16:05:58 it looks like we're done for todays 16:06:04 thank you folks for participating 16:06:10 thanks! 16:06:11 have a nice day/evening 16:06:13 #endmeeting