14:01:01 #startmeeting monasca 14:01:02 Meeting started Wed Sep 6 14:01:01 2017 UTC and is due to finish in 60 minutes. The chair is rhochmuth. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:01:03 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:01:06 The meeting name has been set to 'monasca' 14:01:09 o/ 14:01:12 hello 14:01:14 o/ 14:01:16 o/ 14:01:31 hi everyone 14:01:38 o/ 14:01:45 witek is out today 14:01:58 so, it guess it is back to me this week 14:02:13 no agenda items have been posted 14:03:02 do we have some items to add to the agenda 14:03:11 possibly some updates from last week 14:04:00 hi 14:04:05 hi 14:04:09 I've just got a thought about the mid-cycle meeting. I think those of us involved in Ceilosca would like to join if it isn't going to derail things 14:04:20 hi tomiago and BeHavee 14:04:29 hello 14:04:43 joadavis: sounds like a good topic to discuss 14:04:43 hello 14:04:51 i've been out of the loop on that for a while 14:05:23 I would like to discuss updating the kafka-> persister python library 14:05:34 Yeah, we definitely could use some discussion around architecture going forward, especially as Ceilometer project is pivoting to be very Gnocchi directed 14:06:40 I hope we can continue to improve metering related functions for Monasca users 14:06:48 hi 14:06:56 ok, so let's start with the kafka library first 14:07:06 #topic afka-> persister python library 14:07:28 nseyvet: you have the floor 14:07:55 First, I would liek to know how to go forward and submit updating the python kafka library to a later version? 14:08:17 It seems to be using an old version, with offset manipulations in ZK atm 14:08:36 we would like to move it forward to kafka 0.10+ 14:08:41 there were problems with the newer versions with stability and performance 14:08:52 i don't know if those have been resolved 14:08:58 have you tests 14:09:22 someone in our team made a quick benchmark and performance was about 20% higher 14:09:37 but we have not tested for stability 14:09:52 with pure Python implementation or using librdkafka? 14:09:58 we basically forked the library in monasca-common 14:11:01 because upstream openstack wanted to use the newever version and it wouldn't work for us 14:11:29 also one person in our team was looking shortly on upgrading it and he said that going 0.9 to 2.6 is like using completely new library... 14:11:44 yes, that is also our finding 14:12:15 so, how to move forward? Create a "specifications"? 14:12:42 I dont have the exact info as to which lib was looked at.. sorry 14:12:53 creating a specification would be a good start 14:13:38 if newer Pykafka is so different then maybe https://github.com/confluentinc/confluent-kafka-python also would be good to look at 14:13:56 i think the confluent library is a good one to use 14:14:07 Ok. I take note. And I agree on the confluent one. 14:14:14 however, the problem is that is requires C code to be built as well 14:14:29 so we will submit a specification. 14:14:33 so install/deployment is more complicated 14:14:47 hmmm, true 14:15:01 there was another library that joe keen was looking at 14:15:44 forgot the name of it 14:15:47 newer PyKafka have two backends: pure Python and faster librdkafka 14:15:53 that is the one 14:16:01 thx Dobroslaw_ 14:16:03 (same as in confluent-kafka-python) 14:16:51 are there any C dependencies in the python persister today? Because when I look at the docker images, it lists some 14:17:58 like "linux-headers libxml2-dev libxslt-dev" etc 14:19:19 hi, all 14:21:00 We will submit a specification. Then we see. 14:21:06 thanks for info. 14:21:15 ok 14:21:27 thanks nseyvet 14:22:23 what should the next topic be 14:22:30 maybe go over to pykafka in monasca-common? we could set pure python as default, and make it possible to use librdkafka if one want 14:23:24 tobiajo: good idea, i think that is similar to what joe proposed 14:23:26 too 14:25:04 Ceilosca? 14:25:12 #topic ceilosca 14:25:40 Sure. :) Do we want to put a slice of time on the midcycle agenda for Ceilosca? 14:26:53 i lost my link to the etherpad for the mid cycle 14:27:04 some topics could include changes in Ceilometer project (like deprecation of Ceilometer API), other sources of metrics (like Prometheus) that could be funneled in to Monasca... 14:27:13 but, yes, someone should add that as a agenda item and propose a date 14:27:16 Try https://etherpad.openstack.org/p/monasca_queens_midcycle 14:27:24 thx 14:27:49 ... and other things like events and alarms that could be run through Monasca to match any feature set from the Telemetry project. 14:28:40 Are there any attendees from the telemetry project? 14:29:16 I doubt any telemetry project folks are planning to attend the virtual mid-cycle for Monasca 14:30:12 If you're proposing changes to align/match their metrics/events, wouldn't it make sense to invite them? 14:31:07 Ceilosca already matches the metrics (by matches I mean funnels ceilometer metrics into Monasca). Monasca already has some work around events. 14:31:46 We can invite them, I just doubt they have much interest 14:32:25 They might be interested in the discussion around TSDB support, as gnocchi fits the bill. 14:32:47 14:33:16 sure, i would be interested in a gnocchi discussion 14:33:45 there were some questions around metrics injest rates and query performance 14:36:02 Is there a right way to invite them? Open post to #openstack-telemetry? 14:40:28 how different are the telemetry vs monasca objectives? 14:41:20 is anyone here that can help to answer/clarify 14:42:17 I can't anyway. 14:42:33 Try a post to the mailing list to invite the telemetry team 14:43:30 thx efoley 14:43:31 I don't think I'm qualified to articulate it. :) I have opinions but I'd just get myself in trouble 14:43:46 i'll leave that to witek to handle i guess 14:44:12 I'll take an action item to come up with a focused item or two re: Ceilometer for the mid cycle meetings 14:44:22 witek is off this week IIRC 14:44:29 thx joadavis 14:47:05 so are there more topics to address for today 14:47:13 tsdb updates 14:47:14 ? 14:47:25 griddb 14:48:51 akiraYo: sure 14:49:05 #topic griddb 14:49:11 I'm working to improve DB healthcheck. 14:49:39 thx, is that the refactoring that i requested 14:50:00 yes. 14:50:18 great, i'll take a look 14:51:26 thx 14:51:53 akiraYo: did you poast a new patch set? 14:51:59 i don't see one listed 14:52:12 not yet 14:52:26 are you planning too? 14:52:31 improving unittests 14:52:37 thx 14:52:50 i thought you had posted already, just clarifying 14:53:37 ok. 14:55:12 and I got mails about requirements of TSDB 14:56:16 so, i'm going to have to close the meeting down shortly 14:56:41 any final comments or requests? 14:57:13 all good. ty 14:57:19 have a nice day 14:57:38 thanks everyone 14:57:39 thank you 14:57:47 #endmeeting