15:00:29 <rbrndt> #startmeeting monasca
15:00:30 <openstack> Meeting started Wed Apr 26 15:00:29 2017 UTC and is due to finish in 60 minutes.  The chair is rbrndt. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:00:31 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:00:35 <openstack> The meeting name has been set to 'monasca'
15:00:39 <rbrndt> Hello
15:00:42 <witek> hi
15:01:24 <koji> hello
15:01:31 <haruki> hello
15:01:52 <hosanai> o/
15:02:02 <shinya_kwbt> o/
15:02:03 <rbrndt> I see we have something on the agenda this time, specifically the keystone uwsgi item
15:02:20 <Neptu> hello
15:02:33 <witek> yes, it's a little more than this actually
15:02:37 <Neptu> I have one more item to discuss
15:03:03 <Neptu> Jose Romero from op5
15:03:30 <rbrndt> Alright, well perhaps we can start with the keystone item on the agenda first, which is also from you?
15:03:40 <Neptu> yep
15:03:47 <rbrndt> Take it away
15:04:15 <witek> Keystone is from me
15:04:35 <rbrndt> Sorry, I was just going off who entered it on the agenda
15:04:39 <witek> we have investigated failing gate jobs in the last week
15:04:59 <witek> and found out that devstack moved to using uwsgi
15:06:28 <witek> keystone and other services are hosted on apache now with port 80 and aliases
15:07:38 <witek> Tomasz and Artur try to adapt the devstack plugin to reflect that
15:08:07 <witek> but we still have errors on Java tempest tests
15:09:43 <witek> so the current status is, that for about two weeks now we cannot merge anything because of failing tempests
15:10:07 <witek> python looks good now, Java still not ready
15:10:50 <rbrndt> Do you have some output or a failing review with the errors from the java side? Would be good to have on the meeting record
15:10:56 <witek> I was wondering if we could set Java tempests as non-voting until we solve the issue?
15:12:54 <rbrndt> I'm generally against turning off tests, because it is very easy to never turn them back on.
15:14:03 <rbrndt> That being said, we have been planning on moving away from java, and most new features are not planned to be added to the java side
15:15:12 <rbrndt> Anyone else have an opinion?
15:16:41 <witek> I generally agree with you about switching off the tests not being goot, but it is quite a blocker at the moment
15:18:14 <rbrndt> Alright, well I think Roland should probably way in on this, but if you want to submit a review to change the tests to non-voting perhaps we can discuss more there
15:18:28 <Neptu> for my prespective as new java is only necesary for legacy then... we had this problem this week and we checkout a a previos version of devstack
15:19:20 <Neptu> which is not best but helps to have it all working
15:20:12 <rbrndt> The java has long been a bit of a thorn in the side when dealing with devstack and CI, so dealing with it in a long term solution could be very helpful
15:21:06 <rbrndt> Any last thoughts on that topic?
15:22:09 <witek> I will add the change to project-config and let's discuss in review
15:22:32 <rbrndt> Sounds good. That covers the agenda then. Neptu, you had another topic?
15:22:37 <Neptu> yes
15:22:54 <rbrndt> ty witek
15:23:01 <Neptu> we just got started with monasca here at op5 to use it as core of one of the new products
15:23:16 <Neptu> so we have been analizing the different parts
15:23:23 <Neptu> and we see an issue with influxdb
15:23:36 <Neptu> since after version 0.9 the cluster side is paid version
15:23:59 <Neptu> so we cannot really scale up monasca without paying licences
15:24:07 <Neptu> and I was wondering ig there is something on the roadmap
15:24:40 <Neptu> to add support for cluster of the metridb throw another persistor to ... other metric db
15:25:29 <rbrndt> That is something that has been a bother for a while. We've explored several options, with cassandra being the most promising
15:26:03 <rbrndt> However, there are issues with the data model required for cassandra at higher metrics loads. It seems not viable for large scale.
15:26:22 <Neptu> graphite cairos maybe?
15:27:07 <Neptu> donno i just want to undestand that is not an stopper and we could integrate with something else semi-easily
15:27:27 <Neptu> we are 3 ingeneers now in the project and growing but we lack the experience on monasca
15:27:52 <Neptu> but we could code another persistor if its a must
15:28:13 <rbrndt> Let me see if I can find some documentation of our research into cassandra. Modifying the persister should be the easy part
15:28:36 <Neptu> i think so
15:28:42 <rbrndt> The API needs a new respository as well which is a little more involved, but still very doable
15:28:59 <rbrndt> The real key is the data model
15:29:25 <rbrndt> The API is setup to slice and dice the data a lot of different ways, which doesn't work well in cassandra
15:29:38 <rbrndt> and most other cassandra based options
15:30:05 <witek> Neptu: which data rates are you aiming, how many metrics/s do you want to store?
15:30:05 <rbrndt> We had vertica integrated for a while, but that is also a paid option
15:30:13 <Neptu> so you kind of build different representation of the same data
15:30:25 <rbrndt> Yes
15:30:32 <Neptu> witek: conceptually management is talking about 9M /minute
15:30:38 <rbrndt> Good question wwitek
15:30:44 <Neptu> maybe more
15:31:15 <Neptu> I know from some other companies had that done with monasca here in stockholm
15:31:29 <Neptu> so should be duable with influx
15:31:37 <Neptu> but maybe we do not have influx any more
15:31:37 <Neptu> ...
15:31:49 <rbrndt> 9M /minute is about ...
15:32:12 <rbrndt> 150,000 / sec
15:32:18 <Neptu> 9M/minute is a buzzword here in the office
15:32:23 <rbrndt> that's generally higher than we've run monasca before
15:32:23 <witek> InfluxDB reported 900k/s on single node
15:32:37 <Neptu> metrics
15:32:53 <Neptu> or Kbps
15:32:55 <rbrndt> In our testing, influx has been significantly lower than the 900k /s
15:33:13 <witek> metrics
15:33:35 <Neptu> i mean at this point i have no requirement of how many instances i need to spin to get the job done
15:33:58 <Neptu> the question is with a single machine is a bottlenek for sure
15:34:03 <witek> https://docs.influxdata.com/influxdb/v1.1/administration/differences/
15:34:28 <witek> for HA you can use influxdb-relay or kafka
15:34:54 <rbrndt> In deed. I think the simplest answer (though not the easiest) is if you can find a replacement, we'd love to have a non-paid one.
15:35:01 <Neptu> but influx is used only for storage or aswell runs statistics and searches...
15:35:22 <Neptu> i have experience with mongodb but might not be the best one
15:35:41 <Neptu> i would like to further talk with someone that has the experience
15:35:47 <Neptu> so we can maybe solve this somehow
15:35:57 <Neptu> we can allocate resources probably for it
15:36:04 <Neptu> is a question of proper guidance
15:36:25 <Neptu> so we do code and recode 10 times before is workcable
15:36:34 <rbrndt> I would be happy to talk to you about it in more detail sometime. I can also raise this to roland who probably has the data from our week investigating cassandra
15:36:49 <Neptu> awesome
15:37:19 <Neptu> jromero@op5.com
15:37:26 <Neptu> pop me a mail and we continue from there
15:37:32 <rbrndt> Sounds good.
15:37:39 <witek> please put me on cc
15:37:46 <Neptu> yep
15:37:55 <Neptu> will be nice if you had a #monasca channel
15:38:02 <Neptu> im always on irc
15:38:08 <witek> #openstack-monasca :)
15:38:12 <Neptu> am
15:38:13 <Neptu> good
15:38:23 <Neptu> :)
15:38:23 <rbrndt> You can usually find me in the monasca channel if you have quick questions
15:38:39 <Neptu> yes i will add it to my znc
15:38:55 <rbrndt> Alright, I think that wraps that up for now. Anyone else have any topics? Open floor
15:38:57 <Neptu> thanks so we continue from there
15:39:23 <witek> I wanted to ask, who will attend the Summit?
15:39:37 <sc> I'm going
15:39:46 <rbrndt> I was mistaken in the last meeting, Roland will be attending as usual
15:40:20 <jamesg_> Dan and myself (from Suse) will be going
15:41:05 <witek> sc: I'm sorry, I cannot 'decode' your nick :(
15:41:23 <sc> sc@linux.it aka stefano.canepa@hpe.com
15:41:30 <witek> :)
15:41:36 <witek> sorry
15:42:02 <sc> witek: NP, if you need help for the training session just drop me an email
15:42:21 <witek> Cristiano and me will join from Fujitsu EST
15:42:43 <witek> sc: thanks, I'm still working on material
15:42:49 <haruki> Witek: as you know, I will join from Fujitsu Japan
15:43:33 <sc> witek: I'm starting to prepare my slides tomorrow ;-)
15:43:41 <witek> :)
15:43:53 <rbrndt> Hope it goes well
15:44:19 <rbrndt> Ok, well any more topics?
15:44:30 <rbrndt> if not we can all have 15 minutes back
15:44:59 <witek> thank you Ryan
15:45:21 <rbrndt> Thanks everyone
15:45:28 <sc> see you
15:45:37 <rbrndt> #endmeeting