15:00:34 <witek> #startmeeting monasca
15:00:35 <openstack> Meeting started Wed Nov 22 15:00:34 2017 UTC and is due to finish in 60 minutes.  The chair is witek. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:00:36 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:00:39 <openstack> The meeting name has been set to 'monasca'
15:00:50 <witek> hello everyone
15:00:59 <cbellucci> o/
15:01:13 <jgr> Hello
15:01:17 <haruki> hello
15:01:18 <sgrasley1> Hello
15:01:27 <christian_op5> Hello
15:01:46 <witek> I hope the new time works for everyone
15:02:16 <witek> sorry once again for the short notice
15:02:53 <witek> #topic Slack templating
15:03:32 <witek> hi dougsz
15:03:35 <tobiajo> ho
15:03:37 <witek> we just started
15:03:38 <dougsz> hey all
15:04:09 <witek> dougsz: your stage
15:04:29 <dougsz> Any opinions on https://review.openstack.org/#/c/519630/ ?
15:04:47 <dougsz> An effort to make the slack channel notifier a bit more user friendly
15:04:56 <witek> +1
15:05:04 <pilgrimstack> o/
15:05:27 <witek> it's definitely adding value
15:05:45 <witek> would it make sense to add default template?
15:06:03 <dougsz> I think it would
15:06:04 <witek> similar as for Jira plugin?
15:06:22 <dougsz> Yep, I can add that
15:06:43 <dougsz> So change the default behaviour to use a template, or just provide a default template?
15:08:20 <witek> just adding a default template would be enough
15:08:37 <witek> that's how Jira plugin is implemented, I guess
15:08:58 <dougsz> thanks - i'll take a closer look at that.
15:09:18 <witek> one formal thing: we don't use blueprints anymore
15:09:49 <witek> could you please move the blueprint to StoryBoard?
15:09:57 <dougsz> yep - sorry about that. I'll move it over.
15:10:19 <witek> thanks
15:11:16 <witek> if you're interested in notifications, you might want to have a look at other changes needing care
15:11:28 <witek> https://review.openstack.org/435136
15:11:34 <witek> https://review.openstack.org/461917
15:12:22 <witek> they are little older, but would be nice to have them in
15:12:39 <dougsz> Thanks - I'll take a look at those.
15:13:09 <witek> thanks dougsz
15:13:36 <dougsz> So that's my question answered
15:13:59 <witek> #topic Grafana - Horizon integration
15:14:34 <witek> jgr: that's yours
15:14:43 <jgr> This one is basically me being a pain about the reviews that will make it work
15:14:56 <witek> :)
15:15:10 <jgr> https://github.com/servicecatalog/crowbar-escm/commits/master
15:15:21 <jgr> Oops, wrong URL
15:15:45 <jgr> https://review.openstack.org/#/c/513366/
15:15:55 <jgr> https://review.openstack.org/518017
15:16:18 <jgr> witek is already looking at them but they could use another pair of eyes
15:16:35 <witek> I'm wondering if we couldn't make Grafana 4 link style default
15:17:01 <jgr> With both of these in it will be possible to use the Horizon session for monasca-api authentication in monasca-grafana-datasource
15:17:54 <jgr> Hmm. Depends on how many folks use the old style patched Grafana 1.5 still...
15:18:06 <witek> so we can use Grafana anonymous login and redirect from Horizon
15:18:23 <witek> jgr: I think minority
15:19:00 <jgr> Anonymous login, and in the future Keystone (for creating persistent dashboards on a per-user basis)
15:20:38 <jgr> I can switch the default over, but I'd rather not break existing installations (even if they're a minority).
15:21:27 <witek> we could also update devstack plugin
15:21:29 <witek> ?
15:22:03 <jgr> That uses Grafana 4 with Token right now, correct?
15:22:20 <witek> right
15:22:39 <jgr> I can do that, but I'd like to do it in a separate review if possible.
15:22:52 <witek> fine
15:23:11 <jgr> I'm under some pressure to get this into our RPMs and the dead line for that is getting a bit close...
15:23:25 <witek> reviews welcome!
15:23:46 <dougsz> I'll have to go and have a read of background
15:23:47 <jgr> But I'll create a task for it and take care of the Devstack plugin once the RPMs are there.
15:24:07 <dougsz> Will it help move away from the forked grafana with keystone support?
15:25:02 <witek> it's a workaround I would say, because you cannot use Grafana's user management
15:25:11 <jgr> dougsz: no, that's a separate issue. This is more of a stopgap measure.
15:25:27 <dougsz> ah ok
15:25:48 <jgr> dougsz: The problem with using the Horizon session is this approach's feature of not requiring a Grafana login.
15:25:56 <jgr> dougsz: Instead, it simply uses the Horizon cookie.
15:26:08 <dougsz> ah I see
15:26:30 <jgr> dougsz: by not requiring a Grafana login there's no Grafana account for every Keystone user where they could create and store their own dashboards...
15:26:59 <witek> and own datasources, and apps...
15:27:20 <dougsz> yeah, that would be a problem for us, we're particularly interested in tenant monitoring
15:28:54 <jgr> Yeah. That definitely needs Keystone authentication in Grafana.
15:29:07 <witek> dougsz: if you need instructions on testing, contact jgr, he's always on irc :)
15:29:22 <dougsz> ok, thanks!
15:29:59 <witek> #topic libvirt metrics aggregations
15:30:39 <jgr> That one's mine as well
15:30:50 <jgr> Let's start with a pretty picture: http://btw23.de/tmp/libvirt_graphs.png
15:31:07 <jgr> To make that pretty picture happen https://review.openstack.org/521909 is needed :-)
15:31:26 <jgr> (because Grafana sucks at aggregating data with too many dimensions...)
15:32:13 <jgr> Right now we have a vm.host_alive_status value for every VM on every compute node.
15:33:23 <jgr> vm.host_alive_status only has a few possible values, so from an operator's point of view it really makes sense to have separate metrics for each possible value and aggregate these on a per-compute-node basis.
15:34:01 <jgr> That allows you to e.g. graph how many running and how many crashed VMs you have on each compute node.
15:35:01 <witek> sounds like another good change
15:35:19 <jgr> Yep...and it is another change that could use eyes on it :-)
15:35:55 <dougsz> Interesting - we have something similar focused on baremetal https://www.stackhpc.com/baremetal-cloud-capacity.html
15:37:15 <witek> thanks for the link, will read it after the meeting
15:38:27 <witek> jgr: libvirt related, have you observed problems reported here:
15:38:32 <witek> https://storyboard.openstack.org/#!/story/2001303
15:38:39 <nseyvet> sorry I forgot the time
15:39:25 <haruki> > jgr I'll try to check https://review.openstack.org/#/c/521909/
15:39:36 <jgr> witek: no, but I'm not using monasca-setup for libvirt (I'm configuring it from Crowbar)
15:39:40 <witek> nseyvet: the time changed, so we have to get used to it :)
15:39:46 <jgr> haruki: thanks!
15:41:48 <witek> haruki: would you have a chance to test detection plugin as well?
15:43:37 <witek> #topic Bugfix tempest tests
15:44:21 <witek> jgu: your turn
15:45:11 <jgu> I just wanted to buy folks if you have a few minutes for reviews on this quick fix in the tempest test data setup.
15:45:19 <jgu> s/buy/bug
15:45:46 <witek> jgu: could you add some logs to the case when the tempest fail?
15:46:14 <witek> Story would be the best place ;)
15:46:18 <jgu> yes, where to add to?
15:46:29 <jgu> i see file a story boardt ticket first?
15:46:52 <jgu> sure I can do that
15:46:54 <witek> would be great
15:46:56 <witek> thanks
15:47:46 <jgu> it shows up in the java gate job failure for influxdb test case too
15:48:41 <haruki> witek, jgr: sorry for rate replay, I'd like to check monasca-setup  and return status of libvirt so I try to check (with Koji I hope:)).
15:48:43 <witek> always, or just sporadically?
15:48:56 <witek> haruki: thanks
15:49:01 <jgu> https://review.openstack.org/#/c/520869/ (fix some tempest test data setup checks)
15:49:49 <jgu> sporadically on my local devstack... but seems to be consistent for the Influx Java patch gate job for the cassandra patches
15:50:32 <witek> yes, I have also observed sporadic tempest failures
15:51:30 <jgu> the test data setup check wasn't complete enough so there is intermittent timing issue
15:52:42 <witek> I have seen errors in test_create_metrics with empty dimensions queried
15:54:42 <witek> to sum up, we need more reviews
15:54:51 <witek> I will try myself to review more
15:55:01 <jgu> witek: thanks
15:55:31 <witek> anything else for closing?
15:55:31 <nseyvet> anyone using monasca-agent in production?
15:55:51 <nseyvet> and how does one usually deploy the agent + plugins?
15:57:41 <jgr> We deploy it in a hybrid manner currently.
15:58:05 <witek> Helion distibution is used in production as far as I know
15:58:06 <jgr> We use multiple monasca-setup on the machine with the Monasca services.
15:58:28 <jgr> That takes multiple consecutive monasca-setup invocations.
15:58:56 <jgr> On the compute nodes we currently generate the check plugin configurations from our own configuration management.
15:59:47 <jgr> On the Openstack controller we run monasca-setup as well.
16:00:26 <witek> I have to close the meeting
16:00:44 <witek> bye everyone
16:00:58 <haruki> thank you
16:00:59 <jgu> bye
16:01:03 <jgr> See you
16:01:05 <witek> #endmeeting