| *** openstackgerrit has joined #openstack-watcher | 02:27 | |
| openstackgerrit | Merged openstack/python-watcherclient master: Use upper-constraints in tox runs https://review.opendev.org/658567 | 02:27 |
|---|---|---|
| openstackgerrit | Merged openstack/watcher master: Remove watcher.openstack.common=WARN from _DEFAULT_LOG_LEVELS https://review.opendev.org/658399 | 02:38 |
| openstackgerrit | Merged openstack/watcher-tempest-plugin master: check status when live migration https://review.opendev.org/658725 | 02:38 |
| openstackgerrit | Merged openstack/watcher-tempest-plugin master: Add tempest voting https://review.opendev.org/656460 | 02:38 |
| openstackgerrit | Merged openstack/watcher master: allow building docs without ceilometer client https://review.opendev.org/658802 | 02:38 |
| openstackgerrit | licanwei proposed openstack/watcher-specs master: Add Placement helper https://review.opendev.org/656448 | 03:04 |
| *** adiantum has joined #openstack-watcher | 05:21 | |
| openstackgerrit | chenker proposed openstack/watcher master: Update Sphinx requirement https://review.opendev.org/658965 | 05:41 |
| openstackgerrit | licanwei proposed openstack/watcher master: update api version history https://review.opendev.org/658966 | 05:43 |
| *** adiantum has quit IRC | 05:55 | |
| openstackgerrit | chenker proposed openstack/watcher master: Update Sphinx requirement https://review.opendev.org/658965 | 06:05 |
| openstackgerrit | licanwei proposed openstack/watcher master: Update migration notification https://review.opendev.org/658973 | 07:15 |
| *** tssurya has joined #openstack-watcher | 09:42 | |
| openstackgerrit | chenker proposed openstack/watcher-specs master: Spec for support-keystoneclient-option.rst https://review.opendev.org/659008 | 09:52 |
| openstackgerrit | chenker proposed openstack/watcher master: support-keystoneclient-option https://review.opendev.org/658102 | 10:12 |
| openstackgerrit | chenker proposed openstack/watcher master: support-keystoneclient-option https://review.opendev.org/658102 | 10:44 |
| openstackgerrit | chenker proposed openstack/watcher master: support-keystoneclient-option https://review.opendev.org/658102 | 10:46 |
| openstackgerrit | chenker proposed openstack/watcher master: support-keystoneclient-option https://review.opendev.org/658102 | 11:20 |
| openstackgerrit | chenker proposed openstack/watcher master: Fix_inappropriate_name https://review.opendev.org/659042 | 11:28 |
| *** josecastroleon has quit IRC | 11:35 | |
| openstackgerrit | Dantali0n proposed openstack/watcher master: [WIP] Grafana proxy datasource to retrieve metrics https://review.opendev.org/649341 | 12:38 |
| openstackgerrit | chenker proposed openstack/watcher master: Fix_inappropriate_name https://review.opendev.org/659042 | 12:38 |
| *** mriedem has joined #openstack-watcher | 12:49 | |
| openstackgerrit | Dantali0n proposed openstack/watcher master: [WIP] Grafana proxy datasource to retrieve metrics https://review.opendev.org/649341 | 13:09 |
| openstackgerrit | chenker proposed openstack/watcher master: Use base_strategy's add_action_migrate method https://review.opendev.org/659062 | 13:35 |
| mriedem | Dantalion: so at cern you guys use https://github.com/cernops/collectd-puppet to collect metrics on the compute nodes rather than use something like ceilometer-agent right? | 13:52 |
| openstackgerrit | chenker proposed openstack/watcher master: Use base_strategy's add_action_migrate method https://review.opendev.org/659062 | 13:55 |
| Dantalion | mriedem: Yes metrics are either collected via collectd or using lemon although the later is being deprecated as we speak | 14:07 |
| Dantalion | mriedem: Collected metrics are fed into several different types of databases but mostly influxdb and these databases are configured into grafana | 14:07 |
| Dantalion | mriedem: we collect metrics both for compute nodes and for instances although the typically end up in different databases for scalability reasons | 14:08 |
| mriedem | b/c you've got tons more instances than nodes, right | 14:14 |
| Dantalion | mriedem: Yes and scaling horizontally with influxdb on the same databases has an expensive license fee while setting up multiple independent instances of influx does not | 14:15 |
| mriedem | right now we've still got ceilometer-agent on all the compute nodes which is a major perf issue, so looking at alternatives to that | 14:15 |
| mriedem | and we've got all this data in mongodb right now, but there isn't a clean way it looks like to use mongodb as a data source for grafana | 14:15 |
| Dantalion | I can imagine, in general collectd has been working very well for us and is lightweight altough on the influx side that requires some more grunt to work properly. | 14:23 |
| openstackgerrit | Dantali0n proposed openstack/watcher-specs master: Spec for Grafana proxy datasource https://review.opendev.org/659099 | 15:17 |
| openstackgerrit | Dantali0n proposed openstack/watcher master: [WIP] Grafana proxy datasource to retrieve metrics https://review.opendev.org/649341 | 15:49 |
| mriedem | Dantalion: has cern ever thought/talked about sharding watcher per cell? i'm not sure how feasible that is, just wondering if it would help scale watcher's db and mq if it were isolated per cell rather than global. | 15:50 |
| *** openstackgerrit has quit IRC | 15:54 | |
| *** tssurya has quit IRC | 17:06 | |
| *** openstackgerrit has joined #openstack-watcher | 17:25 | |
| openstackgerrit | Dantali0n proposed openstack/watcher-specs master: Spec for Grafana proxy datasource https://review.opendev.org/659099 | 17:25 |
| openstackgerrit | Matt Riedemann proposed openstack/watcher master: Enhance the collector_plugins option help text https://review.opendev.org/659133 | 18:19 |
| mriedem | i'm a little confused about this change to the compute model collector based on audit scope https://review.opendev.org/#/c/640585/10/watcher/decision_engine/model/collector/nova.py@180 - on startup of the decision-engine service, it looks like it will never pre-load the data model even though the spec seems to imply that it should: https://review.opendev.org/#/c/641606/2/specs/stein/approved/scope-for-watcher-datamodel.rst | 18:43 |
| mriedem | doesn't this mean that the compute cdm is always built fresh for every compute-related audit? | 18:43 |
| mriedem | regardless of whether or not the audit is scoped | 18:44 |
| mriedem | also, i know i'm really late to the part of this change in stein, but i've left several comments about potential problems in the compute cdm builder https://review.opendev.org/#/c/640585/ | 20:02 |
| mriedem | at the least there are performance inefficiencies in using the compute API | 20:02 |
| mriedem | i tried to summarize in the ML | 20:59 |
| openstackgerrit | Matt Riedemann proposed openstack/watcher master: Remove dead code from NovaClusterDataModelCollector https://review.opendev.org/659175 | 21:49 |
| openstackgerrit | Matt Riedemann proposed openstack/watcher master: WIP: Profile the nova CDM builder code https://review.opendev.org/659183 | 23:10 |
| *** irclogbot_1 has quit IRC | 23:45 | |
| *** irclogbot_3 has joined #openstack-watcher | 23:49 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!