d4c8683e68
This patch consits of: * New ops interface grafana_dashboard based on the existing reactive interface * New ops http interface. * Dashboard registers grafana dashboards with grafana. These dashboards are taken from upstream ceph *1. Ideally the charm would pick these up from ceph packaging but they are currently not included. With the exception of host-details.json and hosts-overview.json the dashboards are unchanged. host-details.json and hosts-overview.json both needed changed to make the compatable with the metrics telegraf is sending. * alertmanager-service and prometheus relations using the http ops interface allowing there api endpoints to be registered with the dashboard. NOTE: ceph-mon has an existing relation with Prometheus ( ceph-mon:prometheus <-> prometheus:target ) but prometheus does not publish its api endpoint on that relation. So, the dashboard adds a prometheus:website <-> ceph-dashboard:prometheus relation. This allows the dashboard to set set-prometheus-api-host. *1 https://github.com/ceph/ceph/tree/master/monitoring/grafana/dashboards Change-Id: Ic4522cc601895c9a79489df985a6e81fa70fb9e5 |
||
|---|---|---|
| src | ||
| tests | ||
| unit_tests | ||
| .flake8 | ||
| .gitignore | ||
| .gitreview | ||
| .jujuignore | ||
| .stestr.conf | ||
| .zuul.yaml | ||
| actions.yaml | ||
| build-requirements.txt | ||
| config.yaml | ||
| LICENSE | ||
| metadata.yaml | ||
| osci.yaml | ||
| README.md | ||
| requirements-dev.txt | ||
| requirements.txt | ||
| test-requirements.txt | ||
| tox.ini | ||
Overview
The ceph-dashboard configures the [Ceph Dashboard][ceph-dashboard-upstream]. The charm is intended to be used in conjunction with the ceph-mon charm.
Usage
Configuration
See file config.yaml for the full list of options, along with their
descriptions and default values.
Deployment
We are assuming a pre-existing Ceph cluster.
Deploy the ceph-dashboard as a subordinate to the ceph-mon charm.
juju deploy ceph-dashboard
juju relate ceph-dashboard ceph-mon
Embedded Grafana Dashboards
To enable the embedded grafana dashboards within the Ceph dashboard some additional relations are needed.
juju relate ceph-dashboard:grafana-dashboard grafana:dashboards
juju relate ceph-dashboard:prometheus prometheus:website
juju relate ceph-mon:prometheus prometheus:target
juju relate ceph-osd:juju-info telegraf:juju-info
juju relate ceph-mon:juju-info telegraf:juju-info
Grafana, Telegraf and Prometheus should be related in the standard way
juju relate grafana:grafana-source prometheus:grafana-source
juju relate telegraf:prometheus-client prometheus:target
juju relate telegraf:dashboards grafana:dashboards
Grafana must be using https so either supply a certificates and key via the ssl_* charm config options or add a vault relation.
juju deploy grafana:certificates vault:certificates
Grafana should be set with the following charm options:
juju config grafana anonymous=True
juju config grafana allow_embedding=True
juju config grafana install_plugins="https://storage.googleapis.com/plugins-community/vonage-status-panel/release/1.0.11/vonage-status-panel-1.0.11.zip,https://storage.googleapis.com/plugins-community/grafana-piechart-panel/release/1.6.2/grafana-piechart-panel-1.6.2.zip"
Telegraf should be set with the following charm options:
juju config telegraf hostname="{host}"
NOTE: That is "{host}" verbatim, nothing needs to be substituted.
Currently the dashboard cannot autodect the api endpoint of the grafana service, so the end of the deployment run the following:
juju config ceph-dashboard grafana-api-url="https://<IP of grafana unit>:3000"
Enabling Prometheus Alerting
To enable Prometheus alerting, add the following relations:
juju relate ceph-dashboard:prometheus prometheus:website
juju relate ceph-mon:prometheus prometheus:target
juju relate ceph-dashboard:alertmanager-service prometheus-alertmanager:alertmanager-service
juju relate prometheus:alertmanager-service prometheus-alertmanager:alertmanager-service