This project is about running Grafana on Kubernetes with Prometheus as the datasource in a very opinionated and entirely declarative way. This allows easily operating Grafana highly available as if it was a stateless application - no need to run a clustered database for your dashboarding solution anymore!
Note that at this point this is primarily about getting into the same state as kube-prometheus currently is. It is about packaging up Grafana as a reusable component, without dashboards. Dashboards are to be defined when using this Grafana package.
This repository exists because the Grafana stack in kube-prometheus has gotten close to unmaintainable due to the many steps of generation and it's a very steep learning curve for newcomers.
Since Grafana v5, Grafana can be provisioned with dashboards from files. This project is primarily about generating a set of useful Grafana dashboards for use with and on Kubernetes using with Prometheus as the datasource.
In this repository everything is generated via jsonnet:
- The Grafana dashboard sources configuration.
- The Grafana dashboard datasource configuration, is part of the configuration, and is then simply rendered to json.
- The Grafana dashboard definitions are defined as part of the configuration. For example, dashboard definitions can be developed with the help of grafana/grafonnet-lib.
- The Grafana Kubernetes manifests with the help of ksonnet/ksonnet-lib.
With a single jsonnet command the whole stack is generated and can be applied against a Kubernetes cluster.
You need a running Kubernetes cluster in order to try this out, with the kube-prometheus stack deployed on it as have Docker installed to and be able to mount volumes correctly (this is not the case when using the Docker host of minikube).
For trying this out provision minikube with these settings:
minikube start --kubernetes-version=v1.9.3 --memory=4096 --bootstrapper=kubeadm --extra-config=kubelet.authentication-token-webhook=true --extra-config=kubelet.authorization-mode=Webhook --extra-config=scheduler.address=0.0.0.0 --extra-config=controller-manager.address=0.0.0.0
Use this package in your own infrastructure using jsonnet-bundler:
jb install github.com/brancz/kubernetes-grafana/grafana
An example of how to use it could be:
local grafana = import 'grafana/grafana.libsonnet';
{
local basic =
(grafana {
_config+:: {
namespace: 'monitoring-grafana',
},
}).grafana,
apiVersion: 'v1',
kind: 'List',
items:
basic.dashboardDefinitions +
[
basic.dashboardSources,
basic.dashboardDatasources,
basic.deployment,
basic.serviceAccount,
basic.service {
spec+: { ports: [
port {
nodePort: 30910,
}
for port in super.ports
] },
},
],
}This builds the entire Grafana stack with your own dashboards and a configurable namespace.
Simply run:
$ jsonnet -J vendor example.jsonnet
This setup is optimized to work best when Grafana is used declaratively, so when adding dashboards they are added declaratively as well. In jsonnet there are libraries available to avoid having to repeat boilerplate of Grafana dashboard json. An example with the grafana/grafonnet-lib:
local grafana = import 'github.com/grafana/grafonnet-lib/grafonnet/grafana.libsonnet';
local dashboard = grafana.dashboard;
local row = grafana.row;
local prometheus = grafana.prometheus;
local template = grafana.template;
local graphPanel = grafana.graphPanel;
local grafana = import 'grafana/grafana.libsonnet';
local grafanaWithDashboards =
(grafana
{
_config+:: {
namespace: 'monitoring-grafana',
grafana+:: {
dashboards+:: {
'my-dashboard.json':
dashboard.new('My Dashboard')
.addTemplate(
{
current: {
text: 'Prometheus',
value: 'Prometheus',
},
hide: 0,
label: null,
name: 'datasource',
options: [],
query: 'prometheus',
refresh: 1,
regex: '',
type: 'datasource',
},
)
.addRow(
row.new()
.addPanel(
graphPanel.new('My Panel', span=6, datasource='$datasource')
.addTarget(prometheus.target('vector(1)')),
)
),
},
},
},
}).grafana;
{
apiVersion: 'v1',
kind: 'List',
items:
grafanaWithDashboards.dashboardDefinitions +
[
grafanaWithDashboards.dashboardSources,
grafanaWithDashboards.dashboardDatasources,
grafanaWithDashboards.deployment,
grafanaWithDashboards.serviceAccount,
grafanaWithDashboards.service {
spec+: { ports: [
port {
nodePort: 30910,
}
for port in super.ports
] },
},
],
}If you have many dashboards and would like to organize them into folders, you can do that as well by specifying them in folderDashboards rather than dashboards.
local grafana = import 'grafonnet/grafana.libsonnet';
local dashboard = grafana.dashboard;
local row = grafana.row;
local prometheus = grafana.prometheus;
local template = grafana.template;
local graphPanel = grafana.graphPanel;
local grafana = import 'grafana/grafana.libsonnet';
local grafanaWithDashboards =
(grafana
{
_config+:: {
namespace: 'monitoring-grafana',
grafana+:: {
folderDashboards+:: {
Services: {
'regional-services-dashboard.json': (import 'dashboards/regional-services-dashboard.json'),
'global-services-dashboard.json': (import 'dashboards/global-services-dashboard.json'),
},
AWS: {
'aws-ec2-dashboard.json': (import 'dashboards/aws-ec2-dashboard.json'),
'aws-rds-dashboard.json': (import 'dashboards/aws-rds-dashboard.json'),
'aws-sqs-dashboard.json': (import 'dashboards/aws-sqs-dashboard.json'),
},
ISTIO: {
'istio-citadel-dashboard.json': (import 'dashboards/istio-citadel-dashboard.json'),
'istio-galley-dashboard.json': (import 'dashboards/istio-galley-dashboard.json'),
'istio-mesh-dashboard.json': (import 'dashboards/istio-mesh-dashboard.json'),
'istio-pilot-dashboard.json': (import 'dashboards/istio-pilot-dashboard.json'),
},
},
},
},
}).grafana;
{
apiVersion: 'v1',
kind: 'List',
items:
grafanaWithDashboards.dashboardDefinitions +
[
grafanaWithDashboards.dashboardSources,
grafanaWithDashboards.dashboardDatasources,
grafanaWithDashboards.deployment,
grafanaWithDashboards.serviceAccount,
grafanaWithDashboards.service {
spec+: { ports: [
port {
nodePort: 30910,
}
for port in super.ports
] },
},
],
}Using the kubernetes-mixins, simply install:
$ jb install github.com/kubernetes-monitoring/kubernetes-mixin
And apply the mixin:
local kubernetesMixin = import 'github.com/kubernetes-monitoring/kubernetes-mixin/mixin.libsonnet';
local grafana = import 'grafana/grafana.libsonnet';
{
local basicWithMixin =
(grafana {
_config+:: {
namespace: 'monitoring-grafana',
grafana+:: {
dashboards: kubernetesMixin.grafanaDashboards,
},
},
}).grafana,
apiVersion: 'v1',
kind: 'List',
items:
basicWithMixin.dashboardDefinitions +
[
basicWithMixin.dashboardSources,
basicWithMixin.dashboardDatasources,
basicWithMixin.deployment,
basicWithMixin.serviceAccount,
basicWithMixin.service {
spec+: { ports: [
port {
nodePort: 30910,
}
for port in super.ports
] },
},
],
}To generate, again simply run:
$ jsonnet -J vendor example-with-mixin.jsonnet
This yields a fully configured Grafana stack with useful Kubernetes dashboards.
Grafana can be run with many different configurations. Different organizations have different preferences, therefore the Grafana configuration can be arbitrary modified. The configuration happens via the the $._config.grafana.config variable. The $._config.grafana.config field is compiled using jsonnet's std.manifestIni function. Additionally you can specify your organizations' LDAP configuration through $._config.grafana.ldap variable.
For example to modify Grafana configuration and set up LDAP use:
local grafana = import 'grafana/grafana.libsonnet';
{
local customIni =
(grafana {
_config+:: {
namespace: 'monitoring-grafana',
grafana+:: {
config: {
sections: {
metrics: { enabled: true },
'auth.ldap': {
enabled: true,
config_file: '/etc/grafana/ldap.toml',
allow_sign_up: true,
},
},
},
ldap: |||
[[servers]]
host = "127.0.0.1"
port = 389
use_ssl = false
start_tls = false
ssl_skip_verify = false
bind_dn = "cn=admin,dc=grafana,dc=org"
bind_password = 'grafana'
search_filter = "(cn=%s)"
search_base_dns = ["dc=grafana,dc=org"]
|||,
},
},
}).grafana,
apiVersion: 'v1',
kind: 'List',
items:
customIni.dashboardDefinitions +
[
customIni.config,
customIni.dashboardSources,
customIni.dashboardDatasources,
customIni.deployment,
customIni.serviceAccount,
customIni.service {
spec+: { ports: [
port {
nodePort: 30910,
}
for port in super.ports
] },
},
],
}The config object allows specifying an array of plugins to install at startup.
local grafana = import 'grafana/grafana.libsonnet';
{
local basic =
(grafana {
_config+:: {
namespace: 'monitoring-grafana',
grafana+:: {
plugins: ['camptocamp-prometheus-alertmanager-datasource'],
},
},
}).grafana,
apiVersion: 'v1',
kind: 'List',
items:
basic.dashboardDefinitions +
[
basic.dashboardSources,
basic.dashboardDatasources,
basic.deployment,
basic.serviceAccount,
basic.service {
spec+: { ports: [
port {
nodePort: 30910,
}
for port in super.ports
] },
},
],
}There are a number of things missing for the Grafana stack and tooling to be fully migrated.
If you are interested in working on any of these, please open a respective issue to avoid duplicating efforts.
- A tool to review Grafana dashboard changes on PRs. While reviewing jsonnet code is a lot easier than the large Grafana json sources, it's hard to imagine what that will actually end up looking like once rendered. Ideally a production-like environment is spun up and produces metrics to be graphed, then a tool could take a screenshot and Grafana snapshot of the rendered Grafana dashboards. That way the changes can not only be reviewed in code but also visually. Similar to point 2 this should eventually be it's own project.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent