This document describes how to get up and running with a new Grafana instance on Kubernetes.
The first step is to install the Grafana operator to a namespace in your cluster.
There are two options for this procedure, automated via Ansible, or manually running kubectl/oc commands.
Run make operator/deploy
To deploy the latest released image of the operator, This uses the default operator.yaml and
Grafana.yaml resources found in the deploy
directories and subdirectories.
Warning: The following make recipe uses the latest master Image. It should only be used for testing, for production please use the tagged releases.
Run make operator/deploy/master
to deploy the image from the master branch of the operator.
Cluster admin install cluster resources. For more details and additional parameters see grafana-operator-cluster-resources.yaml.
ansible-playbook deploy/ansible/grafana-operator-cluster-resources.yaml \
-e k8s_host=https://ocp.example.xyz \
-e k8s_username=admin1 \
-e k8s_password=secret \
-e grafana_operator_namespace=grafana
Optional: If grafana_operator_args_scan_all
is set to true
for the grafana-operator-namespace-resources.yaml
playbook then Cluster Admin needs to run this playbook to allow operator to scan all namespaces for dashboards For more
details and additional parameters
see grafana-operator-cluster-dashboards-scan.yaml
.
ansible-playbook deploy/ansible/grafana-operator-cluster-dashboards-scan.yaml \
-e k8s_host=https://ocp.example.xyz \
-e k8s_username=admin1 \
-e k8s_password=secret \
-e grafana_operator_namespace=grafana
Self provisioner install operator For more details and additional parameters see grafana-operator-namespace-resources.yaml.
ansible-playbook deploy/ansible/grafana-operator-namespace-resources.yaml \
-e k8s_host=https://ocp.example.xyz \
-e k8s_username=project_creator \
-e k8s_password=secret \
-e grafana_operator_namespace=grafana
Follow this documentation Deploying the Grafana operator in minikube
To create a namespace named grafana
run:
$ kubectl create namespace grafana
Create the custom resource definitions that the operator uses:
$ kubectl create -f deploy/crds
Create the operator roles:
$ kubectl create -f deploy/roles -n grafana
If you want to scan for dashboards in other namespaces you also need the cluster roles:
$ kubectl create -f deploy/cluster_roles
To deploy the operator to that namespace you can use deploy/operator.yaml
:
$ kubectl create -f deploy/operator.yaml -n grafana
Check the status of the operator pod:
$ kubectl get pods -n grafana
NAME READY STATUS RESTARTS AGE
grafana-operator-78cfcbf8db-ssrgq 1/1 Running 0 17s
Please take note of this chart when wanting to deploy custom grafana images through either --grafana-image
operator
flags or baseImage
CR spec fields.
This chart shows how the operator prioritises which image will be used for the deployment, and the versions that it's known to support. Only the grafana image specified in code here will be supported in unit/e2e tests and as part of the operator, any other specified grafana image through these options may not work as expected.
The operator accepts a number of flags that can be passed in the args
section of the container in the deployment:
-
--grafana-image
: overrides the Grafana image, defaults todocker.io/grafana/grafana
on grafanas official dockerhub, the version of which can be found here. -
--grafana-image-tag
: overrides the Grafana tag. Seecontroller_config.go
for default. -
--grafana-plugins-init-container-image
: overrides the Grafana Plugins Init Container image, defaults toquay.io/integreatly/grafana_plugins_init
. -
--grafana-plugins-init-container-tag
: overrides the Grafana Plugins Init Container tag, defaults to0.0.3
. -
--grafonnet-location
: overrides the location of the grafonnet library. Defaults to/opt/grafonnet-lib
. Only useful when running the operator locally. -
--scan-all
: watch for dashboards in all namespaces. This requires the operator service account to have cluster wide permissions toget
,list
,update
andwatch
dashboards. Seedeploy/cluster_roles
. -
--namespaces
: watch for dashboards in a list of namespaces. Mutually exclusive with--scan-all
. -
--zap-level=n
: set the logging level for the operator, leaving out this flag will only log Errors and error related info, current options are:--zap-level=1
: show all Info level logs
See deploy/operator.yaml
for an example.
If using the automated Ansible installer see the grafana-operator-namespace-resources.yaml - Parameters for the equivalent parameters.
Create a custom resource of type Grafana
, or use the one in deploy/examples/Grafana.yaml
.
The resource accepts the following properties in it's spec
:
-
baseImage: Specifies a custom grafana image for this deployment.
- Warning! this overwrites the
--grafana-image
Operator flag, please refer to the grafana image support chart.
- Warning! this overwrites the
-
initImage: Specifies a custom grafana plugins init image for this deployment.
- Warning! this overwrites the
--grafana-plugins-init-container-image
Operator flag, please refer to the grafana image support chart.
- Warning! this overwrites the
-
dashboardLabelSelector: A list of either
matchLabels
ormatchExpressions
to filter the dashboards before importing them. -
containers: Extra containers to be added to the Grafana deployment. Can be used for example to add auth proxy side cars.
-
secrets: A list of secrets that are added as volumes to the deployment. Useful in combination with extra
containers
or when extra configuraton files are required. -
configMaps: A list of config maps that are added as volumes to the deployment. Useful in combination with extra
containers
or when extra configuraton files are required. -
config: The properties used to generate
grafana.ini
. All properties defined in the official documentation are supported although some of them are not allowed to be overridden (path configuration). Seedeploy/examples/Grafana.yaml
for an example. -
ingress: Allows configuring the Ingress / Route resource (see here).
-
service: Allows configuring the Service resource (see here).
-
serviceAccount: Allows configuring the Service account (see here).
-
deployment: Allows configuring the deployment (see here).
-
resources: Allows configuring the requests and limits for the Grafana pod ( see here).
-
client: Grafana client options (see here).
-
jsonnet: Label selector for jsonnet libraries (see here).
-
livenessProbeSpec: Defines the time, in seconds, to be used for each field in the readiness probe configuration ( see here)
-
readinessProbeSpec: Defines the time, in seconds, to be used for each field in the readiness probe configuration ( see here)
NOTE: by default no Ingress or Route is created. It can be enabled with spec.ingress.enabled
.
To create a new Grafana instance in the grafana
namespace, run:
$ kubectl create -f deploy/examples/Grafana.yaml -n grafana
Get the URL of the instance and open it in a browser:
$ kubectl get ingress -n grafana
NAME HOSTS ADDRESS PORTS AGE
grafana-ingress grafana.apps.127.0.0.1.nip.io 80 28s
When the config object in the Grafana
CR is modified, the grafana.ini
will be automatically updated and Grafana will
be restarted.
By default the operator will not create an Ingress or Route. This can be enabled via spec.ingress
in the Grafana
CR.
The operator will create a Route when running on OpenShift, otherwise an Ingress. Various other properties can also be
configured:
spec:
ingress:
enabled: <Boolean> # Create an Ingress (or Route if on OpenShift)
ingressClassName: <String> # Sets ingress ingressClassName
hostname: <String> # Sets the hostname. Assigned automatically on OpenShift if not provided
tlsEnabled: <Boolean> # Enable TLS on Ingress
tlsSecretName: <String> # TLS secret name in the same namespace
targetPort: <String> # Which port to target on the service
termination: <String> # TLS termination type (defaults to edge, other options are reencrypt or passthrough)
labels: # Additional labels for the Ingress or Route
app: grafana
...
annotations: # Additional annotations for the Ingress or Route
app: grafana
...
path: # Sets the path of the Ingress. Ignored for Routes
pathType: <String> # Sets pathType: ImplementationSpecific, Exact, Prefix (defaults to ImplementationSpecific)
spec:
serviceAccount:
type: object
properties:
skip:
type: boolean
description: setting this to `True` will stop the operator from reconciling the `grafana-serviceaccount`
serviceaccount, Leaving this field empty is equivalent to setting it to`False`
annotations:
type: object
description: Additional annotations for the serviceaccount
labels:
type: object
description: Additional labels for the serviceaccount
Various properties of the Service can be configured:
spec:
service:
name: # Configurable name for the grafana-service
labels: # Additional labels for the Service
app: grafana
...
annotations: # Additional annotations for the Service
app: grafana
...
type: NodePort # Set Service type, either NodePort, ClusterIP or LoadBalancer
ports: # Additional ports to add to the service
- name: grafana-proxy
port: 9091
protocol: TCP
targetPort: ...
NOTE: Service name must adhere to a DNS-1035 label which must consist of lower case alphanumeric characters or '-',
start with an alphabetic character, and end with an alphanumeric character (e.g. 'my-name', or 'abc-123', regex used for
validation is '[a-z]([-a-z0-9]*[a-z0-9])?
. If the name doesn't match this RegEx, the operator will fail to create the
new service, however, the old service will still remain available until a new one is created.
Various properties of the Deployment can be configured:
spec:
deployment:
labels: # Additional labels for the Deployment
app: grafana
...
annotations: # Additional annotations for the Deployment
app: grafana
...
replicas: <Number> # Number of replicas. If more than one is selected, a shared database should be configured.
nodeSelector:
app: grafana # Assign grafana pods to a run on specific nodes.
...
tolerations: # Run grafana pods on tainted nodes.
...
affinity: # Affinity parameters shall support node, pod affinity and anti-affinity usecases.
...
securityContext: # Run grafana pods with pod security context
...
containerSecurityContext: # Run grafana pods with container security context
...
envFrom: # Environment variables from Secret or ConfigMap. The key of data becomes the env name.
...
hostNetwork: <bool> # hostNetwork option, DNSPolicy will be accordingly changed (ClusterFirst by default and ClusterFirstWithHostNet for hostNetwork: true)
...
skipCreateAdminAccount: <bool> # Skip creating the admin account when providing custom credentials from a secret.
...
priorityClassName: <string> # Assign a priorityClass name to the grafana pod. Empty by default.
...
extraVolumes: <array> # Append extra volumes to the Grafana deployment
...
extraVolumeMounts: <array> # Append extra volume mounts
...
strategy: # Optional. The DeploymentStrategy to set for the Grafana deployment. Defaults to 25%/25% RollingUpdate if unset.
...
httpProxy:
Enabled: bool # Whether a HTTP(S) proxy should be used for outbound requests from Grafana and when resolving plugins.
URL: string # The URL of the proxy
NOTE: Some key's are common to both in securityContext and containerSecurityContext, in that case containerSecurityContext has precendence over securityContext. ContainerSecurityContext defined in deployment will also apply to the init-container.
Grafana dashboards are imported using the Grafana API. The following options are available to configure the API access:
spec:
client:
timeout: <Number> # Timeout in seconds for API requests (defaults to 5 seconds).
preferService: <Boolean> # If an Ingress or Route is available, the operator will attempt to use those for API access. This flag forces it to use the Service instead.
When not using an external database, Grafana creates a SQLite database. By default, the location of this database is ephemeral but can be configured:
spec:
dataStorage:
labels: # Additional labels for the PVC
app: grafana
...
annotations: # Additional annotations for the PVC
app: grafana
...
accessModes: # An array of access modes, e.g. `ReadWriteOnce`
...
size: <Quantity> # Requested size, e.g. `10Gi`
class: <String> # Storage class name
Jsonnet specific configuration options.
spec:
jsonnet:
libraryLabelSelector: <LabelSelector> # Selector to discover config maps containing jsonnet libraries
spec:
livenessProbeSpec:
initialDelaySeconds: <Int32>
timeoutSeconds: <Int32>
periodSeconds: <Int32>
sucessThreshold: <Int32>
failureThreshold: <Int32>
readinessProbeSpec:
initialDelaySeconds: <Int32>
timeoutSeconds: <Int32>
periodSeconds: <Int32>
successThreshold: <Int32>
failureThreshold: <Int32>
Both LivenessProbeSpec and ReadinessProbeSpec share the same fields which serve the same purpose.
- initialDelaySeconds: Number of seconds after the container has started before liveness probes are initiated. More info
- timeoutSeconds: Number of seconds after which the probe times out. Defaults to 1 second. Minimum value is 1.
- periodSeconds: How often (in seconds) to perform the probe. Default to 10 seconds. Minimum value is 1.
- successThreshold: Minimum consecutive successes for the probe to be considered successful after having failed. Defaults to 1. Must be 1 for liveness and startup. Minimum value is 1.
- failureThreshold:: When a probe fails, Kubernetes will try failureThreshold times before giving up. Giving up in case of liveness probe means restarting the container. In case of readiness probe the Pod will be marked Unready. Defaults to 3. Minimum value is 1.