Skip to content

Latest commit

 

History

History
252 lines (197 loc) · 8.99 KB

README.md

File metadata and controls

252 lines (197 loc) · 8.99 KB
Raw

docs

The checkly-operator was designed to run inside a kubernetes cluster and listen for events on specific CRDs and ingress resources. With the help of it you can set up:

Installation

We currently supply an installation yaml file, this is present in the releases.

The file holds the following resources:

  • namespace
  • CRDs
  • RBAC
  • deployment

In order for the operator to work, you need to supply secrets which hold your checklyhq.com API Key and Account ID. See docs on how to create an API key and get account ID. The operator expects the following environment variables:

        env:
        - name: CHECKLY_API_KEY
          valueFrom:
            secretKeyRef:
              key: CHECKLY_API_KEY
              name: checkly
        - name: CHECKLY_ACCOUNT_ID
          valueFrom:
            secretKeyRef:
              key: CHECKLY_ACCOUNT_ID
              name: checkly

The following steps are an easy example on how to get started with the operator, it is not a production ready method, for example we're not using any secrets managers, you should not create secrets and commit them to git like in the below example, we're only deploying one replica, while the operator does support HA deployments.

If you just want to try out the checkly-operator, you need a local kubernetes installation, the easiest might be Rancher Desktop, see the docs for the installation, once done, come back to this doc.

Download install.yaml

First we'll download the provided install.yaml files, please change the version number accordingly, we might have newer releases since we've written these docs.

export CHECKLY_OPERATOR_RELEASE=v1.7.0
wget "https://github.com/checkly/checkly-operator/releases/download/$CHECKLY_OPERATOR_RELEASE/install-$CHECKLY_OPERATOR_RELEASE.yaml" -O install.yaml
unset CHECKLY_OPERATOR_RELEASE

Feel free to edit the install.yaml file to your liking, usually you'd want to change:

  • checkly-operator deployment replica count
  • checkly-operator deployment CPU and Memory resources
  • log levels via the --zap-log-level CLI options, valid options are debug, info, error

You can apply the install.yaml, this will create the namespace, we need this to create the secrets in the next step:

kubectl apply -f install.yaml

Controller Domain

We're using a domain name for finalizers and annotations, the default value is k8s.checklyhq.com, but it can be changed by supplying the --controller-domain=other.domain.tld runtime option.

This option allows you to run multiple independent deployments of the operator and each would handle different resources based on the controller domain configuration.

Create secret

Grab your checklyhq.com API key and Account ID, the official docs can help you get this information. Substitute the values into the below command:

export CHECKLY_API_KEY=<api-key-from-checklyhq.com>
export CHECKLY_ACCOUNT_ID=<org-id-from-checklyhq.com>
kubectl create secret generic -n checkly-operator-system checkly \
  --from-literal=CHECKLY_API_KEY=$CHECKLY_API_KEY \
  --from-literal=CHECKLY_ACCOUNT_ID=$CHECKLY_ACCOUNT_ID
unset CHECKLY_API_KEY
unset CHECKLY_ACCOUNT_ID

If you check your pod, you should be able to see the pods starting:

kubectl get pods -n checkly-operator-system

Configuration

We will next create a check group, alert channels and api checks through the custom CRDs. The operator was written with a specific opinion:

  • all checks should belong to a specific group
  • group configuration should be inherited by each check which is part of it
  • alert channels are added to groups, not to individual checks

Based on the above, the order of creation should be:

  1. alert channel
  2. check group
  3. api checks

Reference to resources are done based on the kubernetes internal naming, as in the metadata.name field.

Please look at the below examples and change the supplied data so it fits your needs the best. Save the example into individual files and apply them when ready:

kubectl apply -f <name-of-the-file>.yaml

Alert channel

See the docs on what alert channels are and alert-channels for the options we support.

The following configuration will send all alerts to an email address:

apiVersion: k8s.checklyhq.com/v1alpha1
kind: AlertChannel
metadata:
  name: checkly-operator-test-alertchannel
spec:
  email:
    address: "[email protected]"

The AlertChannel resource is cluster scoped, it can be used in any check group but it also means that the name of the resource has to be unique in the kubernetes cluster.

Once applied you can check if it worked:

kubectl get alertchannel checkly-operator-test-alertchannel

You can also view the alert channel on the checklyhq.com dashboard.

Check group

See the docs on what check groups are and check-group for the options we support.

The following configuration adds the above alert channel to all the checks that will be part of the group check:

apiVersion: k8s.checklyhq.com/v1alpha1
kind: Group
metadata:
  name: checkly-operator-test-group
  labels:
    environment: "local"
spec:
  locations:
    - eu-west-1
    - eu-west-2
  alertchannel:
    - checkly-operator-test-alertchannel

The Group resource is cluster scoped, it can be used in any api check but it also means that the name of the resource has to be unique in the kubernetes cluster.

Once applied you can check if it worked:

kubectl get group checkly-operator-test-group

You can also view the check group on the checklyhq.com dashboard.

Tags

Any labels added to the Group resource will be added as tags to the group, these groups are inherited by the checks.

Locations

To see a full list of locations supported by checklyhq.com, see the docs. We're using the location codes, private locations should be technically supported, we just haven't tested them.

API Checks

See the docs on what API checks are and api-checks for the options we support.

We're currently only supporting API checks which perform a GET request.

The following configuration monitors the https://foo.bar/baz endpoint, expects a return code of 200, it's added to the above created group and are muted so they do not send an alert:

apiVersion: k8s.checklyhq.com/v1alpha1
kind: ApiCheck
metadata:
  name: checkly-operator-test-1
  namespace: default
  labels:
    service: "foo"
spec:
  endpoint: "http://foo.bar/baz"
  success: "200"
  frequency: 10 # Default 5
  muted: true # Default "false"
  group: "checkly-operator-test-group"

You can create multiple ApiCheck resources which point to the same group:

apiVersion: k8s.checklyhq.com/v1alpha1
kind: ApiCheck
metadata:
  name: checkly-operator-test-2
  namespace: default
  labels:
    service: "bar"
spec:
  endpoint: "https://checklyhq.com"
  success: "200"
  frequency: 10 # Default 5
  muted: true # Default "false"
  group: "checkly-operator-test-group"

ApiCheck resources are namespace scoped, they have to have a unique name in each namespace.

Once applied you can check if it worked:

kubectl get apichecks -n default

You can also view the checks on the checklyhq.com dashboard.

Ingresses

See ingress for more details on how we utilize ingress resources.

We can create an ingress object and add annotations to it:

---
apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
  name: ingress-sample
  namespace: default
  annotations:
    k8s.checklyhq.com/enabled: "true"
    k8s.checklyhq.com/path: "/baz"
    # k8s.checklyhq.com/endpoint: "foo.baaz" - Default read from spec.rules[0].host
    # k8s.checklyhq.com/success: "200" - Default "200"
    k8s.checklyhq.com/group: "checkly-operator-test-group"
    # k8s.checklyhq.com/muted: "false" # If not set, default "true"
spec:
  rules:
    - host: "foo.bar"
      http:
        paths:
          - path: /
            pathType: ImplementationSpecific
            backend:
              service:
                name: test-service
                port:
                  number: 8080

Check if it worked:

kubectl get ingress -n default
kubectl get apicheck -n default

You can also view the checks on the checklyhq.com dashboard.