Skip to content
This repository has been archived by the owner on Apr 25, 2022. It is now read-only.

Latest commit

 

History

History
247 lines (156 loc) · 9.67 KB

hack_loki_operator.md

File metadata and controls

247 lines (156 loc) · 9.67 KB

Hacking on Loki Operator

Loki Operator is the Kubernetes Operator for Loki.

Hacking on Loki Operator using kind

kind is a tool for running local Kubernetes clusters using Docker container "nodes". kind was primarily designed for testing Kubernetes itself, but may be used for local development or CI.

Requirements

  • Install kubectl or Openshift CLI for communicating with the cluster. The guide below will be using kubectl for the same.
  • Create a running Kubernetes cluster using kind.
  • A container registry that you and your Kubernetes cluster can reach. We recommend quay.io.

Installation of Loki Operator

  • Build and push the container image and then deploy the operator with:

    make oci-build oci-push deploy REGISTRY_ORG=$YOUR_QUAY_ORG VERSION=latest

    where $YOUR_QUAY_ORG is your personal quay.io account where you can push container images.

    The above command will deploy the operator to your active Kubernetes cluster defined by your local kubeconfig. The operator will be running in the default namespace.

  • You can confirm that the operator is up and running using:

    kubectl get pods

    You should see controller-manager-xxxx and minio-xxxx pods running.

  • Now create a LokiStack instance to get the various components of Loki up and running:

    kubectl apply -f hack/lokistack_dev.yaml

    This will create distributor, compactor, ingester, querier and query-frontend components.

    Confirm that all components are up and running for deployments using:

    kubectl rollout status deployment/<DEPLOYMENT_NAME>

    where <DEPLOYMENT_NAME> is the name of the deployment and can be found using:

    kubectl get deployments

    Confirm that all are up and running for statefulsets using:

    kubectl rollout status statefulset/<STATEFULSET_NAME>  

    where <STATEFULSET_NAME> is the name of the statefulset and can be found using:

    kubectl get statefulsets

Cleanup

To cleanup deployments of the operator, you can use:

make undeploy

It will undeploy controller from the configured Kubernetes cluster in ~/.kube/config

Hacking on Loki Operator on OpenShift

Requirements

  • Install kubectl or Openshift CLI for communicating with the cluster. The guide below will be using kubectl for the same.
  • Create a running OpenShift cluster on AWS.
  • A container registry that you and your OpenShift cluster can reach. We recommend quay.io.
  • Create an S3 bucket in one of the AWS Regions.

Installation of Loki Operator

  • Build and push the container image [2] and then deploy the operator with:

    make olm-deploy REGISTRY_ORG=$YOUR_QUAY_ORG VERSION=$VERSION

    where $YOUR_QUAY_ORG is your personal quay.io account where you can push container images and $VERSION can be any random version number such as v0.0.1.

    The above command will deploy the operator to your active Openshift cluster defined by your local kubeconfig. The operator will be running in the openshift-logging namespace.

  • You can confirm that the operator is up and running using:

    kubectl -n openshift-logging get pods
  • Now you need to create a storage secret for the operator. This can be done using:

    make olm-deploy-example-storage-secret

    OR

    ./hack/deploy-example-secret.sh openshift-logging

    This secret will be available in openshift-logging namespace. You can check the hack/deploy-example-secret.sh file to check the content of the secret.

  • Now you need to create a gateway secret [3] for the operator. This can be done using:

    kubectl -n openshift-logging create secret generic test1 \
      --from-literal=clientID="<CLIENT_ID>" \
      --from-literal=clientSecret="<CLIENT_SECRET>" \
      --from-literal=issuerCAPath="<ISSUER_CA_PATH>"
    
  • Once the object storage secret is created, you can now create a LokiStack instance to get the various components of Loki up and running:

    kubectl -n openshift-logging apply -f hack/lokistack_gateway_dev.yaml

    This will create distributor, compactor, ingester, querier, query-frontend and lokistack-gateway components.

    Confirm that all are up and running for deployments using:

    kubectl -n openshift-logging rollout status deployment/<DEPLOYMENT_NAME>

    where <DEPLOYMENT_NAME> is the name of the deployment and can be found using:

    kubectl -n openshift-logging get deployments

    Confirm that all are up and running for statefulsets using:

    kubectl -n openshift-logging rollout status statefulset/<STATEFULSET_NAME>  

    where <STATEFULSET_NAME> is the name of the statefulset and can be found using:

    kubectl -n openshift-logging get statefulsets

    If you don't want lokistack-gateway component [1] then you can skip it by removing the --with-lokistack-gateway args from the loki-operator-controller-manager deployment:

    kubectl -n openshift-logging edit deployment/loki-operator-controller-manager

    Delete the flag --with-lokistack-gateway from the args section and save the file. This will update the deployment and now you can create LokiStack instance using:

    kubectl -n openshift-logging apply -f hack/lokistack_dev.yaml

    This will create distributor, compactor, ingester, querier and query-frontend components only.

Cleanup

To cleanup deployments of the operator, you can use:

make olm-undeploy

It will cleanup deployments of the operator bundle, and the operator via OLM on an OpenShift cluster selected via ~/.kube/config

Notes

[1] lokistack-gateway is an optional component deployed as part of Loki Operator. It provides secure access to Loki's distributor (i.e. for pushing logs) and query-frontend (i.e. for querying logs) via consulting an OAuth/OIDC endpoint for the request subject.

[2] If you get multiple images as options, and you are required to select one of them then select docker.io/library/golang:1.16

[3] The OIDC configuration expects clientID, clientSecret and issuerCAPath which should be provided via a Kubernetes secret that the LokiStack admin provides upfront.

Each tenant Secret is required to match:

  • metadata.name with TenantsSecretsSpec.Name.
  • metadata.namespace with LokiStack.metadata.namespace.

Basic Troubleshooting on Hacking on Loki Operator

New changes are not detected by Loki Operator

Suppose you made some changes to the Loki Operator's code and deployed it, but the changes are not visible when it runs. This happens when the deployment pulls the old image of the operator because of the imagePullPolicy being set to IfNotPresent. Therefore, you need to make some changes to make your deployment pull a new image always:

  • Go to config/manager/manager.yaml file.

  • Set the imagePullPolicy to Always i.e.,

    imagePullPolicy: Always
  • Deploy the operator again.

kubectl using old context

It is possible that when you use two different clusters - one is kind cluster and the other is OpenShift cluster, you might need to switch between clusters to test your changes. There is a possibility that once you switch between clusters, the kubectl might not switch the context automatically and hence you might need to do this manually to correctly communicate with your cluster.

  • List all the available context:

    kubectl config get-contexts

    The * mark against the context shows the one in use currently.

  • Set the context name you want to use now:

    kubectl config use-context $CONTEXTNAME

    where $CONTEXTNAME is the context name you want to use now from the previous step.

The Loki Operator giving Missing Secrets / Invalid Secrets error

You have probably forgotten to create the gateway secrets because of which the operator runs in degraded condition. Follow the steps mentioned in the step-by-step guide to create the gateway secret first. Once done, you can now create the LokiStack instance.

Verify this by checking the conditions field:

kubectl get lokistack lokistack-dev -o yaml

For OpenShift, the above command would be:

kubectl -n openshift-logging get lokistack lokistack-dev -o yaml

The Loki Operator giving Mandatory Configuration / Incompatible Configuration error

This usually happens when the LokiStack CR is wrongly configured for the lokistack-gateway. Please read the enhancement proposal to figure out the correct way to configure it.