Skip to content

Latest commit

 

History

History
174 lines (124 loc) · 6.07 KB

HACKING.md

File metadata and controls

174 lines (124 loc) · 6.07 KB
Raw

Getting started

Prerequisites

In the case this operator is being used without the Cluster Logging Operator or Jaeger Operator:

  • the namespace where the Elasticsearch cluster will be deployed into needs to be created. By default our make targets use openshift-logging, however they can be overridden by setting DEPLOYMENT_NAMESPACE.

  • certificates must be provided in the form of a secret where the name matches the pattern <elasticsearch_cr_name>-secret. For the convenience of local development the make target deploy-example-secret can be used to do this.

OpenShift

To build a local Elasticsearch Operator image and deploy the example CR you can use:

make deploy-example

Alternatively, if you just want to deploy the Elasticsearch Operator and provide your own CR afterwards you can use:

make deploy

There is an example CR provided in the hack directory that can be created with:

oc create -f hack/cr.yaml

Customize your cluster

Image customization

The operator is designed to work with quay.io/openshift-logging/elasticsearch6 image. To use a different image edit config/manager/manage.yaml and re-run make bundle.

Storage configuration

Storage is configurable per individual node type. Possible configuration options:

  • Empty directory
  • PersistentVolume generated by StorageClass (if storage class is left off the cluster default is used)

Elasticsearch cluster topology customization

Decide how many nodes you want to run.

Exposing elasticsearch service with a route

Obtain the CA cert from Elasticsearch.

oc extract secret/elasticsearch --to=. --keys=admin-ca

Create a re-encrypt route.

In the Re-encrypt termination example, use the contents of the file admin-ca for spec.tls.destinationCACertificate. You do not need to set the spec.tls.key, spec.tls.certificate and spec.tls.caCertificate parameters shown in the example.

Supported features

Kubernetes TBD+ and OpenShift TBD+ are supported.

  • SSL-secured deployment (using Opendistro)
  • Insecure deployment (requires different image)
  • Index per tenant
  • Elasticsearch 6.x support
  • Master role
  • Client role
  • Data role
  • Clientdata role
  • Clientdatamaster role
  • Prometheus monitoring
  • Status monitoring
  • Rolling restarts

Testing

In a real deployment OpenShift monitoring will be installed. However for testing purposes, you should install the monitoring CRDs:

make deploy

E2E Testing

To run the e2e tests run:

make test-e2e-olm

This assumes:

  • the Elasticsearch Operator pod is already deployed and running
  • the Elasticsearch Operator Registry pod is already deployed and running

Dev Testing

You should first ensure that you have podman available.

To test on an OKD cluster, you can run:

make deploy

To remove created API objects:

make uninstall

Building a Universal Base Image (UBI) based image

You must first oc login api.ci.openshift.org. You'll need these credentials in order to pull images from the UBI registry.

The image build process for UBI based images uses a private yum repo. In order to use the private yum repo, you will need access to https://github.com/openshift/release/blob/master/ci-operator/infra/openshift/release-controller/repos/ocp-4.1-default.repo and https://github.com/openshift/shared-secrets/blob/master/mirror/ops-mirror.pem Note that the latter is private and requires special permission to access.

The best approach is to clone these repos under $GOPATH/src/github.com/openshift which the build scripts will pick up automatically. If you do not, the build script will attempt to clone them to a temporary directory.

Running a full integration test with a 4.x cluster

If you have a local clone of origin-aggregated-logging under $GOPATH/github.com/openshift/origin-aggregated-logging you can use its hack/get-cluster-run-tests.sh to build the logging, elasticsearch-operator, and cluster-logging-operator images; deploy a 4.x cluster; push the images to the cluster; deploy logging; and launch the logging CI tests.

Pushing an image to a 4.x cluster

You'll need podman in addition to docker.

If you used the new openshift-installer, it creates a user named kubeadmin with the password in the file installer/auth/kubeadmin_password.

If you already built the image with make image, then it will not be built again until a change is detected in a prerequisite file (defined in the Makefile) You need oc login to the cluster before pushing the image to the register.

oc login --token=<token> --server=https://api.<cluster>:6443
# Or oc login -u kubeadmin -p <password>, if using openshift installer
make deploy-image

Upgrading operator-sdk

The elasticsearch operator relies on operator-sdk to scaffold and generate code and manifests. To upgrade the generated code to a new operator-sdk version, we need to follow the instructions in operator-sdk's migration guide.

In addition, we must ensure that the k8s dependencies in the operator's go.mod match the selected version of operator-sdk. For example, for operator-sdk v0.19.x, check the k8s dependencies:

As a result, we can determine the versions of the k8s dependencies in the operator's go.mod.

Generate Go code, CRD, RBAC, and bundle

# Generate Go code, CRD, ...
make generate

# Generate bundle directory
make bundle