postgres-operator

Developer Guide

Read this guide if you want to debug the operator, fix bugs or contribute new features and tests.

Setting up Go

Postgres Operator is written in Go. Use the installation instructions if you don’t have Go on your system. You won’t be able to compile the operator with Go older than 1.17. We recommend installing the latest one.

Go projects expect their source code and all the dependencies to be located under the GOPATH. Normally, one would create a directory for the GOPATH (i.e. ~/go) and place the source code under the ~/go/src sub directories.

Given the schema above, the Postgres Operator source code located at github.com/zalando/postgres-operator should be put at -~/go/src/github.com/zalando/postgres-operator.

export GOPATH=~/go
mkdir -p ${GOPATH}/src/github.com/zalando/
cd ${GOPATH}/src/github.com/zalando/
git clone https://github.com/zalando/postgres-operator.git

Building the operator

We use Go Modules for handling dependencies. When using Go below v1.13 you need to explicitly enable Go modules by setting the GO111MODULE environment variable to on. The make targets do this for you, so simply run

make deps

This would take a while to complete. You have to redo make deps every time your dependencies list changes, i.e. after adding a new library dependency.

Build the operator with the make docker command. You may define the TAG variable to assign an explicit tag to your Docker image and the IMAGE to set the image name. By default, the tag is computed with git describe --tags --always --dirty and the image is registry.opensource.zalan.do/acid/postgres-operator

export TAG=$(git describe --tags --always --dirty)
make docker

Building the operator binary (for testing the out-of-cluster option):

make

The binary will be placed into the build directory.

Deploying self build image

The fastest way to run and test your Docker image locally is to reuse the Docker environment from minikube or use the load docker-image from kind. The following steps will get you the Docker image built and deployed.

# minikube
eval $(minikube docker-env)
make docker

# kind
make docker
kind load docker-image registry.opensource.zalan.do/acid/postgres-operator:${TAG} --name <kind-cluster-name>

Then create a new Postgres Operator deployment.

Deploying manually with manifests and kubectl

You can reuse the provided manifest but replace the version and tag. Don’t forget to also apply configuration and RBAC manifests first, e.g.:

kubectl create -f manifests/configmap.yaml
kubectl create -f manifests/operator-service-account-rbac.yaml
sed -e "s/\(image\:.*\:\).*$/\1$TAG/" -e "s/\(imagePullPolicy\:\).*$/\1 Never/" manifests/postgres-operator.yaml | kubectl create  -f -

# check if the operator is coming up
kubectl get pod -l name=postgres-operator

Deploying with Helm chart

Yoy can reuse the provided Helm chart to deploy local operator build with the following command:

helm install postgres-operator ./charts/postgres-operator --namespace zalando-operator --set image.tag=${TAG} --set image.pullPolicy=Never

Code generation

The operator employs K8s-provided code generation to obtain deep copy methods and K8s-like APIs for its custom resource definitions, namely the Postgres CRD and the operator CRD. The usage of the code generation follows conventions from the K8s community. Relevant scripts live in the hack directory:

update-codegen.sh triggers code generation for the APIs defined in pkg/apis/acid.zalan.do/,
verify-codegen.sh checks if the generated code is up-to-date (to be used within CI).

The /pkg/generated/ contains the resultant code. To make these scripts work, you may need to export GOPATH=$(go env GOPATH)

References for code generation are:

Relevant pull request See comments there for minor issues that can sometimes broke the generation process.
Code generator source code
Code Generation for CustomResources - intro post on the topic
Code generation in Prometheus and etcd operators

To debug the generated API locally, use the kubectl proxy and kubectl --v=8 log level to display contents of HTTP requests (run the operator itself with --v=8 to log all REST API requests). To attach a debugger to the operator, use the -outofcluster option to run the operator locally on the developer’s laptop (and not in a docker container).

Debugging the operator

There is a web interface in the operator to observe its internal state. The operator listens on port 8080. It is possible to expose it to the localhost:8080 by doing:

kubectl --context minikube port-forward $(kubectl --context minikube get pod -l name=postgres-operator -o jsonpath={.items..metadata.name}) 8080:8080

The inner query gets the name of the Postgres Operator pod, and the outer one enables port forwarding. Afterwards, you can access the operator API with:

curl --location http://127.0.0.1:8080/$endpoint | jq .

The available endpoints are listed below. Note that the worker ID is an integer from 0 up to ‘workers’ - 1 (value configured in the operator configuration and defaults to 4)

/databases - all databases per cluster
/workers/all/queue - state of the workers queue (cluster events to process)
/workers/$id/queue - state of the queue for the worker $id
/workers/$id/logs - log of the operations performed by a given worker
/clusters/ - list of teams and clusters known to the operator
/clusters/$team - list of clusters for the given team
/clusters/$team/$namespace/$clustername - detailed status of the cluster, including the specifications for CRD, master and replica services, endpoints and statefulsets, as well as any errors and the worker that cluster is assigned to.
/clusters/$team/$namespace/$clustername/logs/ - logs of all operations performed to the cluster so far.
/clusters/$team/$namespace/$clustername/history/ - history of cluster changes triggered by the changes of the manifest (shows the somewhat obscure diff and what exactly has triggered the change)

The operator also supports pprof endpoints listed at the pprof package, such as:

/debug/pprof/
/debug/pprof/cmdline
/debug/pprof/profile
/debug/pprof/symbol
/debug/pprof/trace

It’s possible to attach a debugger to troubleshoot postgres-operator inside a Docker container. It’s possible with gdb and delve. Since the latter one is a specialized debugger for Go, we will use it as an example. To use it you need:

Install delve locally

go get -u github.com/derekparker/delve/cmd/dlv

Add following dependencies to the Dockerfile

RUN apk --no-cache add go git musl-dev
RUN go get -d github.com/derekparker/delve/cmd/dlv

Update the Makefile to build the project with debugging symbols. For that you need to add gcflags to a build target for corresponding OS (e.g. GNU/Linux)

-gcflags "-N -l"

Run postgres-operator under the delve. For that you need to replace ENTRYPOINT with the following CMD:

CMD ["/root/go/bin/dlv", "--listen=:DLV_PORT", "--headless=true", "--api-version=2", "exec", "/postgres-operator"]

Forward the listening port

kubectl port-forward POD_NAME DLV_PORT:DLV_PORT

Attach to it

dlv connect 127.0.0.1:DLV_PORT

Unit tests

Prerequisites:

make deps
make mocks

To run all unit tests, you can simply do:

go test ./pkg/...

In case if you need to debug your unit test, it’s possible to use delve:

dlv test ./pkg/util/retryutil/
Type 'help' for list of commands.
(dlv) c
PASS

To test the multi-namespace setup, you can use

./run_operator_locally.sh --rebuild-operator

It will automatically create an acid-minimal-cluster in the namespace test. Then you can for example check the Patroni logs:

kubectl logs acid-minimal-cluster-0

Unit tests with Mocks and K8s Fake API

Whenever possible you should rely on leveraging proper mocks and K8s fake client that allows full fledged testing of K8s objects in your unit tests.

To enable mocks, a code annotation is needed: Mock code gen annotation

To generate mocks run:

make mocks

Examples for mocks can be found in: Example mock usage

Examples for fake K8s objects can be found in: Example fake K8s client usage

End-to-end tests

The operator provides reference end-to-end (e2e) tests to ensure various infrastructure parts work smoothly together. The test code is available at e2e/tests. The special registry.opensource.zalan.do/acid/postgres-operator-e2e-tests-runner image is used to run the tests. The container mounts the local e2e/tests directory at runtime, so whatever you modify in your local copy of the tests will be executed by a test runner. By maintaining a separate test runner image we avoid the need to re-build the e2e test image on every build.

Each e2e execution tests a Postgres Operator image built from the current git branch. The test runner creates a new local K8s cluster using kind, utilizes provided manifest examples, and runs e2e tests contained in the tests folder. The K8s API client in the container connects to the kind cluster via the standard Docker bridge network. The kind cluster is deleted if tests finish successfully or on each new run in case it still exists.

End-to-end tests are executed automatically during builds (for more details, see the README in the e2e folder):

make e2e

End-to-end tests are written in Python and use flake8 for code quality. Please run flake8 before submitting a PR.

Introduce additional configuration parameters

In the case you want to add functionality to the operator that shall be controlled via the operator configuration there are a few places that need to be updated. As explained here, it’s possible to configure the operator either with a ConfigMap or CRD, but currently we aim to synchronize parameters everywhere.

When choosing a parameter name for a new option in a Postgres cluster manifest, keep in mind the naming conventions there. We use camelCase for manifest parameters (with exceptions for certain Patroni/Postgres options) and snake_case variables in the configuration. Only introduce new manifest variables if you feel a per-cluster configuration is necessary.

Note: If one option is defined in the operator configuration and in the cluster manifest, the latter takes precedence.

Go code

Update the following Go files that obtain the configuration parameter from the manifest files:

Postgres manifest parameters are defined in the api package. The operator behavior has to be implemented at least in k8sres.go. Validation of CRD parameters is controlled in crds.go. Please, reflect your changes in tests, for example in:

Updating manifest files

For the CRD-based configuration, please update the following files:

the default OperatorConfiguration
the CRD’s validation
the CRD’s validation in the Helm chart

Add new options also to the Helm chart’s values file file. It follows the OperatorConfiguration CRD layout. Nested values will be flattened for the ConfigMap. Last but no least, update the ConfigMap manifest example as well.

Updating documentation

Finally, add a section for each new configuration option and/or cluster manifest parameter in the reference documents:

It also helps users to explain new features with examples in the administrator docs.

This site is open source. Improve this page.