The PostgreSQL Operator is an open source project hosted on GitHub.
This guide is intended for those wanting to build the Operator from source or contribute via pull requests.
The target development host for these instructions is a CentOS 7 or RHEL 7 host. Others operating systems are possible, however we do not support building or running the Operator on others at this time.
The following environment variables are expected by the steps in this guide:
||$HOME/odev||Golang project directory|
||$GOPATH/src/github.com/crunchydata/postgres-operator||Operator repository location|
||$PGOROOT/installers/ansible/roles/pgo-operator/files||Operator Config Template Directory|
||centos7||Base OS for container images|
||kubectl||Cluster management tool executable|
||crunchydata||Container image prefix|
||pgo||Kubernetes namespace for the operator|
examples/envs.shcontains the above variable definitions as well as others used by postgres-operator tools
- The development host has been created, has access to
yumupdates, and has a regular user account with
sudorights to run
GOPATHpoints to a directory containing
- The development host has
$GOPATH/binadded to its
PATHenvironment variable. Development tools will be installed to this path. Defining a
GOBINenvironment variable other than
$GOPATH/binmay yield unexpected results.
- The development host has
gitinstalled and has cloned the postgres-operator repository to
$GOPATH/src/github.com/crunchydata/postgres-operator. Makefile targets below are run from the repository directory.
- Deploying the Operator will require deployment access to a Kubernetes or OpenShift cluster
- Once you have cloned the git repository, you will need to download the CentOS 7 repository files and GPG keys and place them in the
$PGOROOT/confdirectory. You can do so with the following code:
cd $PGOROOT curl https://api.developers.crunchydata.com/downloads/repo/rpm-centos/postgresql12/crunchypg12.repo > conf/crunchypg12.repo curl https://api.developers.crunchydata.com/downloads/repo/rpm-centos/postgresql11/crunchypg11.repo > conf/crunchypg11.repo curl https://api.developers.crunchydata.com/downloads/gpg/RPM-GPG-KEY-crunchydata-dev > conf/RPM-GPG-KEY-crunchydata-dev
Configuring build dependencies is automated via the
setup target in the project Makefile:
The setup target ensures the presence of:
PATHas described in the prerequisites
- EPEL yum repository
gocompiler version 1.13+
- NSQ messaging binaries
buildahOCI image building tool version 1.14.9+
By default, docker is not configured to run its daemon. Refer to the docker post-installation instructions to configure it to run once or at system startup. This is not done automatically.
Code generation is leveraged to generate the clients and informers utilized to interact with the
various Custom Resources
pgclusters) comprising the PostgreSQL Operator declarative API. Code generation is provided
by the Kubernetes code-generator project,
and the following two
Make targets are included within the PostgreSQL Operator project to both
determine if any generated code within the project requires an update, and then update that code
# Check to see if an update to generated code is needed: make verify-codegen # Update any generated code: make update-codegen
Therefore, in the event that a Custom Resource defined within the PostgreSQL Operator API
$PGOROOT/pkg/apis/crunchydata.com) is updated, the
verify-codegen target will indicate that
an update is needed, and the
update-codegen target should then be utilized to generate the
updated code prior to compiling.
.repofile in the
confdirectory before proceeding.
You will build all the Operator binaries and Docker images by running:
This assumes you have Docker installed and running on your development host.
By default, the Makefile will use buildah to build the container images, to override this default to use docker to build the images, set the IMGBUILDER variable to
The project uses the golang dep package manager to vendor all the golang source dependencies into the
vendor directory. You typically do not need to run any
dep commands unless you are adding new golang package dependencies into the project outside of what is within the project for a given release.
After a full compile, you will have a
pgo binary in
$HOME/odev/bin and the Operator images in your local Docker registry.
Now that you have built the PostgreSQL Operator images, you can now deploy them to your Kubernetes cluster. To deploy the image and associated Kubernetes manifests, you can execute the following command:
If your Kubernetes cluster is not local to your development host, you will need to specify a config file that will connect you to your Kubernetes cluster. See the Kubernetes documentation for details.
Once the PostgreSQL Operator is deployed, you can run the end-to-end regression
test suite interface with the PostgreSQL client. You need to ensure
pgo client executable is in your
$PATH. The test suite can be run
using the following commands:
cd $PGOROOT/testing/pgo_cli GO111MODULE=on go test -count=1 -parallel=2 -timeout=30m -v .
For more information, please follow the testing README in the source repository.
Debug level logging in turned on by default when deploying the Operator.
Sample bash functions are supplied in
examples/envs.sh to view
the Operator logs.
You can view the Operator REST API logs with the
alog bash function.
You can view the Operator core logic logs with the
olog bash function.
You can view the Scheduler logs with the
slog bash function.
These logs contain the following details:
Timestamp Logging Level Message Content Function Information File Information PGO version
Additionally, you can view the Operator deployment Event logs with the
elog bash function.
You can enable the
pgo CLI debugging with the following flag:
pgo version --debug
You can set the REST API URL as follows after a deployment if you are
developing on your local host by executing the
setip bash function.