Skip to content

Latest commit

 

History

History

pipeline

README.md

Tekton pipeline

vscode tekton pipelines extension doc

Bootstrapping

Run the following command to bootstrap a Fybrik instance. By default the instance of Fybrik will be deployed as namespaced scoped, meaning the controller will only watch and process FybrikApplications from a specific namespace. Therefore limiting you to create FybrikApplications from one namespace. By default that namespace is the one specified on the bootstrap script (a.k.a system namespace). You can change this behavior by setting environment variables. For details see Scoping the deployment.

The parameter specified is the system namespace.

. source-external.sh
bash -x bootstrap-pipeline.sh fybrik-myname
# follow on screen instructions

After running the "tkn" command that was returned from the bootstrap script, the following namespaces will be created. Illistrated by using the example system namespace of fybrik-myname:

  1. fybrik-myname - This will be the system namespace where Fybrik is deployed. By default the controller will only process FybrikApplications from this namespace.
  2. fybrik-myname-blueprints - this is the designated namespace for blueprints custom resources. This is also where the data access module is deployed to.
  3. fybrik-myname-app - This namespace is only created if running namespace scoped AND the following environment variable is set: use_application_namespace=true.

Restarting individual tasks

Tasks can be restarted in vscode by right-clicking on the task in the tekton pipelines extension.

From command line, it can be done with this script:

# ./restart-task.sh run-integration-tests
# # Where run-integration-task is a grep match for the task you want to restart

knative eventing is used to restart all downstream tasks. Code rebuilds will trigger image rebuilds. Image rebuilds will trigger helm upgrades.

Scoping the deployment

In the following section the system namespace is the namespace specified when running the bootstrap script.

By default the bootstrap script will deploy Fybrik namespaced scoped. Deploying as namespaced scoped allows multiple instances of Fybrik to be installed in unique namespaces in the same cluster. You can change this behavior but you must do so with care.

Environment variables

Name Default Valid Values Description
cluster_scoped false true, false Indicates if deploy as cluster scoped meaning the controller will watch and process FybrikApplications from all namespaces. For caveats see tips
use_application_namespace false true, false Specifies if you want the bootstrap script to create a namespace for FybrikApplications. This will be ignore if cluster_scoped=true. If set to false the system namespace will be used.

Tips

  1. When deploying a Fybrik instance as cluster scoped - Disclaimer: Only use cluster scope if you know what you’re doing.

    1. There can only be one instance of Fybrik deployed in cluster.
    2. The controller will be setup so it watches all namespaces for FybrikApplications.
    3. You cannot run namespaced scoped and cluster scoped in same cluster.

  2. When deploying a Fybrik instance as namespaced scoped, you can only create FybrikApplication custom resources from one namespace. The default behavior set up by the pipeline, will allow creating FybrikApplication resourcss from the system-app namespace. You can override this behavior by setting the use_application_namespace environment variable to false. The bootstrap code will create the system-app namespace and necessary objects to allow creating FybrikApplications. Hint: You will see the application namespace in the -p mesh-for-data-values= in the tkn command that is returned from running the bootstrap script.

Building code from your development environment

Set the following environment variable to point at your code repo, and it will be copied to the volume tekton tasks will mount rather than cloning code from github

github_workspace="/path/to/fybrik"
. source-external.sh