Peanut Kubernetes Pipelines
Peanut is an operator for running pipelines in Kubernetes.
What is a pipeline?
In its current form, a PeanutPipeline is a set of stages that take the form of a Directed Acyclic Graph (https://en.wikipedia.org/wiki/Directed_acyclic_graph)
This means that stages in a pipeline can depend on other stages and will only be executed if parent stages were successful.
What is a stage?
A stage is a template for a Kubernetes Job that the operator will end up scheduling on the cluster.
Deploying the operator
TODO
Examples
Creating a PeanutStage:
apiVersion: pipeline.peanut.dev/v1alpha1
kind: PeanutStage
metadata:
name: stage-sample
spec:
image: busybox
command: ["sleep"]
args: ["10"]
---
apiVersion: pipeline.peanut.dev/v1alpha1
kind: PeanutStage
metadata:
name: stage-sample-two
spec:
image: busybox
command: ["sleep"]
args: ["15"]
Creating a PeanutPipeline:
apiVersion: pipeline.peanut.dev/v1alpha1
kind: PeanutPipeline
metadata:
name: pipeline-sample
spec:
stages:
- name: "First Stage"
type: stage-sample
- name: "Second Stage"
type: stage-sample-two
- name: "Third Stage"
type: stage-sample
dependsOn:
- "Second Stage"
- name: "Fourth Stage"
type: stage-sample
dependsOn:
- "Second Stage"
- name: "Fifth Stage"
type: stage-sample-two
dependsOn:
- "Third Stage"
- "Fourth Stage"
This creates the following pipeline
+---------+
| |
| First |
+-->+ Stage |
| | |
| +---------+
+------+
| +--------+ +---------+
| | | | |
+-->+ Second | | Third +----+
| Stage +--+-->+ Stage | |
| | | | | | +---------+
+--------+ | +---------+ | | |
| +-->+ Fifth |
| +--------+ | | Stage |
| | | | | |
| | Fourth | | +---------+
+-->+ Stage +-----+
| |
+--------+
Executing our pipeline:
apiVersion: pipeline.peanut.dev/v1alpha1
kind: PeanutPipelineExecution
metadata:
name: peanutpipelineexecution-sample
spec:
pipelineName: pipeline-sample
The operator will schedule the jobs according to the definition created.
If a stage fails all stages in the same branch won't be scheduled.
At the moment there isn't much flexibility on the behaviour of the pipeline when a stage fails.
But Why?
I started this project to learn how to develop an operator for k8s using kubebuilder.
Can I use this in production?
I would strongly advise against it, use at your own risk. This is very much a work in progress and not well tested.
Contributing
Feel free to fork the project and submit a PR