Service PPS
Enable a pipeline to be treated as a long-running service.
December 4, 2023
Spec #
This is a top-level attribute of the pipeline spec.
{
"pipeline": {...},
"transform": {...},
"service": {
"internalPort": int,
"externalPort": int
},
...
}
Attributes #
| Attribute | Description |
|---|---|
internalPort | The port that the user code binds to inside the container. |
externalPort | The port on which it is exposed through the NodePorts functionality of Kubernetes services. |
Behavior #
- When enabled,
transform.cmdis not expected to exit and will restart if it does. - The service becomes exposed outside the container using a Kubernetes service.
- You can access the service at
http://<kubernetes-host>:<externalPort>. - The Service starts running at the first commit in the input repo.
When to Use #
You should use the service field in a HPE ML Data Management Pipeline Spec when you want to expose your pipeline as a Kubernetes service, and allow other Kubernetes services or external clients to connect to it.
Example scenarios:
Microservices architecture: If you are building a microservices architecture, you may want to expose individual pipelines as services that can be accessed by other services in the cluster. By using the service field to expose your pipeline as a Kubernetes service, you can easily connect it to other services in the cluster.
Client access: If you want to allow external clients to access the output of your pipeline, you can use the service field to expose your pipeline as a Kubernetes service and provide clients with the service’s IP address and
externalPort.Load balancing: By exposing your pipeline as a Kubernetes service, you can take advantage of Kubernetes’ built-in load balancing capabilities. Kubernetes automatically load balances traffic to the service’s IP address and
externalPortacross all the replicas of the pipeline’s container.