Triggering on Pub/Sub Messages

Trigger pipelines from Pub/Sub messages.

In order to programmatically trigger pipelines, you can configure Spinnaker to subscribe and listen to a Pub/Sub topic and push messages to the configured topic. This can be used to trigger pipelines during CI jobs, from the command line, or from a third-party system. The message payload will be available in the Pipeline’s execution.

Note: It’s possible to configure multiple pipelines to trigger off of a single Pub/Sub message.

Only Google and AWS Pub/Sub systems are supported. See the detailed instructions for Google Pub/Sub and AWS SQS/SNS .

Prerequisites

Adding a Pub/Sub trigger to a pipeline

Assuming you have created a pipeline, under Configuration, select Add Trigger and make its type selector Pub/Sub.

To select a Pub/Sub subscription to trigger from, select values for Pub/Sub System Type and Subscription Name. Note that the subscription must be configured before it is available to select in the UI.

Payload constraints

If you want to ensure that a Pub/Sub trigger only fires when a certain message payload arrives, you can provide Payload Constraints in the trigger. These are key/value pairs where the key must be found in the incoming message payload, and the value must match using regex.

For example, if we had configured:

For clarity, the constraints are mykey = myvalue and bing = b.*p.

For clarity, the constraints are mykey = myvalue and bing = b.*p.

The following message payload would be accepted:

{
  "mykey": "myvalue",
  "bing": "boooop",
  "x": ["1", "2", "3"]
}

But this message payload would be rejected (pipeline would not trigger):

{
  "mykey": "myvalue",
  "x": ["1", "2", "3"]
}

Passing parameters

Say your pipeline accepted some parameters (for example, the desired stack to deploy to), you can make this explicit by adding a pipeline parameter on the same configuration screen as the Pub/Sub trigger:

For more information on how to use pipeline parameters, see the pipeline expressions guide.

For more information on how to use pipeline parameters, see the pipeline expressions guide.

Warning: there are several reserved parameter keys (names) that cause unexpected behavior and failures if overwritten by a pipeline parameter definition. See the list of reserved parameter and evaluate variable key names .

If you were to manually execute this pipeline, you would be prompted with the following dialogue:

If instead you were to trigger this pipeline with a Pub/Sub Trigger, you could supply each parameter a value inside a key/value map called parameters in the message body. Take the following message payload for example:

{
  "parameters": {
    "stack": "prod"
  }
}

☞ Note: If you had selected the Required checkbox for a parameter without providing a default, the pipeline doesn’t trigger if a parameter is not present. The difference between this and the preconditions covered earlier is that when a precondition isn’t met, Spinnaker will not even try to run the pipeline. However, when a required parameter doesn’t exist, Spinnaker will try and fail to run a pipeline, surfacing a “Failed Execution” in the UI.

Passing artifacts

If your pipeline requires artifacts (for example, a Kubernetes manifest file stored in GCS), you can make this explicit by defining an Expected Artifact and assigning it to the Pub/Sub Trigger. Under Artifact Constraints, select Define a new artifact… and supply the artifact details, as shown below:

In order for this to work, you need to supply the required artifact in the Pub/Sub message payload, and configure Spinnaker so that it can translate the Pub/Sub payload into a Spinnaker artifact.

If you’re using GCR, you can use the --message-format flag and Spinnaker will translate the payload automatically:

hal config pubsub google subscription edit my-gcr-subscription \
  --message-format GCR

Otherwise, you need to supply a translation template so that Spinnaker can translate the Pub/Sub payload into a Spinnaker artifact. To do this, create a Jinja template . Note that the output of the Jinja transform must be a JSON list of Spinnaker artifacts. The translation template itself can be any valid Jinja transform.

Use the following hal command to tell Spinnaker how to find the template:

hal config pubsub google subscription edit my-gcs-subscription \
  --template-path /path/to/jinja/template

Example

Let’s say you have a message payload containing the following structure:

{
  ...
  "location": "gs://jtk54-artifacts/manifest.yml"
  ...
}

You can use the following Jinja template to translate the above into the Spinnaker artifact format:

[
  {
    "type": "gcs/object", # static type.
    "reference": "{{"{{ location "}}}}", # 'location' in the Pub/Sub payload.
  }
]