Google Cloud Deploy Adds Canary and Parallel Deployment Support

Published on April 29, 2023

Google Cloud Deploy now supports canary and parallel deployments. Both features are compatible with Google Kubernetes Engine, Cloud Run, and Anthos. For more advanced rollout strategies, the features can be combined.

Using canary deployments, traffic can be split between the old release and the new release in order to facilitate the validation that the new code is working as expected. Typically, traffic is split so that the new release receives a small percentage of the overall traffic. As validation and testing demonstrate that the new code release is stable, more traffic can be switched over to it. In Google Cloud Deploy, you can advance the percentage of traffic directed to the new release by either using the console or by calling gcloud deploy beta rollouts advance.

Both phase and job status and rollout logs are available for troubleshooting purposes. Failing jobs can be retried, terminated, or ignored as necessary. Furthermore, the entire canary deployment can be canceled or rolled back.

If one deployment fails, the entire rollout fails. Parallel deployments can deploy multiple targets in lockstep. As a result, the various deployments are treated as one deployment, where all targets must pass in order to move onto the stage. When used with a multi-target, a controller rollout orchestrates multiple child rollouts. Each child rollout operates within its own execution environment and performs its own deployment.

As a result, qsprod-a and qsprod-b are child targets of qsprod-multi. In order for qsprod-multi to be considered successful, both qsprod-a and qsprod-b must be successful.

It is possible to combine canary deployments with parallel deployments. When combined, it is not possible to advance a child rollout directly, but only the controller rollout can be advanced. As a result, all child rollouts are automatically advanced. In this manner, child rollouts are prevented from advancing ahead of other children.

Within a combined deployment, only child rollouts can be retried or ignored. Cancelling a rollout, however, can only be done at the controller level. The entire rollout will remain in the IN PROGRESS status if a child rollout fails, but some children are still marked as IN PROGRESS.

API Gateway can be used to set up canary deployments within AWS. A percentage of traffic can be directed to the new code using canary settings on a deployment stage. In Azure, canary deployments can be performed using the Kubernetes manifest task’s canary strategy

You can find more information about the release on the Google Cloud blog or within the tutorials and documentation. The Google Cloud community forums are a good place to ask questions and provide feedback.