Service Revisions
Last updated
Last updated
In VESSL, a Revision holds important information about how an inference server works. It's like a recipe that includes details about the model used, the resources it needs, how it can automatically adjust itself, and where it's accessible. Since models can change over time, VESSL encourages to manage revisions make it easy to manage different setups and go back to older ones if needed.
To start a new Revision, go to the list of Revisions and click on "New revision."
When you click "New revision," you'll see a page where you can set up the new Revision. This is where you tell VESSL which model to use, how much power it needs, whether it should grow or shrink automatically, and how people will connect to it.
Here, you add some notes about what this Revision is for:
Metadata: Metadata of the revision.
Message: Write a short message explaining what this Revision is meant for.
Deployment Spec: Resource requirements and configuration for the Revision.
Resource: CPU, RAM, and GPU resources to allocate to the Revision.
Docker Image: The Docker Image to use for the Revision.
Start Command: The command to run inside the container. This is like running a command in the terminal on your computer.
Environment Variables: Environment variables to inject into the container.
Port: The port to expose from the container. For example, if you're using a BentoML model server, you'll want to expose port 3000 and use the HTTP protocol to access the service endpoint.
Advanced Options: Additional configuration for the Revision.
Autoscaling: Automatically scale the Revision up or down based on demand.
Min: The minimum number of replicas to keep running.
Max: The maximum number of replicas to scale out.
Target metric: The metric to use for scaling up or down.
CPU: The CPU usage of the Revision.
Memory: The memory usage of the Revision.
GPU: The GPU usage of the Revision.
Launch this revision immediately: Start the Revision as soon as it's created.
Once you create a Revision, you can see basic information about the Revision in the list of revisions and quickly perform actions needed for service operation.
Actions on the Revision list are as follows:
Start: Deploy the Revision to the cluster immediately according to the current settings.
Stop: Immediately stop the Revision that is deployed and running.
Scale: Adjust the number of replicas of the deployed and running Revision.
Reproduce: Create a new Revision with the same settings as the selected Revision.
Delete: Delete the selected Revision.
Volume Mount: You can mount datasets, model files, code, and more as folders for use in Revision. For more information, please see
Service Account Name: The Kubernetes service account to connect to the container. This is commonly used with , , to control what cloud resources the container can access.
