Upgrading Containers in a Fargate Cluster in ECS
We have containers running as tasks in a Fargate cluster in ECS. We would like to upgrade, but I want to confirm the following default procedure is acceptable:
- Old task (Container) is running
- New task (container) is created with newer image
- New task becomes healthy
- Old task is removed
The concern is that an older version of Decisions is still connected to the database while a newer version is booting up and upgrading the database.
Comments
-
There should not be an issue when performing rolling updates for minor upgrade. The default recommendation is to perform this during a maintenance window. A for your particular use case, I would recommend you first test this on a UAT/QA environment to ensure 100% uptime, and manually test your major processes with unit tests for 100% certainty.