An ECS service rolling deployment stops every running task before any replacement is healthy. The service is configured so the minimum healthy percent is zero. What should change so the service keeps serving?
Select an answer to reveal the explanation.
Short Explanation
A renovation that empties the store before the new shelves arrive is a closed store. Your minimum healthy percent is zero, so ECS may stop every task before a replacement is healthy. Raise that floor so a task keeps serving while new ones start.
Full Explanation
minimumHealthyPercent is the lower limit, as a percent of desired count rounded up, on tasks that must stay running during a rolling deployment. At zero that limit is no tasks, so the scheduler may stop every running task before a replacement is healthy. Raising it holds running capacity in place while new tasks start. Setting the desired count to zero is an outage you perform by hand, and restoring it later does not keep a task up during the deploy. A maximum percent of zero does not leave room to launch a replacement. The daemon scheduling strategy runs one task per container instance. It is a different service shape, and it does not repair the rolling-deployment floor on this replica service. Caveat: the upper percent is rounded down and must still allow a replacement to start. If the ceiling and the floor leave no room to launch, the deployment cannot move even after the floor is raised. Operational check: start a deployment and confirm the running task count never drops to zero, and confirm a new task is healthy before the task it replaces is stopped.