Automatic rebalancing of container-based services for high availability

A system comprising:

• a first one or more electronic devices to implement a container service in a multi-tenant provider network, wherein the container service includes instructions that upon execution cause the container service to:
• launch a plurality of containers, wherein each container of the plurality of containers is executed on computing infrastructure within one of a plurality of availability zones (AZs) of the multi-tenant provider network, and wherein the container service places each container in a particular AZ of the plurality of AZs according to a placement strategy,
• identify an actual distribution of the plurality of containers across the plurality of AZs,
• determine that the actual distribution of the plurality of containers differs from an expected distribution of the plurality of containers based on the placement strategy, and
• rebalance placement of the plurality of containers by:
• terminating at least one container of the plurality of containers in a first AZ of the plurality of AZs, and
• launching at least one new container in a second AZ of the plurality of AZs;
• and
• a second one or more electronic devices to implement an AZ performance monitoring service in the multi-tenant provider network, wherein the AZ performance monitoring service including instructions that upon execution cause the AZ performance monitoring service to:
• identify an operational issue affecting the first AZ, wherein the operational issue affects performance of containers placed by the container service in the first AZ, and
• provide data identifying the operational issue affecting the first AZ to the container service.

A computer-implemented method comprising:

• launching, by a container service of a cloud provider network, a plurality of containers, wherein the plurality of containers is associated with a task or service definition defining parameters for launching the plurality of containers, wherein each container of the plurality of containers is executed on computing infrastructure within one of a plurality of availability zones (AZs) of the cloud provider network, and wherein the container service places each container in a particular AZ of the plurality of AZs according to a placement strategy;
• identifying an actual distribution of the plurality of containers across the plurality of AZs;
• determining that the actual distribution of the plurality of containers across the plurality of AZs differs from an expected distribution of the plurality of containers across the plurality of AZs based on the placement strategy;
• receiving a request to exclude the first AZ from the plurality of AZs as a possible destination for the container service to launch containers based on the task or service definition;
• and
• responsive to the request to exclude the first AZ, rebalancing placement of the plurality of containers by:
• terminating at least one container of the plurality of containers in a first AZ of the plurality of AZs, and
• responsive to the terminating at least one container, launching at least one new container in a second AZ of the plurality of AZs.