dlt uses several containers in its test setup. Most define the policy restart: unless-stopped in their docker-compose.

This means these containers are started when booting the developer's machine and run most of the time in the background. This has a noticeable impact on CPU during intensive usage. The developer is required to manually spin down containers each time via docker container stop $(docker ps -q -a)

Changes

• set policy restart: "on-failure" which doesn't start containers on boot.
• remove the version: field; this has been deprecated and ignored for some time
• set container_name with a dlt_test_ prefix to facilitate container management and prevent spawning duplicate containers

Future steps

• some docker compose use docker networks and named volumes; those can be harder to cleanup. Would be useful to minimize resource usage of dev machines or provide cleanup scripts