It is okay when some subprocesses respawn, they have to finish finally with the expected exit code. The reason some processes depend on mysql which sometimes starts later but it’s not a problem at all.
I have tried running PMM Server on Amazon ECS myself, I got success running “pmm-server” container. However, it was not clear to me what to do with the data volumes. Looks like “pmm-data” container is something not suitable for ECS, as all containers there should be disposable there (one off tasks). So I assume you have to manage volumes manually. Then it’s a question how to create them w/o being initialized first on the pmm-server image.
It seems to me it’s easier to install docker on EC2 and spin up containers yourself. You probably don’t need to scale “pmm-server” in the way ECS designed for (web/app front/end backends).