Our current ECS setup only allows having a fixed amount of tasks per service.

We should be able to scale out/in on demand depending on utilization, to handle traffic spikes more gracefully

This issue is about setting up terraform to support auto scaling for ECS tasks

Additional considerations

Fixed memory limits: The idea should be to have fixed memory / cpu limits and set up percentage thresholds. For example: "if memory usage is > 60%, spawn a new instance".

This might be an issue with oonimeasurements, since it tends to have sharp spikes in memory usage that go well over 4000% of its default memory allocation. We should think about whether we really want fixed memory for this service

Example memory usage of ooniprobe:

Example memory usage of oonimeasurements:

Also note that the memory allocations should also be adjusted to keep them under 100% on normal load