Set firelens mem_buf_limit by default

[yinyic]

Summary

awsfirelens logging driver uses a Fluentbit-based image. Fluentbit supports mem_buf_limit configuration parameter which controls the amount of buffer memory that can be used by a log input. This prevents excessive memory usage of the buffer which may lead to container and task OOM kill. (see https://docs.fluentbit.io/manual/administration/backpressure)

mem_buf_limit is not enabled by Fluentbit by default, making buffer unbounded. In this PR we let ECS Agent sets a default mem_buf_limit when generating Fluentbit configuration.

Implementation details

The value of the mem_buf_limit is determined as follows:

1. if firelens container memoryReservation is specified, mem_buf_limit = memoryReservation / 2. We resolve memoryReservation from container host config property (a json blob).
2. else if firelens container memory is specified, mem_buf_limit = memory / 2. We resolve memory from container property directly.
   (memoryReservation and memory are optional parameters indicating container soft and hard memory limit respectively. The goal here is to scale the mem_buf_limit in accordance with the memory allocation for the firelens container. The scaling factor is derived from this recommendation)
3. if neither memoryReservation nor memory is specified, we set mem_buf_limit to a fixed amount of 25MB (rationale)

Main changes are:

• Updated NewFirelensResource method to take in a new containerMemoryLimit argument, and updated caller to pass in this argument after resolving the value from container memory and memoryReservation
• Added container helper method GetMemoryReservationFromHostConfig to retrieve container memoryReservation
• Updated firelens generateConfig method to write Mem_Buf_Limit config for the default input, after resolving the value from resolveMemBufLimit helper method

Testing

New tests cover the changes: yes

Description for the changelog

Set firelens mem_buf_limit by default

Additional Information

Does this PR include breaking model changes? If so, Have you added transformation functions?

N/A

Does this PR include the addition of new environment variables in the README?

N/A

Licensing

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.