Wesley Pettit

With it we get: ``` [2022/08/16 21:37:27] [ warn] [output:s3:s3.0] Buffer is almost full: current_buffer_size=19483080, store_dir_limit_size=20000000 bytes [2022/08/16 21:37:33] [error] [output:s3:s3.0] Buffer is full: current_buffer_size=19997432, new_data=41193, store_dir_limit_size=20000000 bytes ```

Doc PR: https://github.com/fluent/fluent-bit-docs/pull/956/files

@Claych I think a commit message that makes it more clear why we made this change would be something like: `out_s3: use configured retry_limit instead of hardcoded MAX_UPLOAD_ERRORS`

@Claych The commit messages are used to create the changelog, that's why its important they are clear and end-user friendly IMO. Please change the commit message to something like what...

@Claych when this is ready to merge, this should be one commit.

@mwarzynski Thank you for contributing this! I have read the description of the PR and I agree with your thinking/plan at a high level. I think simply fetching all tags...

@mwarzynski I performed some initial review, you may ping me once all of these comments are addressed for the next round of review. Thanks!

@mwarzynski Sorry I should have pinged this before, if possible, it would be ideal if we can re-use the stuff built by @matthewfala for mocking IMDS: https://github.com/fluent/fluent-bit/blob/master/tests/internal/aws_credentials_ec2.c#L92 https://github.com/fluent/fluent-bit/blob/master/tests/internal/aws_client_mock.c > It...

@mwarzynski Thanks for the doc PR, that's all that is needed on that front: https://github.com/fluent/fluent-bit-docs/pull/897 I will review that and this soon. I was on vacation last week.

@mwarzynski Before merge, this should be two commits probably. One for all aws filter code changes, and one for the testing changes.