S3Client guidance in long-running multi-threaded process

[rvernica]

I have a long-running (in months) process that is multi-threaded. Once in a while, it does a set of S3 requests (get, put, delete, list). What is the best way of using the S3Client class?

1. Should I have one S3Client instance per process?
2. Should I have one S3Client instance per batch of requests?
3. Should I have one S3Client instance for each request?

We tried all three but not sure which one is best. Our initial implementation used option (2) which worked fine for about a year. Recently we consistently run into this issue. Once the issue happens in a batch, any new S3Client we create will have this issue. Since then, we tried option (1) and option (3). In both, we have avoided the issue but we are not sure which is the recommended use.

[jmklix]

We recommend option #1 because it will be to most resource efficient and work best with this sdk.

[github-actions[bot]]

Hello! Reopening this discussion to make it searchable.