Retry initiateMultipartUpload

[sopel39]

# Hive/Iceberg/Delta
* Improve resilience of insert queries. ({issue}`issuenumber`)

[findepi]

cc @alexjo2144 @findinpath @homar

[alexjo2144]

Does this happen often? Curious what kind of errors from AWS are you seeing

[findinpath]

Could you please add some context / description to this fix?
It would be helpful to know in which situations does the initiation of the multipart upload request fail.

[sopel39]

The failure can happen during scaled insert when there is high pressure on S3. Usually these are timeout exceptions

[findinpath]

Usually these are timeout exceptions

Why don't we add a retry mechanism for the specific case of timeouts then?

[sopel39]

@findinpath this

                            if (e instanceof AmazonS3Exception) {
                                switch (((AmazonS3Exception) e).getStatusCode()) {
                                    case HTTP_FORBIDDEN, HTTP_NOT_FOUND, HTTP_BAD_REQUEST -> throw new UnrecoverableS3OperationException(bucket, key, e);
                                }
                            }

is what you will find in other places that retry S3 calls (in one form or other)