How to handle gzip'd gzip'd (twice) files in S3

[markwan (Mark)]

A client is providing AWS cloud trail logs via an S3 bucket. The S3 input plugin will read .gz files.

However the problem is that the file, eg "cloudtrail_log_2019-01-22.gz" is actually gzip'd twice, and should really be named cloudtrail_log_2019-01-22.gz.gz"

I have tried the gzip_lines codec but that assumes each line in a file is gzip'd not the whole file.

Has anyone else seen this problem and found a solution? Right now I'm thinking about a pipeline using s3 input which just unzips and outputs to a file on the local drive, but suffixing with a .gz on the inner gzip file. Then a second pipeline which reads from that using file codec on the new .gz file.

Thoughts? Comments?

[markwan (Mark)]

Client has changed data. Now the files in S3 are not double gzip'd. However the file does not have a .gz extension. This means s3 input does not automatically recognise data as gzip'd.

Using gzip_lines codec looks like it's looking for each line to be gzip data, not a whole file that is gzip'd. So data cannot be parsed using that codec.

Is the only solution to extend the s3 input plugin to allow an optional is_gzip flag that forces gzip behaviour? or is there a simpler solution available?