Figuring out why a particular RGW request is painful and often requires annoyingly verbose debug levels. Gathering the logs and making sense of them is not an easy task, especially when load balancers and multiple radosgw instances are involved.

It would be fantastic if the RGW extended the "blkin" work [1] such that a client could originate the trace identifier. With this implemented end to end we could see a marvelous waterfall chart of where latencies come from.

While typically reserved for response headers, perhaps we could support either "x-amz-id-2" or "x-amz-request-id" as request headers to pass the trace identifier.

If someone is using a proxy, eg. HAproxy, they would of course need to configure it to pass the header along.


[1] https://github.com/ceph/blkin
[2] https://docs.aws.amazon.com/AmazonS3/latest/API/RESTCommonResponseHeaders.html