Using the “X-AMZN-Trace-Id” Header for Request Tracing Through AWS Load Balancer

Request ID! I added that to the Amazon web servers about 20 years ago! My code probably doesn’t exist anymore but it’s funny to see it referenced on HN this many years later.

I recently implemented support for this in the prometheus-exporter gem.

https://github.com/discourse/prometheus_exporter/pull/150

What's annoying about the encoded time value is version 1 (the only version still) is in seconds precision. Makes the value non-viable for any sort of queue time metric. Better than nothing, but would have been nice if it were micro.

I think this the AWS docs explain this fairly well.

https://docs.aws.amazon.com/elasticloadbalancing/latest/appl...

The format for the field is

     X-Amzn-Trace-Id: Field=version-time-id

and you're allowed to add custom fields that don't follow this format and they will be preserved.

Have a look at this propagator from Opentelemetry: https://github.com/open-telemetry/opentelemetry-go-contrib/b...

If you extract this trace id and use it as your request id you can follow it from the load balancer all the way to the end.

I wonder if the system propagates "baggage" headers. OpenTelemetry uses this header to propagate span/trace IDs.

https://www.w3.org/TR/baggage/

Wasnt there an rfc recommending against the use of X- prefix headers? I guess that didnt hold