elastic should trace the response even http status code is not 2xx

[Jack47]

Please use the following questions as a guideline to help me answer
your issue/question without further inquiry. Thank you.

Which version of Elastic are you using?

[ ] elastic.v2 (for Elasticsearch 1.x)
[X] elastic.v3 (for Elasticsearch 2.x)

Please describe the expected behavior

elastic should trace the response even the http status code is not 2xx.

Please describe the actual behavior

I want to find specific doc by it's id, this doc doesn't exist in ES, then elastic returned an error with status code 401. I set trace log for elastic and try to figure out what happened, but I can't see the corresponding http response in trace log.

Any steps to reproduce the behavior?

Try to find an not exist doc by it's id

[olivere]

I'll investigate if I can change Client.PerformRequest accordingly. Thanks for reporting.

[freb]

I believe the issue I'm having is the same. I had a scroll request that would return an error on:

results, err := scroll.Do(ctx)

Which I then use to panic. The panic takes place because the response is a 400, but I only get a portion of the actual response that is included in err and not the full trace as configured on the client. This is all I see:

panic: elastic: Error 400 (Bad Request): all shards failed [type=search_phase_execution_exception]

I ended up using the trace output of the request to send a curl request. It's response included the additional detail I needed:

"reason":"No mapping found for [flow.] in order to sort on"

Adding reason to the error or tracing the raw response before the error is returned would help.

[olivere]

@freb Yes, fixing the tracing is an open issue, but it's fixed soon by #553. You can get access to all error details that ES returns by type checking on Error.

[freb]

Sorry! I didn't see the separate issue. Good to know about type checking Error, I hadn't thought of that.

Anyway, thanks for the awesome library. I'm enjoying it much more than official Python library.

[olivere]

Closing for it is too old. If you still see this as a problem, please file a new issue and link back to this one. Thanks.