[
https://issues.apache.org/jira/browse/ARROW-17079?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17567141#comment-17567141
]
Philipp Moritz commented on ARROW-17079:
----------------------------------------
Thanks for your quick answer!
For my use case, a better printout with the right error message and the AWS SDK
call that failed would already go a very long way and would probably be
sufficient. A format that allows programmatic extraction would be nice at some
point in the future since it would allow better error messages for higher level
libraries, but for now we could just print the raw error message from Arrow :)
> Improve error message propagation from AWS SDK
> ----------------------------------------------
>
> Key: ARROW-17079
> URL: https://issues.apache.org/jira/browse/ARROW-17079
> Project: Apache Arrow
> Issue Type: Improvement
> Components: C++
> Affects Versions: 8.0.0
> Reporter: Philipp Moritz
> Priority: Minor
>
> Dear all,
> I'd like to see if there is interest to improve the error messages that
> originate from the AWS SDK. Especially for loading datasets from S3, there
> are many things that can go wrong and the error messages that (Py)Arrow gives
> are not always the most actionable, especially if the call involves many
> different SDK functions. In particular, it would be great to have the
> following attached to each error message:
> * A machine parseable status code from the AWS SDK
> * Information as to exactly which AWS SDK call failed, so it can be
> disambiguated for Arrow API calls that use multiple AWS SDK calls
> In the ideal case, as a developer I could reconstruct the AWS SDK call that
> failed from the error message (e.g. in a form the allows me to run the API
> call via the "aws" CLI program) so I can debug errors and see how they relate
> to my AWS infrastructure. Any progress in this direction would be super
> helpful.
>
> For context: I recently was debugging some permissioning issues in S3 based
> on the current error codes and it was pretty hard to figure out what was
> going on (see
> [https://github.com/ray-project/ray/issues/19799#issuecomment-1185035602).]
>
> I'm happy to take a stab at this problem but might need some help. Is
> implementing a custom StatusDetail class for AWS errors and propagating
> errors that way the right hunch here?
> [https://github.com/apache/arrow/blob/50f6fcad6cc09c06e78dcd09ad07218b86e689de/cpp/src/arrow/status.h#L110]
>
> All the best,
> Philipp.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
