MS Batch Translation -> Add option to abort after a certain number of errors

Issue #507 resolved
Fredrik L created an issue

Trying to translate a large file using an expired MS Hub Engine takes a long time. The step still tries to translate every single segment multiple time and receives error code 400. In the case of the expired engine every segment will fail and the error logged. Might be useful to be able to abort after a certain number of failures.

Comments (4)

  1. Chase Tingley

    In general, the error handling on this connector is not good. IIRC, some errors are reported as "unable to get authorization token" when they are actually a different error.

    Also (and this is a separate issue), the connector doesn't deal with MS Hub's internal limit on batch queries (10k chars/request).

  2. Fredrik L reporter

    Thanks for the info Chase, I'm not familiar with that "unable to get authorization token" issue. For the 10k limit maybe we could add some character counter, but I think it's not something we run into very often after we lowered the batch size.

    In this case we're running it as a Rainbow cmd pipeline so just thinking higher level in the Step counting the number of errors from the Connector and basically skipping sending it to the connector if there are to many errors. If the Hub engine is undeployed, quote exceeded, or the credentials are wrong(?), every call will fail after multiple attempts like:

    ERROR: Query response code: 400: Bad Request

    ERROR: Query response code: 400: Bad Request

    ERROR: Query response code: 400: Bad Request

    ERROR: Failed to get Microsoft Translator access token after 3 tries. Skipping query for <segment content>

    So all remaining http attempts are unnecessary and the log file gets bloated.

  3. Chase Tingley

    Ahh, that 400 error may be what I was thinking of -- this happened to me a few months ago and my memory was fuzzy. I thought there was an "unable to get authorization token" message somewhere in the exception handling, but maybe not. Either way, it could be better.

  4. Log in to comment