Customer is receiving a timeout error

Customer is receiving a timeout error for long files

Introduction

If a request hasn’t received a 200 response for a long period of time it may time out. When a request times out, the JSON response will return this error message:

“error”: “Server error, developers have been alerted.” When we check our datadog logs with the transcript_id for the request, you may find a message that indicates that the transcript timed out, like the following:

jobs did not finish within timeout period

Procedures

1. Ask the customer for a transcript id and the audio file.

2. Open Datadog and search @transcript_id:<transcript-id> on the logs page.

3. Once the logs have finished loading, look for logs that are colour coded red. If you click on the logs, a side panel will appear with more information.

4. If you find jobs did not finish within timeout period among the logs, that confirms the request timed out.

5. Try to reproduce the error with the audio file. To do this, process the file with the same parameters used in the customer’s request. You can find the features/models used by searching “Received transcript:” “<transcript-id>” and clicking on the “Received transcript” log message.

6. If you find that the error is reproducible using the same features/models, try processing the audio file with every model we offer, one by one. This is so that we can provide engineering with additional info on which models are unable to complete before a transcript request times out for long audio files.

7. Once you have the information on which models/features result in a time-out error, create a bug report with your findings.