Open
Description
We noticed that certain unretryable unstructured
errors like negative seek value -1
and [Errno 22] Invalid argument
were being swallowed by the default retry mechanism provided by the JS client.
Since the JS client does not allow overriding the retryable statusCode
values to something more restrictive (e.g. we only want to retry 503
), we created our own BackoffStrategy
in order to fail sooner, e.g.
new BackoffStrategy(
500, // initialInterval (ms)
10 * 1000, // maxInterval (ms)
1.5, // exponent
20 * 1000, // maxElapsedTime (ms)
)
However, in our testing, we found that no retry requests are ever actually being sent. We suspect that this is due to how Speakeasy may be consuming the file buffer the first time but we cannot confirm. For now, we've disabled retries at the client level entirely by setting strategy: "none"
but this is not our ideal state.
We are on version 0.10.6.
TL;DR
- Can you make the retry status codes configurable to the end user?
- Are we using the retry mechanism correctly? Can you confirm our behavior vs the expected behavior?
Metadata
Metadata
Assignees
Labels
No labels