Github user steveloughran commented on the issue:
https://github.com/apache/spark/pull/20490
@rdblue thanks. That was what I thought (the output coordinator doesn't
tell incoming speculative work to abort until any actively committing task
attempt has returned, I was just worried after the conversation.
In a lot of the Hadoop FS code, InterruptedException is converted to
`InterruptedIOException` to allow it to trickle up, but as other IOEs subclass
that (socket/connect timeouts), you can't assume that `InterruptedIOException`
implied job was interrupted, only that some IO went wrong.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
