Thanks, We ran into differnet errors and then realized it was OOM issue
which was causing different parts to be failed.
Flink was buffering too much data as we were reading too fast from source.
Reducing the speed fixed the issue.
However, I am curious how to achieve the same with S3 apart from l
This seems to be irrelevant to the issue for KyroSerializer in recent
discussions [1]. which has been fixed in 1.4.3, 1.5.0 and 1.6.0.
On a quick glance, this might have been a corrupted message in your
decoding, for example a malformed JSON string.
--
Rong
[1] https://issues.apache.org/jira/brow
