Elphas Toringepi created FLINK-33068:
----------------------------------------
Summary: Handle DNS errors in AWS connectors as recoverable
Key: FLINK-33068
URL: https://issues.apache.org/jira/browse/FLINK-33068
Project: Flink
Issue Type: Improvement
Components: Connectors / AWS
Affects Versions: 1.17.1, 1.16.2, 1.15.4
Reporter: Elphas Toringepi
Fix For: 1.17.1, 1.16.2, 1.15.4
AWS connectors are not treating DNS errors as recoverable using exponential
back-off resulting in Flink job restarts on transient errors and small timeouts.
DNS errors typically manifest as UnknownHostException which is not handled as a
recoverable exception and potentially not retried.
Here is an [example of an
exception|https://github.com/apache/flink-connector-aws/blob/21c64ce38d3a6fd4fd9b4abda416c2de1a926780/flink-connector-aws/flink-connector-kinesis/src/main/java/org/apache/flink/streaming/connectors/kinesis/proxy/KinesisProxy.java#L459]
that is recoverable with exponential-backoff.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
