Sam Whittle created BEAM-10808:
----------------------------------
Summary: StreamingDataflowWorker streaming rpcs due not always
observe stream failures, until timeout
Key: BEAM-10808
URL: https://issues.apache.org/jira/browse/BEAM-10808
Project: Beam
Issue Type: Bug
Components: runner-dataflow
Reporter: Sam Whittle
Assignee: Sam Whittle
The StreamingDataflow worker communicates to Streaming Engine with grpc
streaming rpcs. Occasionaly with service failures, the error is not observed by
the streaming rpc until the deadline is exceeded. As the deadline is multiple
minutes, this adds unnecessary latency to pipelines that are otherwise keeping
up with seconds delay.
Testing has shown that grpc channel health checks are not sufficient but that
application level health checks ensure that streaming rpc health issues are
detected earlier than the stream timeout.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
