Arnaud Linz created FLINK-5305:
----------------------------------
Summary: Documentation of collect() should mention parameter akka.framesize
Key: FLINK-5305
URL:
https://issues.apache.org/jira/browse/FLINK-5305 Project: Flink
Issue Type: Improvement
Components: Documentation
Affects Versions: 1.1.3
Reporter: Arnaud Linz
Priority: Minor
When calling collect() on a big dataset, if the data size exceeds the value of akka.framesize, which is 10Mb by default, the collect() freezes and some "akka payload size exceeded" logs appear in the yarn logs (but neither in the flink console output nor UI logs).
10Mb is not that big, and its relation to akka.framesize is hard to find in the documentation (in Flink internals section). I believe that the "easy to find" collect documentation should mention this parameter to save hours of users time.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)