Closed Spycsh closed 2 years ago
Flink network stack is based on Netty and it will distribute memory locally (off-heap) for the network buffer. The state transmitted for large time interval may exceed the buffer and cause the connection to be closed. Increasing the parameter statefun.feedback.memory.size
may fix this problem as shown in link
In the latest version, it is fixed by rewriting the application logic.
Undertow exception:
java.io.IOException: UT000128: Remote peer closed connection before all data could be read
when handling queries with start time = 0, end time = 10000000 on email-EU-core dataset, under storage paradigm 3 (TM),
Undertow throughput and undertow network internal timeout need to be explored for future work