Github user NicoK commented on a diff in the pull request:
https://github.com/apache/flink/pull/4552#discussion_r161578518
--- Diff:
flink-tests/src/test/java/org/apache/flink/test/misc/SuccessAfterNetworkBuffersFailureITCase.java
---
@@ -59,7 +59,7 @@ public void testSuccessfulProgramAfterFailure() {
config.setInteger(ConfigConstants.LOCAL_NUMBER_TASK_MANAGER, 2);
config.setLong(TaskManagerOptions.MANAGED_MEMORY_SIZE,
80L);
config.setInteger(ConfigConstants.TASK_MANAGER_NUM_TASK_SLOTS, 8);
-
config.setInteger(TaskManagerOptions.NETWORK_NUM_BUFFERS, 800);
+
config.setInteger(TaskManagerOptions.NETWORK_NUM_BUFFERS, 1024);
--- End diff --
Is that also the reason here? I see that otherwise we get into
`Insufficient number of network buffers` but it does not look as if it was
configured as tightly...
(just want to rule out some memory leak with the new code)
---
