Hi All, Installation of flink clusters where we govern the memory configs by specifying 'taskmanager.memory.process.size' TM memory is auto calculated.
But in this calculation we generally don't check the actual physical memory available. So in case of a wrong configuration where the JVM heap size calculation is below the physical memory but the direct memory goes beyond the physical memory we don't fail the process creation but allow the process to be started. This is a naive question but just wanted to know if it is by design. The reason is that it actually misleads the user as the UI shows the calculated value but in reality memory crunch will lead to OOME at some point. Should we fail the process startup in such cases? Regards Ram
