0

I deployed Heron cluster using aurora-scheduler and Mesos. And when I ran the default WordCountTopology using this cluster, I found the ram demand of aurora task is 4G. However, the WordCountToplogy's configuration as follows:

componentRam: 1G
containerRamRequested: 1G
containerCpuRequested: 2 cores
containerDiskRequeted: 2G

Aurora task.json content is: enter image description here

It shows that this task of aurora needs 4g ram resources. But I don't know why it requests 4G ram. And how to modified this ram requirement? In addition, there are two slave hosts in my heron cluster and these host resources is: enter image description here

Jonas
  • 121,568
  • 97
  • 310
  • 388
Yitian Zhang
  • 314
  • 1
  • 3
  • 18

1 Answers1

1

In addition to the Ram requested by topology's components, there are some additional resources(cpu, memory) requested for heron's daemon processes,e.g stream-manager. Packing additional CPU in RR

A second cause for the larger resource request is due to Aurora only allows homogeneous containers. The packing algorithm will pick maximum container resources as the resource request for all the containers. For example, if a topology has two containers: one requests 2 cpus and the other requests 3 cpus. Then the eventually all containers will request 3 cpus.

Neng
  • 146
  • 2