You are viewing a plain text version of this content. The canonical link for it is here.

Posted to user@hadoop.apache.org by Fabio <an...@gmail.com> on 2015/01/20 10:05:00 UTC

Capacity scheduler locality allocation

Hi everyone,
I am testing the capacity scheduler with SLS. I am simulating _a single 
node_. When an application requires, let's say, 46 mappers, the request 
will be like:

{Priority: 20, Capability: <memory:1024, vCores:1>, # Containers: 46, 
Location: *, Relax Locality: true}
{Priority: 20, Capability: <memory:1024, vCores:1>, # Containers: 46, 
Location: node1, Relax Locality: true}
{Priority: 20, Capability: <memory:1024, vCores:1>, # Containers: 46, 
Location: default-rack, Relax Locality: true}

But then I see the containers are assigned as OFF-SWITCH, thus going to 
satisfy first the "Location: *" request.
Why is it so? Since I have a single node I was expecting all the 
assignment to be node-local. Is this an unusual behavior or is it 
expected in your opinion?

Thanks

Fabio