You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@ambari.apache.org by "robinlin (JIRA)" <ji...@apache.org> on 2015/08/23 06:07:45 UTC
[jira] [Updated] (AMBARI-12860) Ambari in Docker Containers
[ https://issues.apache.org/jira/browse/AMBARI-12860?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
robinlin updated AMBARI-12860:
------------------------------
Description:
Hi
I try the Ambari 2.1 and HDP 2.3 on the Docker Containers which is provisioned by Openstack. And the containers were provisioned on bare-metal machines.
But after the stack was ready, I saw few problems
1) Metrics on utilized CPU and Memory are from physical host not the Docker Containers.
2) The value of properties {noformat}yarn.nodemanager.resource.memory-mb{noformat}
and {noformat}yarn.nodemanager.resource.vcpu-vcores{noformat}
are calculated based on the CPU and Memory on physical machine, which may hosts many Docker containers.
Is there any plan to make these right?
was:
Hi
I try the Ambari 2.1 and HDP 2.3 on the Docker Containers which is provisioned by Openstack. And the containers were provisioned on bare-metal machines.
But after the stack was ready, I saw a problem, the metrics on utilized CPU and Memory are from physical host not the Docker Containers.
Is there any plan to make this right?
Summary: Ambari in Docker Containers (was: Ambari in Docker Containers should acquire the CPU and Memory from (Docker)container stats)
> Ambari in Docker Containers
> ---------------------------
>
> Key: AMBARI-12860
> URL: https://issues.apache.org/jira/browse/AMBARI-12860
> Project: Ambari
> Issue Type: Improvement
> Components: ambari-metrics, ambari-web
> Affects Versions: 2.1.0, trunk
> Environment: Openstack, nova-docker driver, Docker container on bare-metal machine
> Reporter: robinlin
>
> Hi
> I try the Ambari 2.1 and HDP 2.3 on the Docker Containers which is provisioned by Openstack. And the containers were provisioned on bare-metal machines.
> But after the stack was ready, I saw few problems
> 1) Metrics on utilized CPU and Memory are from physical host not the Docker Containers.
> 2) The value of properties {noformat}yarn.nodemanager.resource.memory-mb{noformat}
> and {noformat}yarn.nodemanager.resource.vcpu-vcores{noformat}
> are calculated based on the CPU and Memory on physical machine, which may hosts many Docker containers.
> Is there any plan to make these right?
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)