[ 
https://issues.apache.org/jira/browse/YARN-8734?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16614199#comment-16614199
 ] 

Eric Yang commented on YARN-8734:
---------------------------------

Patch 001 is proof of concept to prevent service from deploying until 
remote_service_dependencies is satisfied.  User can express remote dependencies 
like:

{code}
{
  "name": "sleeper-service",
  "version": "5",
  "remote_service_dependencies": ["my-remote-service"],
  "components" :
  [
    {
      "name": "ping",
      "number_of_containers": 2,
      "artifact": {
        "id": "hadoop/centos:latest",
        "type": "DOCKER"
      },
      "launch_command": "sleep,10000",
      "resource": {
        "cpus": 1,
        "memory": "256"
      }
    }
  ]
}
{code}

The only limitation of remote service dependency is the dependent service must 
be owned by the same user.  YARN restricts who can see the status of other 
jobs.  This is the reason that user A can not obtain status of user B's 
service.  This limitation is likely to be resolved by YARN-8733 remote service 
port check.

> Readiness check for remote service
> ----------------------------------
>
>                 Key: YARN-8734
>                 URL: https://issues.apache.org/jira/browse/YARN-8734
>             Project: Hadoop YARN
>          Issue Type: New Feature
>          Components: yarn-native-services
>            Reporter: Eric Yang
>            Assignee: Eric Yang
>            Priority: Major
>         Attachments: Dependency check vs.pdf, YARN-8734.001.patch
>
>
> When a service is deploying, there can be remote service dependency.  It 
> would be nice to describe ZooKeeper as a dependent service, and the service 
> has reached a stable state, then deploy HBase.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org

Reply via email to