[ 
https://issues.apache.org/jira/browse/SAMZA-2402?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Sanil Jain updated SAMZA-2402:
------------------------------
    Description: 
On job restarts, there might be already pending container placement requests 
from the previous deployment. Since a new deploy can potentially change the job 
state (some containers might lose affinity), we should not allow pending 
control actions to be taken from the previous deployment

Also even if the metastore is eventually consistent, duplicate actions 
delivered to the metastore should be only taken once, allow queuing / dequeuing 
logic for container placement requests

  was:
On job restarts, there might be already pending container placement requests 
from the previous deployment. Since a new deploy can potentially change the job 
state (some containers might lose affinity), we should not allow pending 
control actions to be taken from the previous deployment

Also even if the metastore is eventually consistent, duplicate actions 
delivered to the metastore should be only taken once


> Tie Container placement service and Container placement handler and validate 
> placement requests
> -----------------------------------------------------------------------------------------------
>
>                 Key: SAMZA-2402
>                 URL: https://issues.apache.org/jira/browse/SAMZA-2402
>             Project: Samza
>          Issue Type: New Feature
>            Reporter: Sanil Jain
>            Assignee: Sanil Jain
>            Priority: Major
>          Time Spent: 10m
>  Remaining Estimate: 0h
>
> On job restarts, there might be already pending container placement requests 
> from the previous deployment. Since a new deploy can potentially change the 
> job state (some containers might lose affinity), we should not allow pending 
> control actions to be taken from the previous deployment
> Also even if the metastore is eventually consistent, duplicate actions 
> delivered to the metastore should be only taken once, allow queuing / 
> dequeuing logic for container placement requests



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

Reply via email to