mesos-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Vinod Kone (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (MESOS-6299) Master doesn't remove task from pending when it is invalid
Date Fri, 30 Sep 2016 20:40:20 GMT

     [ https://issues.apache.org/jira/browse/MESOS-6299?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Vinod Kone updated MESOS-6299:
------------------------------
    Description: 
In `Master::_accept()` there are cases when a task is not launched (e.g, agent disconnected,
agent removed or task group is invalid). Doesn't look like we make sure the task is removed
from `framework->pending` in this case. In addition to the memory leak reconciliation will
result in a scheduler thinking the task is still in TASK_STAGING. Same for operators who are
looking at operator endpoints.

Note that pending task not being removed when agent is disconnected/removed is a long standing
bug. We should fix it in all the supported versions.




  was:In `Master::_accept()` there are cases when a task is not launched (e.g, agent disconnected
or task is invalid). Doesn't look like we make sure the task is removed from `framework->pending`
in this case.


> Master doesn't remove task from pending when it is invalid
> ----------------------------------------------------------
>
>                 Key: MESOS-6299
>                 URL: https://issues.apache.org/jira/browse/MESOS-6299
>             Project: Mesos
>          Issue Type: Improvement
>            Reporter: Vinod Kone
>            Assignee: Vinod Kone
>            Priority: Blocker
>              Labels: mesosphere
>             Fix For: 1.1.0
>
>
> In `Master::_accept()` there are cases when a task is not launched (e.g, agent disconnected,
agent removed or task group is invalid). Doesn't look like we make sure the task is removed
from `framework->pending` in this case. In addition to the memory leak reconciliation will
result in a scheduler thinking the task is still in TASK_STAGING. Same for operators who are
looking at operator endpoints.
> Note that pending task not being removed when agent is disconnected/removed is a long
standing bug. We should fix it in all the supported versions.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message