[
https://issues.apache.org/jira/browse/MESOS-4829?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Anand Mazumdar updated MESOS-4829:
----------------------------------
Description:
There are two ways in which a shutdown of executor can be triggered:
1. If it receives an explicit `Shutdown` message from the agent.
2. If the recovery timeout period has elapsed, and the executor still hasn’t
been able to (re-)connect with the agent.
Currently, the executor library relies on the field `grace_period_seconds`
having a default value of 5 seconds to handle the second scenario.
https://github.com/apache/mesos/blob/master/src/executor/executor.cpp#L608
The driver used to trigger the grace period via a constant defined in
src/slave/constants.cpp.
https://github.com/apache/mesos/blob/master/src/exec/exec.cpp#L92
The agent may want to force a shorter shutdown grace period (e.g.
oversubscription eviction may have shorter deadline) in the future. For now, we
can just read the value via an environment variable.
was:
There are two ways in which a shutdown of executor can be triggered:
1. If it receives an explicit `Shutdown` message from the agent.
2. If the recovery timeout period has elapsed, and the executor still hasn’t
been able to (re-)connect with the agent.
Currently, the executor library relies on the field `grace_period_seconds`
having a default value of 5 seconds to handle the second scenario.
https://github.com/apache/mesos/blob/master/src/executor/executor.cpp#L608
The driver used to trigger the grace period via a constant defined in
src/slave/constants.cpp.
https://github.com/apache/mesos/blob/master/src/exec/exec.cpp#L92
The agent may want to force a shorter shutdown grace period (e.g.
oversubscription eviction may have shorter deadline).
> Remove `grace_period_seconds` field from Shutdown event.
> --------------------------------------------------------
>
> Key: MESOS-4829
> URL: https://issues.apache.org/jira/browse/MESOS-4829
> Project: Mesos
> Issue Type: Task
> Reporter: Anand Mazumdar
> Labels: mesosphere
>
> There are two ways in which a shutdown of executor can be triggered:
> 1. If it receives an explicit `Shutdown` message from the agent.
> 2. If the recovery timeout period has elapsed, and the executor still hasn’t
> been able to (re-)connect with the agent.
> Currently, the executor library relies on the field `grace_period_seconds`
> having a default value of 5 seconds to handle the second scenario.
> https://github.com/apache/mesos/blob/master/src/executor/executor.cpp#L608
> The driver used to trigger the grace period via a constant defined in
> src/slave/constants.cpp.
> https://github.com/apache/mesos/blob/master/src/exec/exec.cpp#L92
> The agent may want to force a shorter shutdown grace period (e.g.
> oversubscription eviction may have shorter deadline) in the future. For now,
> we can just read the value via an environment variable.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
