[
https://issues.apache.org/jira/browse/MESOS-868?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Yan Xu updated MESOS-868:
-------------------------
Description:
Currently we use a constant of 10 seconds as both the zookeeper session timeout
and the local timeout. However, the timeout proposed by the client maybe
rejected by ZK if it violates the server's (max)minSessionTimeout.
We should set local timer to use negotiated timeout.
This is to ensure that if the leading master is partitioned, it terminates asap
when it's session is destined to expire.
was:
Currently we use a constant of 10 seconds as both the zookeeper session timeout
and the local timeout. However, the timeout proposed by the client maybe
rejected by ZK if it violates the server's (max)minSessionTimeout.
We should read the negotiated timeout and set local timer to use a value
smaller than that. e.g. half of it.
This is to ensure that if the leading master is partitioned, it terminates asap
when it's session is destined to expire.
> Group should set local session timeout to use half of negotiated ZK timeout
> ---------------------------------------------------------------------------
>
> Key: MESOS-868
> URL: https://issues.apache.org/jira/browse/MESOS-868
> Project: Mesos
> Issue Type: Bug
> Components: general
> Reporter: Yan Xu
> Assignee: Yan Xu
>
> Currently we use a constant of 10 seconds as both the zookeeper session
> timeout and the local timeout. However, the timeout proposed by the client
> maybe rejected by ZK if it violates the server's (max)minSessionTimeout.
> We should set local timer to use negotiated timeout.
> This is to ensure that if the leading master is partitioned, it terminates
> asap when it's session is destined to expire.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
