[
https://issues.apache.org/jira/browse/MESOS-6763?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15743488#comment-15743488
]
Vinod Kone commented on MESOS-6763:
-----------------------------------
Keeping the ticket open for the tests review to land.
commit 441e0767f60a14496de67cf17078e912faf2daeb
Author: Kevin Klues <[email protected]>
Date: Mon Dec 12 15:20:04 2016 -0800
Added a bi-directional heartbeat for IOSwitchboard connections.
Review: https://reviews.apache.org/r/54560/
commit c1f9af5b53f370f9720654b360c373a5cb1ad7fe
Author: Kevin Klues <[email protected]>
Date: Mon Dec 12 15:19:59 2016 -0800
Added a `http_heartbeat_interval` flag to the agent flags.
Review: https://reviews.apache.org/r/54630/
> Add heartbeats to both input/output connections in IOSwitchboard
> ----------------------------------------------------------------
>
> Key: MESOS-6763
> URL: https://issues.apache.org/jira/browse/MESOS-6763
> Project: Mesos
> Issue Type: Improvement
> Reporter: Kevin Klues
> Assignee: Kevin Klues
> Labels: debugging, mesosphere
>
> Some networks will kill idle connections if no data is transfered over them
> within a set amount of time. For example, using AWS's Elastic Load Balancer
> (ELB), the default time to kill a connection is only 60s! Because of this, we
> need a way to send application level heartbeats to keep these connections
> alive for any long running LAUNCH_NESTED_CONTAINER_SESSION,
> ATTACH_CONTAINER_INPUT and ATTACH_CONTAINER_OUTPUT calls.
> We should serve these heartbeats from the IOSwitchboard server rather than
> the agent handlers since the agent essentially acts as a proxy and the
> heartbeats should originate from the actual server being communicated with.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
