[jira] [Commented] (MESOS-5361) Consider introducing TCP KeepAlive for Libprocess sockets.

["haosdent (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/MESOS-5361?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15283537#comment-15283537 ] 

haosdent commented on MESOS-5361:
---------------------------------

{quote}
 Also, operators might start tuning the values for their clusters explicitly once we start supporting it.
{quote}

We could reduce the default {{TCP_KEEPIDLE}}, {{TCP_KEEPINTVL}} and {{TCP_KEEPCNT}} when create the socket instead of operators to change the Linux default configuration. 

{quote}
The default TCP keep alive values on Linux are a joke though
{quote}

Actually it is not a joke, the RFC 1122 standard defined the minimum timeout should not less than 2 hours while we use few seconds instead in the real world...

{code}
4.2.3.6  TCP Keep-Alives

            Implementors MAY include "keep-alives" in their TCP
            implementations, although this practice is not universally
            accepted.  If keep-alives are included, the application MUST
            be able to turn them on or off for each TCP connection, and
            they MUST default to off.

            Keep-alive packets MUST only be sent when no data or
            acknowledgement packets have been received for the
            connection within an interval.  This interval MUST be
            configurable and MUST default to no less than two hours.
{code}

> Consider introducing TCP KeepAlive for Libprocess sockets.
> ----------------------------------------------------------
>
>                 Key: MESOS-5361
>                 URL: https://issues.apache.org/jira/browse/MESOS-5361
>             Project: Mesos
>          Issue Type: Improvement
>          Components: libprocess
>            Reporter: Anand Mazumdar
>              Labels: mesosphere
>
> We currently don't use TCP KeepAlive's when creating sockets in libprocess. This might benefit master - scheduler, master - agent connections i.e. we can detect if any of them failed faster.
> Currently, if the master process goes down. If for some reason the {{RST}} sequence did not reach the scheduler, the scheduler can only come to know about the disconnection when it tries to do a {{send}} itself. 
> The default TCP keep alive values on Linux are a joke though:
> {code}
> . This means that the keepalive routines wait for two hours (7200 secs) before sending the first keepalive probe, and then resend it every 75 seconds. If no ACK response is received for nine consecutive times, the connection is marked as broken.
> {code}
> However, for long running instances of scheduler/agent this still can be beneficial. Also, operators might start tuning the values for their clusters explicitly once we start supporting it.



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