Mailing-List: contact dev-help@flink.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@flink.apache.org
Date: Thu, 12 Jan 2017 15:09:54 +0000 (UTC)
From: "Till Rohrmann (JIRA)" <jira@apache.org>
To: dev@flink.apache.org
Message-ID: <JIRA.13034383.1484233786000.2615.1484233794488@Atlassian.JIRA>
In-Reply-To: <JIRA.13034383.1484233786000@Atlassian.JIRA>
References: <JIRA.13034383.1484233786000@Atlassian.JIRA> <JIRA.13034383.1484233786267@arcas>
Subject: [jira] [Created] (FLINK-5471) Properly inform JobClientActor about
 terminated Mesos framework
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit
archived-at: Thu, 12 Jan 2017 15:10:02 -0000

Till Rohrmann created FLINK-5471:
------------------------------------

             Summary: Properly inform JobClientActor about terminated Mesos framework
                 Key: FLINK-5471
                 URL: https://issues.apache.org/jira/browse/FLINK-5471
             Project: Flink
          Issue Type: Improvement
          Components: Mesos
    Affects Versions: 1.2.0, 1.3.0
            Reporter: Till Rohrmann
            Priority: Minor
             Fix For: 1.3.0


In case that the Mesos framework running Flink terminates (e.g. exceeded number of container restarts) the {{JobClientActor}} is not properly informed. As a consequence, the client only terminates after the {{JobClientActor}} detects that it lost the connection to the JobManager ({{JobClientActorConnectionTimeoutException}}). The current default value for the timeout is 60s which is quite long to detect the connection loss in case of a termination.

I think it would be better to notify the {{JobClientActor}} which allows it to print a better message for the user and also allows it to react quicker.


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