reef-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Andrew Chung (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (REEF-1364) C# Evaluator should attempt to send a failure message back to the Driver on an unhandled Exception
Date Thu, 28 Apr 2016 23:08:12 GMT

     [ https://issues.apache.org/jira/browse/REEF-1364?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Andrew Chung updated REEF-1364:
-------------------------------
    Description: Currently, an Unhandled Exception at the {{System.Threading.Tasks.Task}}
Continuation of a user's Task simply prompts the Evaluator to call {{Environment.Exit(1)}}.
We should fix it such that the Evaluator attempts to send a message back to the Driver before
giving up.  (was: The C# Evaluator is structured as follows:
1. Main {{System.Threading.Tasks.Task}} runs a clock which triggers heartbeats periodically.
2. A {{System.Threading.Tasks.Task}} is fired and forgotten for the user's Task. It updates
the status by setting a shared variable between the two {{System.Threading.Tasks.Task}}.

What should be done is the follows:
{{await}} two {{System.Threading.Tasks.Task}} using {{Task.WaitAny}} , one for the {{ContextManager}}
which handles user's Tasks, the other for heartbeats . If either of those Tasks leaks an {{Exception}},
that means our {{Exception}} handling was not done properly in the {{System.Threading.Tasks.Task}}
that threw the {{Exception}}, in which case we should fail the Evaluator.

The {{ContextManager}} {{System.Threading.Tasks.Task}} should contain the logic for performing
Context and Task-related heartbeats.
The Evaluator {{System.Threading.Tasks.Task}} should contain the logic for performing periodic
heartbeats that notify the Driver that the Evaluator is still running.

This will immensely simplify the Exception handling logic and provide a clearer structure
to the C# Evaluator.)

> C# Evaluator should attempt to send a failure message back to the Driver on an unhandled
Exception
> --------------------------------------------------------------------------------------------------
>
>                 Key: REEF-1364
>                 URL: https://issues.apache.org/jira/browse/REEF-1364
>             Project: REEF
>          Issue Type: Improvement
>          Components: REEF.NET, REEF.NET Evaluator
>            Reporter: Andrew Chung
>            Assignee: Andrew Chung
>
> Currently, an Unhandled Exception at the {{System.Threading.Tasks.Task}} Continuation
of a user's Task simply prompts the Evaluator to call {{Environment.Exit(1)}}. We should fix
it such that the Evaluator attempts to send a message back to the Driver before giving up.



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

Mime
View raw message