You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@reef.apache.org by "Markus Weimer (JIRA)" <ji...@apache.org> on 2017/11/06 17:57:00 UTC
[jira] [Created] (REEF-1951) Make DriverStatus more robust
Markus Weimer created REEF-1951:
-----------------------------------
Summary: Make DriverStatus more robust
Key: REEF-1951
URL: https://issues.apache.org/jira/browse/REEF-1951
Project: REEF
Issue Type: Bug
Components: REEF Bridge
Reporter: Markus Weimer
[REEF-1942] introduces a new HTTP handler to communicate the Driver status to the client. However, as the tests for [REEF-1950] in [#1412|https://github.com/apache/reef/pull/1412] show, this mechanism isn't robust in two ways:
# The Client could call the HTTP endpoint for the first time _before_ it starts. This results in a failed connection. To remedy this, the Client needs to retry the connection.
# The Client could call the HTTP endpoint for the first time _after_ the Driver shuts down. This results in a failure. To remedy this, the Driver's HTTP handler should ensure that it is called at least once, e.g. by scheduling checks on the {{Clock}}.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)