Vladimir Ozerov created IGNITE-4272:
---------------------------------------
Summary: Hadoop: shuffle messages must always be processed
asynchronously wtr to shuffle job or mapper.
Key: IGNITE-4272
URL: https://issues.apache.org/jira/browse/IGNITE-4272
Project: Ignite
Issue Type: Sub-task
Components: hadoop
Affects Versions: 1.8
Reporter: Vladimir Ozerov
Assignee: Vladimir Ozerov
Fix For: 2.0
Currently we run "GridIoManager.send()" from shuffle job. If it hit local node,
then message will be processed synchronously, stalling shuffling process for
other mappers.
Need to ensure that it is always routed to internal pool instead.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
