I tested flink-0.7.0-incubating-bin-hadoop2-yarn.tgz.
It works fine for Apache MRQL in Flink mode:
1) It passed all query tests for MRQL in Flink local mode (25 queries).
2) It was tested successfully for PageRank on an 1.4GB XML file (DBLP
bibliography) using 20 slots on a 40-slot yarn cluster.
I had to make the following changes to the MRQL code:
1) changed some Java imports (eg, from
org.apache.flink.api.java.functions.RichMapFunction to
org.apache.flink.api.common.functions.RichMapFunction )
2) changed the run script to get the flink job manager from
conf/.yarn-properties instead of conf/.yarn-jobmanager
Good job!
Leonidas Fegaras
On 10/08/2014 06:15 AM, Robert Metzger wrote:
Hi guys,
I've created a first release candidate for the 0.7-incubating major release:
http://people.apache.org/~rmetzger/flink-0.7.0-incubating-rc0/
This thread is not an official release vote thread, rather a discussion
thread to collect all open issues that we want to address until we create
the voting release candidate.
Best,
Robert
