0

I understood that giraph-dist-1.2.0-hadoop2-bin.tar.gz binary distribution is built with following maven command and it is officially supported by with hadoop-2.5.1.

"mvn -Phadoop_2 clean install"

I successfully used giraph-dist-1.2.0-hadoop2-bin.tar.gz in pseudo distributed mode on hadoop-2.5.1 in which I configured yarn.

Now, I downloaded giraph-dist-1.2.0-hadoop2-src.tar.gz and successfully build giraph with yarn support, using command and patches taken from Building Giraph with Hadoop, i.e.:

"mvn -Phadoop_yarn -Dhadoop.version=2.5.1 clean package -DskipTests"

Since I already configured yarn in Hadoop 2.5.1, I didn't understand if and what I have to change in Hadoop 2.5.1 configuration about mapred-site.xml and yarn-site.xml in order to use giraph with yarn support?

I think that the main question is: What does it change in Hadoop usage by giraph built with -Phadoop_2 and by giraph built with -Phadoop_yarn?

The only documentation that I found is the following one:

  • Apache Hadoop 2 (latest version: 2.5.1)

    This is the latest version of Hadoop 2 (supporting YARN in addition to MapReduce) Giraph could use. You may tell maven to use this version with "mvn -Phadoop_2 ".

Cristina Bovi
  • 11
  • 4
  • You can try looking at the POM file profiles for each of them. As far as I know, Giraph isn't maintained anymore and JanusGraph is the more current version of a Hadoop graph system – OneCricketeer Aug 29 '18 at 01:20
  • Probably: - with "mvn -Dhadoop2" giraph runs as mapreduce application - with "mvn -Dyarn" giraph runs as yarn application – Cristina Bovi Aug 30 '18 at 13:25
  • A "yarn application" isn't a definitive thing, though. MR2 is only one type of YARN application – OneCricketeer Aug 30 '18 at 13:53

0 Answers0