Titan 1.1 & Tinkerpop 3.1.0 Graph computer troubles

[Damian Wloch]

Hi guys,

I've tried running a simple query on a graph with a traversal using a computer of org.apache.tinkerpop.gremlin.hadoop.process.computer.spark.SparkGraphComputer with local Spark and Cassandra but it just sat the hung and doing seemingly nothing. The steps I took were:

sparkgraph = GraphFactory.open('conf/hadoop-graph/read-cassandra.properties')

gs = sparkgraph.traversal(computer(org.apache.tinkerpop.gremlin.hadoop.process.computer.spark.SparkGraphComputer))

gs.V().count()

In the attachments I've included the config I'm using. Spark is 1.5.1 pulled from the Apache website, Cassandra 2.1.8 from DSE.

I'd expect it to give me a nice 0 for the result, but I get nothing.

[Jason Plurad]

Hi Damian,

Are you running this from the Gremlin Console or from a standalone Java program?

-- Jason

--
You received this message because you are subscribed to the Google Groups "Aurelius" group.
To unsubscribe from this group and stop receiving emails from it, send an email to aureliusgraph...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/aureliusgraphs/b65dd2e3-54b3-41f7-88f3-c828ee8609c1%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Damian Wloch]

From an external application connecting through websockets and from the CLI (though using websockets again). I don't currently have a setup to do this using the gremlin console. I don't see any errors in the server output, but I'll try again using a proper console session when I can.

[Damian Wloch]

I've included as well the log with an actual error when running from the Gremlin CLI.

[Damian Wloch]

I've managed to gleam this might be due library versioning mismatches. Is there a way of getting a full distribution of Titan 1.1 and Tinkerpop >3.1.0 (even if they're pre-release) with all the required libs for connecting to Cassandra and using the SparkGraphComputer? So far I haven't had luck trying to compile both from source.

On Tuesday, 1 March 2016 17:39:15 UTC, Damian Wloch wrote:

[Jason Plurad]

Building from source has come up previously -- https://groups.google.com/d/msg/aureliusgraphs/e5L5M6MQozY/GoywC3prBAAJ
Later on in that thread John Skilling also posted a Dropbox link to a compiled distrubition.

git clone https://github.com/thinkaurelius/titan -b titan11 titan11
cd titan11
mvn clean install -DskipTests=true -Dgpg.skip=true -Paurelius-release
unzip titan-dist/titan-dist-hadoop-2/target/titan-1.1.0-SNAPSHOT-hadoop2.zip

The same mvn command works against the https://github.com/graben1437/titan1withtp3.1 repo.

I'd recommend using the gremlin.sh and gremlin-server.sh packaged in the titan.zip only because the titan.zip includes the spark-gremlin.jar in the lib directory. If you used the standalone apache-gremlin-console.zip, you'd have to install the hadoop-gremlin and spark-gremlin plugins, and since you're building from source, you'd have to make sure that your grapeConfig.xml was set up to pull from your local .m2 repository.

As you work with TinkerPop OLAP, make sure that your HADOOP_GREMLIN_LIBS is set correctly, and that you delete any previous versions of cached jars laying around in HDFS or on your Spark cluster classpath.

-- Jason

[Damian Wloch]

That fixed all my problems. Thank you very much.