How to decouple multiple graph instances from each other in JanusGraph

user_2205969 0
Viewed 969

I'm trying to create multiple graph instances in Janusgraph, but they all seem to have the same reference to one another, and so any action taken on one affects the others (see example below). I want to have these graphs setup as separate instances, distinct from each other, but somewhere in the steps below I'm faltering.

Steps taken to add a new graph to JanusGraph

Goal: have two graphs called graph1, graph2, with traversal objects named g1, g2, respectively, and which are distinct from one another.

  1. Create properties files called graph1.properties, graph2.properties. With contents (for a Cassandra backend):

    gremlin.graph=org.janusgraph.core.JanusGraphFactory
gremlin.graph=org.janusgraph.core.ConfiguredGraphFactory
storage.backend=cql
storage.hostname=127.0.0.1

^ This is where I'm guessing the core issue lies - graph1.properties and graph2.properties have the same contents... but I'm unsure what to change

  1. Add the graphs to the gremlin-server.yaml file, which maps to the newly created graph1.properties and graph2.properties files.

    graphs: {
    graph1: conf/gremlin-server/graph1.properties,
    graph2: conf/gremlin-server/graph2.properties
}

  2. Add traversal object names to the empty-sample.groovy

    globals << [g1 : graph1.traversal(), g2: graph2.traversal()]

Testing

The output below shows that the graphs were created successfully, but also shows that they are referencing eachother.

==>Configured localhost/127.0.0.1:8182-[b7696535-97d9-4b59-b30f-f83707492057]
gremlin> :remote console
==>All scripts will now be sent to Gremlin Server - [localhost/127.0.0.1:8182]-[b7696535-97d9-4b59-b30f-f83707492057] - type ':remote console' to return to local mode
gremlin> g1
==>graphtraversalsource[standardjanusgraph[cql:[127.0.0.1]], standard]
gremlin> g1.V().count()
==>100
gremlin> g2.V().count()
==>100
gremlin> g1.addV('item').property('id', '123')
==>v[327684312]
gremlin> g1.tx().commit()
==>null
gremlin> g1.V().count()
==>101
gremlin> g2.V().count()
==>101                  <-- g2 should have remained at 100
janusgraph gremlin-server tinkerpop3 gremlin graph
1 Answers

Since you're using Cassandra to store your data, to separate two graphs from each other entirely, you need to use a different Cassandra keyspace for each graph, as that's the unit of storage that JanusGraph uses.

As you can see in the JanusGraph Cassandra docs:

  • keyspace: The name of the keyspace to store the JanusGraph graph in. Allows multiple JanusGraph graphs to co-exist in the same Cassandra cluster.

Looking at the JanusGraph configuration reference, we see more information about this configuration parameter:

  • name: storage.cql.keyspace
  • description: "The name of JanusGraph’s keyspace. It will be created if it does not exist."
  • data type: String
  • default value: janusgraph
  • mutability: LOCAL

So, by not specifying this parameter in your config, both graphs are being stored in the default janusgraph keyspace which causes them to conflict with each other.

Since you're using storage.backend=cql, to separate the two graphs, simply provide an explicit parameter storage.cql.keyspace which is different for each of the graphs you want to separate from each other.

user_36186710
Related