I execute the count method on a cluster with a titan graph using the fulgora graph computer and get this common error:
Frame size (20769719) larger than max length (15728640)!
This can be corrected by specifying this in the config file:
storage.cassandra.frame-size-mb=60
However, when using the SparkGraphComputer on a hadoop graph pointing to the same data I get the same error. I have tried all of these settings and none of them work:
titanmr.ioformat.conf.storage.cassandra.frame-size-mb=60
titanmr.ioformat.conf.storage.cassandra.frame-size=60
titan.hadoop.input.conf.storage.cassandra.thrift.frame-size=60
titan.hadoop.input.conf.storage.cassandra.thrift.frame-size-mb=60
storage.cassandra.thrift.frame-size=60
storage.cassandra.thrift.frame-size-mb=60
Does anyone know the current correct setting?
The Cassandra read path is different when you're doing OLAP, so that's why using the Titan setting storage.cassandra.frame-size-mb doesn't work. Try this Cassandra setting: cassandra.thrift.framed.size_mb. It defaults to 15 MB.
References: