Hi Marc, 

from the gremlin console, I get: 


While getting in the server log, on the one side a lot of warnings such as: 


And finally an exception about a temporary failure in storage backend: 


... 
caused by: 



I'll keep the logs this tiem just in case more detail is required. 

Best, 

Carlos 

Hi Carlos,

I read the preceding discussion with Stephen Mallette, which says: From the logs, while loading this graph, Cassandra driver is almost always warning that all batches are over the limit of 5120 (which I haven't found yet where to modify ...).
A complete stacktrace would help indeed, but it strikes me that 5120 equals 20 x 256 while the cassandra cql drivers have the following defaults:

storage.cql.batch-statement-size

The number of statements in each batch

Integer

20

storage.cql.remote-max-requests-per-connection

The maximum number of requests per connection for remote datacenter

Integer

256

Best wishes,    Marc

Hi all,

I'm trying to load a graphml export into janusgraph 0.5.3. not quite big (1.68M nodes, 8.8M edges). However, I reach a point where Tinkerpop layer tells me that the Batch is too large and it crashes (I suspect that there might be an edge with far too much information ... but finding it is difficult).

I've tried to split the graphml file into non-overlapping partitions, but janusgraph does not seem to honor the ID (I'm not actually sure whether this might be omitted at Tinkerpop or Janusgraph level), and when I reach the splitted edges part, it inserts new nodes for both source and target.

Has anyone faced this graphml loading problem before? Should I try to get a GraphSON translated version of my exported graph to run everything more smoothly?

What are the recommendations to deal with this kind of loadings? I'm trying to avoid implementing my own graph loader at this point.

Thank you very much in advance,

Best,

Carlos Bobed