We are running embedded Janusgraph (0.5.3) with an HBase backend (2.1.6) in our Spark jobs. Each Spark executor creates an instance of Janusgraph. At times there can be over 500 executors running simultaneously. Under those conditions, we observe heavy row contention for the ‘configuration’ row that Janusgraph creates as part of the initialization of the HBase table. Is there any recommendation on how to prevent/reduce this HBase row contention? As the row is only created during HBase initialization and is never updated subsequently, can the data held by the configuration row be moved out of HBase and into a static file?
This e-mail may contain information that is privileged or confidential. If you are not the intended recipient, please delete the e-mail and any attachments and notify us immediately.