We used streamset to put data in KUDU cluster and get the following error: com.streamsets.pipeline.api.StageException: KUDU_03 - Errors while writing to Kudu: org.kududb.client.NonRecoverableException: RPC can not complete before timeout: KuduRpc(method=Write, tablet=0cf1a047fbf845d2833806062c379984, attempt=16, DeadlineTracker(timeout=10000, elapsed=9799)) I think it could be much connections to KUDU cluster. How could I improve KUDU performance? And how could I modify the RPC timeout? We used CDH5.8.3+KUDU1.0.1. Thanks a lot. Vince Ng
I don't know how StreamSets integrates with Kudu, but it looks like they're using the Kudu Java client. If you have the capability to modify how the Java client is used, you need to call KuduSession.setTimeoutMillis() on the session objects responsible for writing. The default value is 10000; you can set it much higher (like 60000).
As for generally improving Kudu performance, your best bet is to hop on our Slack channel (https://getkudu-slack.herokuapp.com/) and talk to the developers directly; it'd be far faster than a back-and-forth on this message board.