Presto 0.246 returns a division by zero stacktrace

Hi everybody,

I have upgraded Prestodb from 0.226 to 0.246, and adjusted some config parameters to adapt to the new version. My testing environment is very simple, one query coordinator and one worker node. When I try to execute a query that uses the Hive catalog, I always see the following returned:

java.lang.ArithmeticException: / by zero
	at com.facebook.presto.sql.planner.NodePartitioningManager.createArbitraryBucketToNode(NodePartitioningManager.java:256)
	at com.facebook.presto.sql.planner.NodePartitioningManager.getBucketNodeMap(NodePartitioningManager.java:189)
	at com.facebook.presto.execution.scheduler.SectionExecutionFactory.createStageScheduler(SectionExecutionFactory.java:335)
	at com.facebook.presto.execution.scheduler.SectionExecutionFactory.createStreamingLinkedStageExecutions(SectionExecutionFactory.java:243)
	at com.facebook.presto.execution.scheduler.SectionExecutionFactory.createStreamingLinkedStageExecutions(SectionExecutionFactory.java:221)
	at com.facebook.presto.execution.scheduler.SectionExecutionFactory.createSectionExecutions(SectionExecutionFactory.java:167)
	at com.facebook.presto.execution.scheduler.LegacySqlQueryScheduler.createStageExecutions(LegacySqlQueryScheduler.java:343)
	at com.facebook.presto.execution.scheduler.LegacySqlQueryScheduler.<init>(LegacySqlQueryScheduler.java:233)
	at com.facebook.presto.execution.scheduler.LegacySqlQueryScheduler.createSqlQueryScheduler(LegacySqlQueryScheduler.java:164)
	at com.facebook.presto.execution.SqlQueryExecution.planDistribution(SqlQueryExecution.java:474)
	at com.facebook.presto.execution.SqlQueryExecution.start(SqlQueryExecution.java:347)
	at com.facebook.presto.$gen.Presto_0_246_b4fb470____20210205_091342_1.run(Unknown Source)
	at com.facebook.presto.execution.SqlQueryManager.createQuery(SqlQueryManager.java:254)
	at com.facebook.presto.dispatcher.LocalDispatchQuery.lambda$startExecution$5(LocalDispatchQuery.java:114)
	at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
	at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
	at java.lang.Thread.run(Thread.java:748)

The code that yields the exception seems to be doing a division indeed, but I have no idea why by zero since the worker node is correctly recognized by the coordinator (I checked in the IU as well). Can you give me more info about why this error happens?

[elukey]

@rschlussel the catalog config file properties had a typo in its filename, you were absolutely right, everything works now, thanks a lot for the help 😃

It was very difficult to figure out what was the problem though, without you I wouldn’t have had a chance to figure it out. The java division by zero stacktrace is definitely weird to see (and makes you believe that a bug is happening in the code), so if it is easy to add a friendlier message it would really be nice!

[rschlussel]

I suspect what’s happening is that the worker doesn’t know about the hive catalog. Can you double check that the file is there and in the right directory and try restarting the worker (you always need to restart after config changes, since Presto won’t pick them up automatically)? Another thing to try would be to try adding another connector (e.g. add the tpch connector https://prestodb.io/docs/current/connector/tpch.html) and see if you can query that.