[MAPREDUCE-7128] LocalJobRunner may fail due to metrics name collision

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Assignee: Unassigned
• Reporter: Todd Lipcon
• Created: 11/Aug/18 00:37
• Updated: 11/Aug/18 00:37

Description

Running Hive in local mode, I got the following failure:

	Caused by: org.apache.hadoop.metrics2.MetricsException: Metrics source LocalJobRunnerMetrics-1090927800 already exists!
		at org.apache.hadoop.metrics2.lib.DefaultMetricsSystem.newSourceName(DefaultMetricsSystem.java:152)
		at org.apache.hadoop.metrics2.lib.DefaultMetricsSystem.sourceName(DefaultMetricsSystem.java:125)
		at org.apache.hadoop.metrics2.impl.MetricsSystemImpl.register(MetricsSystemImpl.java:229)
		at org.apache.hadoop.mapred.LocalJobRunnerMetrics.create(LocalJobRunnerMetrics.java:46)
...

It seems that it tries to generate a unique metrics source name using a random 32-bit integer. When running some longer-lived program that runs many separate jobs (such as HS2) this can easily collide. https://en.wikipedia.org/wiki/Birthday_attack has some tables for the probabilities.