[YARN-4544] All the log messages about rolling monitoring interval are shown with WARN level

• Type: Improvement
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 2.7.1
• Fix Version/s: 2.8.0, 3.0.0-alpha1
• Component/s: log-aggregation, nodemanager
• Hadoop Flags: Reviewed
• Assignee: Takashi Ohnishi
• Reporter: Takashi Ohnishi
• Created: 05/Jan/16 18:46
• Updated: 30/Aug/16 04:08
• Resolved: 07/Jan/16 19:16

Description

About yarn.nodemanager.log-aggregation.roll-monitoring-interval-seconds, there are three log messages corresponding to the value set to this parameter, but all of them are shown with the WARN level.

(a) disabled (default)

2016-01-05 22:19:29,062 WARN org.apache.hadoop.yarn.server.nodemanager.containermanager.logaggreg     ation.AppLogAggregatorImpl: rollingMonitorInterval is set as -1. The log rolling mornitoring interval is disabled. The logs will be aggregated after this application is finished.

(b) enabled

2016-01-06 00:41:15,808 WARN org.apache.hadoop.yarn.server.nodemanager.containermanager.logaggregation.AppLogAggregatorImpl: rollingMonitorInterval is set as 7200. The logs will be aggregated every 7200 seconds

(c) enabled but wrong configuration

2016-01-06 00:39:50,820 WARN org.apache.hadoop.yarn.server.nodemanager.containermanager.logaggregation.AppLogAggregatorImpl: rollingMonitorIntervall should be more than or equal to 3600 seconds. Using 3600 seconds instead.

I think it is better to output with WARN only in case (c), but it is ok to output with INFO in case (a) and (b).