[YARN-8691] AMRMClient unregisterApplicationMaster Api's appMessage should have a maximum size

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Duplicate
• Affects Version/s: 2.7.3
• Fix Version/s: 3.0.0-alpha4
• Component/s: resourcemanager
• Assignee: Yicong Cai
• Reporter: Yicong Cai
• Created: 21/Aug/18 06:47
• Updated: 23/Aug/18 08:45
• Resolved: 23/Aug/18 08:38

Description

SparkSQL AM Codegen ERROR，then call unregister AM API and send the error message to RM, RM receive the AM state and update to RMStateStore. The  Codegen error message maybe is huge, (Our case is about 200MB). If the RMStateStore is ZKRMStateStore, it causes the same exception as YARN-6125, but YARN-6125 doesn't cover the unregisterApplicationMaster's message cut.

 

SparkSQL Codegen error message show below:

18/08/18 08:34:54 ERROR codegen.CodeGenerator: failed to compile: org.codehaus.janino.JaninoRuntimeException: Constant pool has grown past JVM limit of 0xFFFF
/* 001 */ public java.lang.Object generate(Object[] references)

{ /* 002 */ return new SpecificSafeProjection(references); /* 003 */ }

/* 004 */
/* 005 */ class SpecificSafeProjection extends org.apache.spark.sql.catalyst.expressions.codegen.BaseProjection {
..........

about 2 million lines.

..........

Issue Links

• duplicates: YARN-6967 Limit application attempt's diagnostic message size thoroughly
• is related to: SPARK-25174 ApplicationMaster suspends when unregistering itself from RM with extreme large diagnostic message
• relates to: YARN-7115 Move BoundedAppender to org.hadoop.yarn.util pacakge
• relates to: YARN-6125 The application attempt's diagnostic message should have a maximum size