[PIG-5309] Problem with tez + union + replicated join

• Type: Bug
• Status: Open
• Priority: Minor
• Resolution: Unresolved
• Affects Version/s: 0.17.0
• Fix Version/s: 0.18.0, 0.17.1
• Component/s: tez
• Assignee: Rohini Palaniswamy
• Reporter: Will Oberman
• Created: 12/Oct/17 14:30
• Updated: 13/Oct/17 22:46

Description

I've been using Pig 0.12.1 for quite some time and am finally upgrading to 0.17. One of my existing scripts failed. I have a workaround (SET pig.tez.opt.union false), but I thought I'd pass on the problem I observed.

In stdout:

ERROR 2017: Internal error creating job configuration.

In the Pig log:

Caused by: java.lang.IllegalArgumentException: Edge [scope-93 : org.apache.pig.backend.hadoop.executionengine.tez.runtime.PigProcessor] -> [scope-83 : org.apache.pig.backend.hadoop.executionengine.tez.runtime.PigProcessor] ({ BROADCAST : org.apache.tez.runtime.library.input.UnorderedKVInput >> PERSISTED >> org.apache.tez.runtime.library.output.UnorderedKVOutput >> NullEdgeManager }) already defined!
	at org.apache.tez.dag.api.DAG.addEdge(DAG.java:272)
	at org.apache.pig.backend.hadoop.executionengine.tez.TezDagBuilder.visitTezOp(TezDagBuilder.java:404)
	at org.apache.pig.backend.hadoop.executionengine.tez.plan.TezOperator.visit(TezOperator.java:259)
	at org.apache.pig.backend.hadoop.executionengine.tez.plan.TezOperator.visit(TezOperator.java:56)
	at org.apache.pig.impl.plan.DependencyOrderWalker.walk(DependencyOrderWalker.java:87)
	at org.apache.pig.impl.plan.PlanVisitor.visit(PlanVisitor.java:46)
	at org.apache.pig.backend.hadoop.executionengine.tez.TezJobCompiler.buildDAG(TezJobCompiler.java:69)
	at org.apache.pig.backend.hadoop.executionengine.tez.TezJobCompiler.getJob(TezJobCompiler.java:120)
	... 20 more

I played around with a minimum viable test script and can cause this to fail:

weblogs = LOAD '/tmp/in/weblogInfo' as (path:chararray, queryMap:map[chararray]); 
featureToExtraData = LOAD '/tmp/in/featureToExtraData' as (feature:chararray, extraData:chararray); 

oldA = FILTER weblogs BY path == '/A';
newA = FILTER weblogs BY path == '/somethingElse';
B = FILTER weblogs BY path == '/B';

oldAFeatures = FOREACH oldA GENERATE queryMap#'feature1' as feature1, queryMap#'feature2' as feature2;
newAFeatures = FOREACH newA GENERATE queryMap#'different1' as feature1, queryMap#'different2' as feature2;
AFeatures = UNION oldAFeatures, newAFeatures;
AFeaturesPlusMore = JOIN AFeatures BY feature1 LEFT, featureToExtraData BY feature USING 'replicated';

BFeatures = FOREACH B GENERATE queryMap#'somethingElseEntirely1' as feature1, queryMap#'somethingElseEntirely2' as feature2;
BFeaturesPlusMore = JOIN BFeatures BY feature1 LEFT, featureToExtraData BY feature USING 'replicated';

STORE AFeaturesPlusMore INTO '/tmp/out/1/AFeaturesPlusMore';
STORE BFeaturesPlusMore INTO '/tmp/out/1/BFeaturesPlusMore';