[PHOENIX-4692] ArrayIndexOutOfBoundsException in ScanRanges.intersectScan

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 4.14.0
• Fix Version/s: 4.14.0, 5.0.0
• Assignee: James R. Taylor
• Reporter: Sergey Soldatov
• Created: 18/Apr/18 07:12
• Updated: 26/Jul/18 01:16
• Resolved: 24/May/18 21:48

Description

ScanRanges.intersectScan may fail with AIOOBE if a salted table is used.

java.lang.ArrayIndexOutOfBoundsException: 1

	at org.apache.phoenix.util.ScanUtil.getKey(ScanUtil.java:333)
	at org.apache.phoenix.util.ScanUtil.getMinKey(ScanUtil.java:317)
	at org.apache.phoenix.compile.ScanRanges.intersectScan(ScanRanges.java:371)
	at org.apache.phoenix.iterate.BaseResultIterators.getParallelScans(BaseResultIterators.java:1074)
	at org.apache.phoenix.iterate.BaseResultIterators.getParallelScans(BaseResultIterators.java:631)
	at org.apache.phoenix.iterate.BaseResultIterators.<init>(BaseResultIterators.java:501)
	at org.apache.phoenix.iterate.ParallelIterators.<init>(ParallelIterators.java:62)
	at org.apache.phoenix.execute.ScanPlan.newIterator(ScanPlan.java:274)
	at org.apache.phoenix.execute.BaseQueryPlan.iterator(BaseQueryPlan.java:364)
	at org.apache.phoenix.execute.HashJoinPlan.iterator(HashJoinPlan.java:234)
	at org.apache.phoenix.execute.DelegateQueryPlan.iterator(DelegateQueryPlan.java:144)
	at org.apache.phoenix.execute.DelegateQueryPlan.iterator(DelegateQueryPlan.java:139)
	at org.apache.phoenix.jdbc.PhoenixStatement$1.call(PhoenixStatement.java:314)
	at org.apache.phoenix.jdbc.PhoenixStatement$1.call(PhoenixStatement.java:293)
	at org.apache.phoenix.call.CallRunner.run(CallRunner.java:53)
	at org.apache.phoenix.jdbc.PhoenixStatement.executeQuery(PhoenixStatement.java:292)
	at org.apache.phoenix.jdbc.PhoenixStatement.executeQuery(PhoenixStatement.java:285)
	at org.apache.phoenix.jdbc.PhoenixStatement.executeQuery(PhoenixStatement.java:1798)

Script to reproduce:

CREATE TABLE TEST (PK1 INTEGER NOT NULL, PK2 INTEGER NOT NULL,  ID1 INTEGER, ID2 INTEGER CONSTRAINT PK PRIMARY KEY(PK1 , PK2))SALT_BUCKETS = 4;
upsert into test values (1,1,1,1);
upsert into test values (2,2,2,2);
upsert into test values (2,3,1,2);

create view TEST_VIEW as select * from TEST where PK1 in (1,2);
CREATE INDEX IDX_VIEW ON TEST_VIEW (ID1);


  select /*+ INDEX(TEST_VIEW IDX_VIEW) */ * from TEST_VIEW where ID1 = 1  ORDER BY ID2 LIMIT 500 OFFSET 0;

That happens because we have a point lookup optimization which reduces RowKeySchema to a single field, while we have more than one slot due salting.
jamestaylor can you please take a look? I'm not sure whether it should be fixed on the ScanUtil level or we just should not use point lookup in such cases.