Details
-
Bug
-
Status: Open
-
Minor
-
Resolution: Unresolved
-
None
-
None
-
None
-
None
Description
I have not done extensive testing on it, but when I specified SALT_BUCKETS=0 while creating an index, I get no error and this is a valid use case to disable salting on index when the data table is salted:
create table tsalt (k INTEGER PRIMARY KEY, v1 INTEGER) SALT_BUCKETS=4;
upsert into tsalt (k, v1) VALUES (1, 100);
create uncovered index tsaltidx on tsalt (PHOENIX_ROW_TIMESTAMP()) SALT_BUCKETS=0;
From schema and hbase regions, it is correctly treated as no salting scenario.
0: jdbc:phoenix:localhost> select salt_buckets from system.catalog where table_name = 'TSALTIDX' and salt_buckets is not null;
--------------
SALT_BUCKETS |
--------------
--------------
No rows selected (0.026 seconds)
hbase:001:0> list_regions 'TSALTIDX'
SERVER_NAME | REGION_NAME | START_KEY | END_KEY | SIZE | REQ | LOCALITY |
----------------------------- | --------------------------------------------------------- | ---------- | ---------- | ----- | ----- | ---------- |
localhost,16020,1708958003582 | TSALTIDX,,1708958225506.a72b20c15cecba23289a03cd6956ec15. | | | 0 | 3 | 0.0 |
1 rows
However, when I query through the index, I get an ArithmeticError for divide by zero.
{{0: jdbc:phoenix:localhost> select /*+ INDEX(TSALT TSALTIDX) */ * from TSALT;Caused by: java.lang.ArithmeticException: / by zero
at org.apache.phoenix.schema.SaltingUtil.getSaltingByte(SaltingUtil.java:79)
at org.apache.phoenix.index.IndexMaintainer.buildDataRowKey(IndexMaintainer.java:916)
at org.apache.phoenix.coprocessor.UncoveredIndexRegionScanner.scanIndexTableRows(UncoveredIndexRegionScanner.java:253)
at org.apache.phoenix.coprocessor.UncoveredIndexRegionScanner.scanIndexTableRows(UncoveredIndexRegionScanner.java:274)
at org.apache.phoenix.coprocessor.UncoveredIndexRegionScanner.next(UncoveredIndexRegionScanner.java:382)
at org.apache.phoenix.coprocessor.BaseRegionScanner.nextRaw(BaseRegionScanner.java:56)
at org.apache.phoenix.iterate.RegionScannerFactory$1.nextRaw(RegionScannerFactory.java:257)
My suspicion is that table cells have number buckets stored as zero, so PTableImpl for the index gets constructed to return 0 from getBucketNum() and this is causing the divide by 0 error.
Attachments
Issue Links
- blocks
-
PHOENIX-7001 Change Data Capture leveraging Max Lookback and Uncovered Indexes
- Resolved