Details
-
Bug
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
1.8.0
Description
Getting an error -" out-of-order key" for a query select v,count(k) from
kudu.test group by v where k is the primary key. This happens only when the
aggregation is done on primary key. Should drill move to the latest kudu
client to investigate this further?
Current drill kudu connector uses org.kududb:kudu-client:0.6.0 from
cloudera repository, where the latest released library
org.apache.kudu:kudu-client:1.2.0 is hosted on maven central. There are a
few breaking changes with the new library:
1. TIMESTAMP renamed to UNIXTIME_MICROS
2. In KuduRecordReader#setup -
KuduScannerBuilder#lowerBoundPartitionKeyRaw renamed to lowerBoundRaw
andKuduScannerBuilder#exclusiveUpperBoundPartitionKeyRaw renamed
exclusiveUpperBoundRaw. Both methods are deprecated.
3. In KuduRecordWriterImpl#updateSchema - client.createTable(name,
kuduSchema) requires CreateTableOperatios as the third argument