Below is just a simple example of hive transactions. These are very useful on slowly changing type 1 dimension tables for which you do not wish to retain history, but only the most recent value of the row. The table must be bucketed, and the transactional=true property must be set on the table as shown below.
0: jdbc:hive2://localhost:10000/brierley> create table t(c int) clustered by (c) into 10 buckets stored as orc tblproperties('transactional'='true');
No rows affected (0.077 seconds)
0: jdbc:hive2://localhost:10000/brierley> insert into table t values(1);
INFO : Tez session hasn't been created yet. Opening session
…snip…
INFO : INFO : Table brierley.t stats: [numFiles=11, numRows=1, totalSize=2871, rawDataSize=0] 0: jdbc:hive2://localhost:10000/brierley> select * from t; +------+--+ | t.c | +------+--+ | 2 | +------+--+ 1 row selected (0.172 seconds) 0: jdbc:hive2://localhost:10000/brierley>