Major Description
when column value from jdbc is null, sqoop will write it into hive table with blank string.
eg
jdbc:
A | B
1 | 1
2 | 2
a | null
hive table:
A | B
1 | 1
2 | 2
a |
because of this, when I use count(distinct B), it return 3, not 2, and it can lead to other problems
String cmd = String.format(Locale.ROOT, "%s/bin/sqoop import" + generateSqoopConfigArgString() + "--connect \"%s\" --driver %s --username %s --password %s --query \"%s AND \\$CONDITIONS\" " + "--target-dir %s/%s --split-by %s --boundary-query \"%s\" --null-string '' " + "--fields-terminated-by '%s' --num-mappers %d", sqoopHome, connectionUrl, driverClass, jdbcUser, jdbcPass, selectSql, jobWorkingDir, hiveTable, splitColumn, bquery, filedDelimiter, mapperNum);
the param '–null=string' should be '
N' instead of blank string ''.
I resolved this problem by replace the param. But it needs be configured in kylin.properties