I assume you¹re using v0.6.x Kylin, which need the source table be
TextInputFormat; From 0.7.x, it will support multiple file formats like
ORC, Avro.

In 0.6.x, Kylin will read the table¹s hdfs file directly, detecting its
field delimiter (you can refer to
https://github.com/KylinOLAP/Kylin/blob/master/dictionary/src/main/java/com
/kylinolap/dict/lookup/FileTableReader.java#L51); If there is a cell¹s
content contains the delimiter, that may cause the problem; You may need
do a data cleanup in the source file.


On 2/25/15, 5:24 PM, "Jakob Stengård" <[email protected]> wrote:

>I get the following error when trying to build a cube with dimensions:
>
>[QuartzScheduler_Worker-1]:[2015-02-25
>10:12:23,362][ERROR][com.kylinolap.job.hadoop.dict.CreateDictionaryJob.run
>(CreateDictionaryJob.java:55)]
>- Bad hive table row, TableDesc [name=TESTDATA_TIME] expect 5 columns, but
>got [2012-07-30 08:56:01.776 2012 7 30 8]
>java.lang.IllegalStateException: Bad hive table row, TableDesc
>[name=TESTDATA_TIME] expect 5 columns, but got [2012-07-30 08:56:01.776
>2012 7 30 8]
>    at
>com.kylinolap.dict.lookup.SnapshotTable.takeSnapshot(SnapshotTable.java:68
>)
>    at
>com.kylinolap.dict.lookup.SnapshotManager.buildSnapshot(SnapshotManager.ja
>va:94)
>    at
>com.kylinolap.cube.CubeManager.buildSnapshotTable(CubeManager.java:208)
>    at ...
>
>
>What format do i need to have dim tables in?
>I've tried ORC, TEXTFILE and SEQUENCEFILE so far, but building the
>dictionary doesn't seem
>to work with any of those.
>
>Best regards
>/Jakob Stengård

Reply via email to