[
https://issues.apache.org/jira/browse/TAJO-1685?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14630831#comment-14630831
]
ASF GitHub Bot commented on TAJO-1685:
--------------------------------------
Github user blrunner commented on the pull request:
https://github.com/apache/tajo/pull/631#issuecomment-122186801
Hi @jihoonson
Replicating local data to all worker don't look well. So I updated the
message for exception.
And I also think that supporting local file system is a fundamental
solution. But we need to put more efforts to implement it and this patch would
be a temporary work before providing local file system.
> Query fails when using table data which located on local file system
> occasionally on fully distributed mode.
> ------------------------------------------------------------------------------------------------------------
>
> Key: TAJO-1685
> URL: https://issues.apache.org/jira/browse/TAJO-1685
> Project: Tajo
> Issue Type: Improvement
> Components: Java Client, SQL Shell
> Reporter: Jaehwa Jung
> Assignee: Jaehwa Jung
>
> Tajo allows that the location of table would be set the path of local file
> system, for example, “file:///home/tajo/xyz”. When querying above table data
> on pseudo distributed mode, the query would finished successfully. Pseudo
> distributed mode for tajo means that TajoMaster and TajoWorker just run on
> the same host. But when querying the data on fully distribute mode, the query
> would failed because the data was’t located on all hosts for running
> TajoWorker. In this case, users would see ambiguous error message as follows.
> {code:xml}
> default> create external table table1 (
> > id int,
> > name text,
> > score float,
> > type text)
> > using text with ('text.delimiter'='|') location
> > 'file:///home/tajo/data.csv'
> > ;
> OK
> default> \d table1;
> table name: default.table1
> table uri: file:///home/tajo/data.csv
> store type: text
> number of rows: unknown
> volume: 60 B
> Options:
> 'text.delimiter'='|'
> schema:
> id INT4
> name TEXT
> score FLOAT4
> type TEXT
> default> select * from table1;
> id, name, score, type
> -------------------------------
> 1, abc, 1.1, a
> 2, def, 2.3, b
> 3, ghi, 3.4, c
> 4, jkl, 4.5, d
> 5, mno, 5.6, e
> (5 rows, 0.081 sec, 60 B selected)
> default> select count(*) from table1;
> ERROR: No error message
> {code}
> It doesn’t seems easy for users to know the cause of the error. We need to
> print a well-defined message for avoiding users confusion.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
