[jira] [Commented] (TRAFODION-2195) create Trafodion table with hbase options DATA_BLOCK_ENCODING and COMPRESSION together won't get all results when executing queries

["Suresh Subbiah (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/TRAFODION-2195?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15457336#comment-15457336 ] 

Suresh Subbiah commented on TRAFODION-2195:
-------------------------------------------

Thank you. I was able to see the same problem with a different situation.

Another workaround will be to alter the table from hbase shell to change encoding to 'NONE'. For example
disable 'TRAFODION.SEABASE.EVENT'
alter 'TRAFODION.SEABASE.EVENT', {NAME => '#1', DATA_BLOCK_ENCODING => 'NONE'}
enable 'TRAFODION.SEABASE.EVENT'
describe 'TRAFODION.SEABASE.EVENT'
major_compact 'TRAFODION.SEABASE.EVENT'

In my case all three encoding options (DIFF, FAST_DIFF & PREFIX) had the same problem.

Could you please state what version of HBase was used when this problem was seen? I saw it on CDH 5.4.

> create Trafodion table with  hbase options DATA_BLOCK_ENCODING and COMPRESSION together  won't get all results when executing queries
> -------------------------------------------------------------------------------------------------------------------------------------
>
>                 Key: TRAFODION-2195
>                 URL: https://issues.apache.org/jira/browse/TRAFODION-2195
>             Project: Apache Trafodion
>          Issue Type: Bug
>            Reporter: QiaoYanke
>
>   I create a table using the follow ddl, and I insert into the table 1 million lines, and the 'select count(*) from event' will get the right result. But when I execute 'select * from event', I only get several hundred lines. 
>   After I create the table just using 'HBASE_OPTIONS( 
> COMPRESSION='snappy',
> MEMSTORE_FLUSH_SIZE = '1073741824') ' without data_block_encoding, I will get the data I insert .
> here is a link make help me just using compression.
> http://hadoop-hbase.blogspot.com/2016/02/hbase-compression-vs-blockencoding_17.html
> CREATE TABLE EVENT (
>   v_date timestamp(6) NOT NULL,
>   sid varchar(16 BYTES) NOT NULL DEFAULT '',
>   uid varchar(20 BYTES) NOT NULL DEFAULT '',
>   vid int unsigned NOT NULL ,
>   idx tinyint NOT NULL,
>   created_at int unsigned NOT NULL ,
>   p_id bigint  NOT NULL ,
>   category varchar(255) CHARACTER SET UTF8 NOT NULL,
>   e_action varchar(255) CHARACTER SET UTF8 NOT NULL,
>   label varchar(255) CHARACTER SET UTF8 NOT NULL,
>   e_value int NOT NULL,
>   is_bounced tinyint NOT NULL DEFAULT 0,
>   primary key (sid,v_date desc,uid,vid,idx)
> )
> salt using 4 partitions on (sid)
> division by (date_trunc('day', v_date))
> HBASE_OPTIONS( DATA_BLOCK_ENCODING = 'FAST_DIFF',
> COMPRESSION='snappy',
> MEMSTORE_FLUSH_SIZE = '1073741824');



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)