Mailing-List: contact issues-help@hive.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@hive.apache.org
Date: Tue, 31 Jan 2017 21:22:51 +0000 (UTC)
From: "Alexander Behm (JIRA)" <jira@apache.org>
To: issues@hive.apache.org
Message-ID: <JIRA.13035752.1484705883000.2743.1485897771831@Atlassian.JIRA>
In-Reply-To: <JIRA.13035752.1484705883000@Atlassian.JIRA>
References: <JIRA.13035752.1484705883000@Atlassian.JIRA> <JIRA.13035752.1484705883897@jira-lw-us.apache.org>
Subject: [jira] (HIVE-15653) Some ALTER TABLE commands drop table stats
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit
archived-at: Tue, 31 Jan 2017 21:22:56 -0000


    [ https://issues.apache.org/jira/browse/HIVE-15653?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15847540#comment-15847540 ] 

Alexander Behm commented on HIVE-15653:
---------------------------------------

Thanks, [~ctang.ma]. I've already added a workaround to Impala, but not using StatsSetupConst.DO_NOT_UPDATE_STATS because sometimes we do want to update stats (e.g. compute stats).
See: http://gerrit.cloudera.org:8080/5731

> Some ALTER TABLE commands drop table stats
> ------------------------------------------
>
>                 Key: HIVE-15653
>                 URL: https://issues.apache.org/jira/browse/HIVE-15653
>             Project: Hive
>          Issue Type: Bug
>          Components: Metastore
>    Affects Versions: 1.1.0
>            Reporter: Alexander Behm
>            Assignee: Chaoyu Tang
>            Priority: Critical
>             Fix For: 2.2.0
>
>         Attachments: HIVE-15653.1.patch, HIVE-15653.2.patch, HIVE-15653.3.patch, HIVE-15653.4.patch, HIVE-15653.5.patch, HIVE-15653.6.patch, HIVE-15653.patch
>
>
> Some ALTER TABLE commands drop the table stats. That may make sense for some ALTER TABLE operations, but certainly not for others. Personally, I I think ALTER TABLE should only change what was requested by the user without any side effects that may be unclear to users. In particular, collecting stats can be an expensive operation so it's rather inconvenient for users if they get wiped accidentally.
> Repro:
> {code}
> create table t (i int);
> insert into t values(1);
> analyze table t compute statistics;
> alter table t set tblproperties('test'='test');
> hive> describe formatted t;
> OK
> # col_name            	data_type           	comment             
> 	 	 
> i                   	int                 	                    
> 	 	 
> # Detailed Table Information	 	 
> Database:           	default             	 
> Owner:              	abehm               	 
> CreateTime:         	Tue Jan 17 18:13:34 PST 2017	 
> LastAccessTime:     	UNKNOWN             	 
> Protect Mode:       	None                	 
> Retention:          	0                   	 
> Location:           	hdfs://localhost:20500/test-warehouse/t	 
> Table Type:         	MANAGED_TABLE       	 
> Table Parameters:	 	 
> 	COLUMN_STATS_ACCURATE	false               
> 	last_modified_by    	abehm               
> 	last_modified_time  	1484705748          
> 	numFiles            	1                   
> 	numRows             	-1                  
> 	rawDataSize         	-1                  
> 	test                	test                
> 	totalSize           	2                   
> 	transient_lastDdlTime	1484705748          
> 	 	 
> # Storage Information	 	 
> SerDe Library:      	org.apache.hadoop.hive.serde2.lazy.LazySimpleSerDe	 
> InputFormat:        	org.apache.hadoop.mapred.TextInputFormat	 
> OutputFormat:       	org.apache.hadoop.hive.ql.io.HiveIgnoreKeyTextOutputFormat	 
> Compressed:         	No                  	 
> Num Buckets:        	-1                  	 
> Bucket Columns:     	[]                  	 
> Sort Columns:       	[]                  	 
> Storage Desc Params:	 	 
> 	serialization.format	1                   
> Time taken: 0.169 seconds, Fetched: 34 row(s)
> {code}
> The same behavior can be observed with several other ALTER TABLE commands.


--
This message was sent by Atlassian JIRA
(v6.3.15#6346)