You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@hbase.apache.org by "Ashish Shinde (JIRA)" <ji...@apache.org> on 2011/01/25 05:46:44 UTC

[jira] Created: (HBASE-3474) HFileOutputFormat to use column family's compression algorithm

HFileOutputFormat to use column family's compression algorithm
--------------------------------------------------------------

                 Key: HBASE-3474
                 URL: https://issues.apache.org/jira/browse/HBASE-3474
             Project: HBase
          Issue Type: Improvement
          Components: mapreduce
    Affects Versions: 0.90.0
         Environment: All
            Reporter: Ashish Shinde


HFileOutputFormat  currently creates HFile writer's using a compression algorithm set as configuration "hbase.hregion.max.filesize" with default as no compression. The code does not take into account the compression algorithm configured for the table's column family.  As a result bulk uploaded tables are not compressed until a major compaction is run on them. This could be fixed by using the column family descriptors while creating HFile writers.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

[jira] [Updated] (HBASE-3474) HFileOutputFormat to use column family's compression algorithm

Posted by "Todd Lipcon (JIRA)" <ji...@apache.org>.

     [ https://issues.apache.org/jira/browse/HBASE-3474?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Todd Lipcon updated HBASE-3474:
-------------------------------

    Attachment: hbase-3474.txt

Hi Ashish. I took your patch and cleaned up the unit tests a bit to share some of the mocking code. I also added a {{getCompressionAlgorithm}} call to {{HFile.Reader}} as you suggested.

I removed one of the unit tests since it seemed to be entirely subsumed by the other one. Also updated the unit tests to check cases like 0-CF and 1-CF.

Can you check over this updated patch and make sure it looks good to you?

> HFileOutputFormat to use column family's compression algorithm
> --------------------------------------------------------------
>
>                 Key: HBASE-3474
>                 URL: https://issues.apache.org/jira/browse/HBASE-3474
>             Project: HBase
>          Issue Type: Improvement
>          Components: mapreduce
>    Affects Versions: 0.92.0
>         Environment: All
>            Reporter: Ashish Shinde
>             Fix For: 0.92.0
>
>         Attachments: hbase-3474.txt, patch3474.txt, patch3474.txt, patch3474.txt
>
>   Original Estimate: 48h
>  Remaining Estimate: 48h
>
> HFileOutputFormat  currently creates HFile writer's using a compression algorithm set as configuration "hbase.hregion.max.filesize" with default as no compression. The code does not take into account the compression algorithm configured for the table's column family.  As a result bulk uploaded tables are not compressed until a major compaction is run on them. This could be fixed by using the column family descriptors while creating HFile writers.

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] Commented: (HBASE-3474) HFileOutputFormat to use column family's compression algorithm

Posted by "Todd Lipcon (JIRA)" <ji...@apache.org>.

    [ https://issues.apache.org/jira/browse/HBASE-3474?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12989061#comment-12989061 ] 

Todd Lipcon commented on HBASE-3474:
------------------------------------

Hey Ashish. This is looking good. A few more comments before I think it's ready to commit:
- If you make configureCompression and createFamilyCompressionMap package-private instead of private, you could add a nice unit test for them. You can use Mockito to make a mock HTable that returns the HColumnDescriptors you want for the sake of the test - if you need a hand with this you can catch me on #hbase IRC
- Probably a good idea to put "families.compression" in a constant, and maybe change it to be more obviously scoped - ie something like "hbase.hfileoutputformat.families.compression"
- When you upload a patch, you should take the diff from the root of the SVN trunk/ - so that it applies with "patch -p0" from inside trunk/.

Thanks!

> HFileOutputFormat to use column family's compression algorithm
> --------------------------------------------------------------
>
>                 Key: HBASE-3474
>                 URL: https://issues.apache.org/jira/browse/HBASE-3474
>             Project: HBase
>          Issue Type: Improvement
>          Components: mapreduce
>    Affects Versions: 0.92.0
>         Environment: All
>            Reporter: Ashish Shinde
>             Fix For: 0.92.0
>
>         Attachments: patch3474.txt, patch3474.txt
>
>   Original Estimate: 48h
>  Remaining Estimate: 48h
>
> HFileOutputFormat  currently creates HFile writer's using a compression algorithm set as configuration "hbase.hregion.max.filesize" with default as no compression. The code does not take into account the compression algorithm configured for the table's column family.  As a result bulk uploaded tables are not compressed until a major compaction is run on them. This could be fixed by using the column family descriptors while creating HFile writers.

-- 
This message is automatically generated by JIRA.
-
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] Commented: (HBASE-3474) HFileOutputFormat to use column family's compression algorithm

Posted by "Todd Lipcon (JIRA)" <ji...@apache.org>.

    [ https://issues.apache.org/jira/browse/HBASE-3474?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12990694#comment-12990694 ] 

Todd Lipcon commented on HBASE-3474:
------------------------------------

Let me know if you need a hand with the test, happy to help. Mockito is strange at first but once you get the hang of it, it's pretty nice

> HFileOutputFormat to use column family's compression algorithm
> --------------------------------------------------------------
>
>                 Key: HBASE-3474
>                 URL: https://issues.apache.org/jira/browse/HBASE-3474
>             Project: HBase
>          Issue Type: Improvement
>          Components: mapreduce
>    Affects Versions: 0.92.0
>         Environment: All
>            Reporter: Ashish Shinde
>             Fix For: 0.92.0
>
>         Attachments: patch3474.txt, patch3474.txt
>
>   Original Estimate: 48h
>  Remaining Estimate: 48h
>
> HFileOutputFormat  currently creates HFile writer's using a compression algorithm set as configuration "hbase.hregion.max.filesize" with default as no compression. The code does not take into account the compression algorithm configured for the table's column family.  As a result bulk uploaded tables are not compressed until a major compaction is run on them. This could be fixed by using the column family descriptors while creating HFile writers.

-- 
This message is automatically generated by JIRA.
-
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] Commented: (HBASE-3474) HFileOutputFormat to use column family's compression algorithm

Posted by "Todd Lipcon (JIRA)" <ji...@apache.org>.

    [ https://issues.apache.org/jira/browse/HBASE-3474?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12987693#action_12987693 ] 

Todd Lipcon commented on HBASE-3474:
------------------------------------

Thanks for taking this on, Ashish. A few comments on the patch:
- for createFamilyCompressionMap, specify in the javadoc that this is run inside the task to deserialize the map back out of the Configuration
- In javadoc for configureCompression, I don't think there's any reason to include the @param table and @param conf since they're kind of obvious
- in createFamilyCompressionMap, 'compression' is misspelled in one var name
- The encoding you've chosen for this config value isn't the best, since we currently do support ',' in a column family name. What about encoding the map like a query string in a URL? foo=bar&baz=blah - and using URLEncoder/URLDecoder to escape the =s and &s that might be in there.

> HFileOutputFormat to use column family's compression algorithm
> --------------------------------------------------------------
>
>                 Key: HBASE-3474
>                 URL: https://issues.apache.org/jira/browse/HBASE-3474
>             Project: HBase
>          Issue Type: Improvement
>          Components: mapreduce
>    Affects Versions: 0.92.0
>         Environment: All
>            Reporter: Ashish Shinde
>             Fix For: 0.92.0
>
>         Attachments: patch3474.txt
>
>   Original Estimate: 48h
>  Remaining Estimate: 48h
>
> HFileOutputFormat  currently creates HFile writer's using a compression algorithm set as configuration "hbase.hregion.max.filesize" with default as no compression. The code does not take into account the compression algorithm configured for the table's column family.  As a result bulk uploaded tables are not compressed until a major compaction is run on them. This could be fixed by using the column family descriptors while creating HFile writers.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.