hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Ted Yu (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (HBASE-9019) Port HBASE-8690: Reduce unnecessary getFileStatus hdfs calls in TTL hfile and hlog cleanners to 0.94
Date Tue, 23 Jul 2013 03:42:48 GMT

     [ https://issues.apache.org/jira/browse/HBASE-9019?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Ted Yu updated HBASE-9019:
--------------------------

    Attachment: 9019.94

Patch for 0.94

Tests touched by the patch pass.

Running test suite now.
                
> Port HBASE-8690: Reduce unnecessary getFileStatus hdfs calls in TTL hfile and hlog cleanners
to 0.94
> ----------------------------------------------------------------------------------------------------
>
>                 Key: HBASE-9019
>                 URL: https://issues.apache.org/jira/browse/HBASE-9019
>             Project: HBase
>          Issue Type: Improvement
>            Reporter: Ted Yu
>            Assignee: Ted Yu
>            Priority: Minor
>         Attachments: 9019.94
>
>
> For each in file in archive dir, the TimeToLiveHFileCleaner need call getFileStatus to
get the modify time of file. Actually the CleanerChore have had the file status when listing
its parent dir. 
> When we set the TTL to 7 days in our cluster for data security, the number of files left
in archive dir is up to 65 thousands. In each clean period, TimeToLiveHFileCleaner will generate
ten thousand getFileStatus call in a short time, which is very heavy for hdfs namenode.
> Fix is to change the path param to FileStatus in isFileDeletable method and reduce unnecessary
getFileStatus hdfs calls in TTL cleaners.
> This issue is to backport this improvement to 0.94

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Mime
View raw message