[ 
https://issues.apache.org/jira/browse/MAPREDUCE-6141?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Jason Lowe updated MAPREDUCE-6141:
----------------------------------
    Attachment: MAPREDUCE-6141v2.patch

Thanks for the review, Jonathan!

bq. Should we consider making mapreduce.jobhistory.recovery.store.class default 
to levelb as part of a follow on jira

Yes, my intention was to get some runtime experience with this and then propose 
we default to this at a later date.

bq. please document the test blocks testTokenStore with comments to help guide 
the reader of the test code

Done.

bq.  if version in db in null upon initialization, I think we want to return 
this as the current version and not 1.0

The intent of getCurrentVersion is to return the version of the code not the 
database.  The code always converts the database contents to conform to the 
current version on startup or fails if the database contents cannot be 
converted (e.g.: major version upgrade with no automatic conversion in the 
code).

bq. Is the mapreduce-default.xml sufficient documentation to use this feature?

I believe so.  This is just another backend choice for history server recovery, 
and the description for the one property specific to this backend should be 
sufficient to know how to configure it.  I wasn't planning on writing up docs 
for choosing one backend or the other, but I can add those if necessary.

> History server leveldb recovery store
> -------------------------------------
>
>                 Key: MAPREDUCE-6141
>                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-6141
>             Project: Hadoop Map/Reduce
>          Issue Type: Improvement
>          Components: jobhistoryserver
>            Reporter: Jason Lowe
>            Assignee: Jason Lowe
>         Attachments: MAPREDUCE-6141.patch, MAPREDUCE-6141v2.patch
>
>
> It would be nice to have a leveldb option to the job history server recovery 
> store.  Leveldb would provide some benefits over the existing filesystem 
> store such as better support for atomic operations, fewer I/O ops per state 
> update, and far fewer total files on the filesystem.



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

Reply via email to