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

Posted to common-dev@hadoop.apache.org by "Jim Kellerman (JIRA)" <ji...@apache.org> on 2007/05/17 22:54:16 UTC

[jira] Created: (HADOOP-1384) HBase code update 0.0.0-4

HBase code update 0.0.0-4
-------------------------

                 Key: HADOOP-1384
                 URL: https://issues.apache.org/jira/browse/HADOOP-1384
             Project: Hadoop
          Issue Type: Improvement
          Components: contrib/hbase
            Reporter: Jim Kellerman
         Assigned To: Jim Kellerman


HClient
- Added logic to detect when a region is no longer being served by the
  cached region server
- Prevent null pointer exception in the get methods because the server
  can return null if the data is not found (contributed by Vuk Ercegovac)
- Disallow creation of reserved table names

HLocking
- Make the lock counter an AtomicInteger

HLog
- Add some DEBUG logging


HMaster
- Main processing loop did not sleep if there were no known region
  servers (contributed by Vuk Ercegovac)
- Move processing of kill list after processing all the incoming
  region server messages 
- deleteTable prevents meta scanner from running so that there will be
  no race condition between them. Previously, if deleteTable was in
  the middle of processing and the meta scanner ran, it could detect a
  region as unassigned and try to reassign it.
- deleteTable does not update the kill list until it has finished
  scanning a meta region. This will prevent lock contention on the
  region server.
- deleteTable will now delete any regions that are not being served
- deleteTable removes unassigned regions from the "to be assigned list"
- fix createTable so it correctly refuses to create a second table
  with the same name as an existing table

HRegion
- deleteRegion is a new static method
- added isClosed method
- closeAndDelete calls new deleteRegion method
- added locking

HRegionInterface
- getRegionInfo now throws NotServingRegionException

HRegionServer
- split or compact checker skips a region if it is closed
- cache flusher skips a region if it is closed
- lessen amount of time closeAndDeleteRegion holds the HRegionServer
  write lock
- getRegionInfo now throws NotServingRegionException
- moved checks for region being served to getRegion method (which
  throws NotServingRegionException)
- fixed bug in close in which the scanner was not actually getting closed
- implements RegionUnavailableListener in splitOrCompactChecker

Leases
- added some DEBUG logging
- fixed bug in which the leaseCheckFrequency was not being set causing
  the lease checker to spin and never sleep. (contributed by Vuk Ercegovac)

NotServingRegionException
- new exception (extends IOException)

RegionUnavailableListener
- new interface that enables split region to call back the caller just
  before the old region is taken off-line

Refactor test cases:
- MiniHBaseCluster modified
- Deleted Environment.java replaced with StaticTestEnvironment.java
- New classes:
  HBaseTestCase
  HBaseClusterTestCase
- New tests:
  TestHLog
  TestTable
- Modified tests:
  TestGet
  TestHBaseCluster
  TestHRegion
  TestScanner


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

Re: [jira] Created: (HADOOP-1384) HBase code update 0.0.0-4

Posted by Doug Cutting <cu...@apache.org>.

Michael Stack wrote:
> May I be added to the list of developers so I can be assigned hbase JIRA 
> issues (I just faxed in a Contributor License Agreement).

You were added some time ago.  I sometimes do this silently once folks 
have submitted a few patches.  Sorry, I probably should have told you!

Note that the ICLA is only required for committers, but never hurts.  I 
think the rationale is that committers can contribute changes without 
creating a patch and checking the license checkbox in jira, so they need 
a blanket agreement rather than the per-contribution agreement in Jira.

Doug

Re: [jira] Created: (HADOOP-1384) HBase code update 0.0.0-4

Posted by Michael Stack <st...@duboce.net>.

May I be added to the list of developers so I can be assigned hbase JIRA 
issues (I just faxed in a Contributor License Agreement).
Thanks,
St.Ack


Doug Cutting wrote:
> Jim Kellerman wrote:
>> So if you're trying out HBase, please feel free to enter Jira issues for
>> the component contrib/hbase if you find something that needs attention.
>>
>> Also if you're working on one of the to do items listed on the Wiki,
>> please enter a Jira and assign it to yourself so we can keep track of
>> who's working on what.
>
> Yes, please!  However folks can't be assigned issues in Jira until 
> they've been added to the Hadoop developers list, which I can do.  So, 
> if you want to be able to assign issues to yourself, please ask.
>
> Doug

Re: [jira] Created: (HADOOP-1384) HBase code update 0.0.0-4

Posted by Doug Cutting <cu...@apache.org>.

Jim Kellerman wrote:
> So if you're trying out HBase, please feel free to enter Jira issues for
> the component contrib/hbase if you find something that needs attention.
> 
> Also if you're working on one of the to do items listed on the Wiki,
> please enter a Jira and assign it to yourself so we can keep track of
> who's working on what.

Yes, please!  However folks can't be assigned issues in Jira until 
they've been added to the Hadoop developers list, which I can do.  So, 
if you want to be able to assign issues to yourself, please ask.

Doug

Re: [jira] Created: (HADOOP-1384) HBase code update 0.0.0-4

Posted by Jim Kellerman <ji...@powerset.com>.

I had replied to Doug off the list saying that I was intending to start
creating more focused jira issues and committing more frequently, much
easier to do now that I am a contrib committer.

This will probably be my last large patch for a while. I had accumulated
a number of changes and had recently broken my copy of HBase with one of
them. (and got it working again this morning).

So if you're trying out HBase, please feel free to enter Jira issues for
the component contrib/hbase if you find something that needs attention.

Also if you're working on one of the to do items listed on the Wiki,
please enter a Jira and assign it to yourself so we can keep track of
who's working on what.

Thanks!

-Jim

On Thu, 2007-05-17 at 15:57 -0700, Jim White wrote:
> Doug Cutting wrote:
> 
>  > Jim,
>  >
> > Rather than making big, omnibus patches every few weeks, you should 
> > contribute a more continuous stream of narrowly targeted changes.  That 
> > would encourage collaboration with others, as their patches could be 
> > interspersed with yours and have less chance of conflicts, no?
> 
> What you say is true Doug, but the most important thing is that he's 
> updating the public project his local changes on a fairly frequent basis.
> 
> Far too many folks simply never get around to pushing their code out at 
> all (typically thinking they'll do it at "1.0" or after "one more 
> thing...").
> 
> I'm sure the size and frequency of patches will follow a natural course 
> of change as HBase matures.
> 
> Also I notice that Jim K just became a committer, and that will probably 
> smooth out the process too.
> 
> And I'm looking forward to giving HBase a whirl myself before too long!
> 
> Jim W
> 
-- 
Jim Kellerman, Senior Engineer; Powerset
jim@powerset.com

Re: [jira] Created: (HADOOP-1384) HBase code update 0.0.0-4

Posted by Jim White <ji...@pagesmiths.com>.

Doug Cutting wrote:

 > Jim,
 >
> Rather than making big, omnibus patches every few weeks, you should 
> contribute a more continuous stream of narrowly targeted changes.  That 
> would encourage collaboration with others, as their patches could be 
> interspersed with yours and have less chance of conflicts, no?

What you say is true Doug, but the most important thing is that he's 
updating the public project his local changes on a fairly frequent basis.

Far too many folks simply never get around to pushing their code out at 
all (typically thinking they'll do it at "1.0" or after "one more 
thing...").

I'm sure the size and frequency of patches will follow a natural course 
of change as HBase matures.

Also I notice that Jim K just became a committer, and that will probably 
smooth out the process too.

And I'm looking forward to giving HBase a whirl myself before too long!

Jim W