accumulo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From ke...@deenlo.com
Subject Re: Review Request: Patch for ACCUMULO-409
Date Mon, 02 Jul 2012 16:16:37 GMT


> On July 2, 2012, 3:13 p.m., Adam Fuchs wrote:
> > /branches/1.4/src/server/src/main/java/org/apache/accumulo/server/zookeeper/DistributedWorkQueue.java,
line 52
> > <https://reviews.apache.org/r/5684/diff/1/?file=117637#file117637line52>
> >
> >     What is special about a queue size of greater than one? Also, this code is repeated
in several spots -- should it be turned into a method?
> 
> kturner wrote:
>     I am not sure if there is way to tell if all threads are busy in the pool other than
the fact that at least one thing is queued.  Basically if all threads are busy, then the tablet
server should not reserve it because another tablet server my have processing capacity.
> 
> Adam Fuchs wrote:
>     The condition would be > 0 and not > 1, right?
>     
>     Seems like there's a potential race condition, where the first thread would just
empty the queue and none of the other threads would make progress.

Could increment an atomic long before adding a task to the thread pool and then decrement
the atomic long when a task finishes.  Could check this atomic long against the thread pool
size before adding to the pool.  This count will include queued and running.  Need to decrement
in a finally block of the task.


- kturner


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/5684/#review8792
-----------------------------------------------------------


On June 30, 2012, 3:10 a.m., kturner wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/5684/
> -----------------------------------------------------------
> 
> (Updated June 30, 2012, 3:10 a.m.)
> 
> 
> Review request for accumulo.
> 
> 
> Description
> -------
> 
> A patch that fixes the issue of failed bulk imports being copied through the master.
> 
> 
> This addresses bug ACCUMULO-409.
>     https://issues.apache.org/jira/browse/ACCUMULO-409
> 
> 
> Diffs
> -----
> 
>   /branches/1.4/src/core/src/main/java/org/apache/accumulo/core/Constants.java 1355540

>   /branches/1.4/src/core/src/main/java/org/apache/accumulo/core/zookeeper/ZooUtil.java
1355540 
>   /branches/1.4/src/server/src/main/java/org/apache/accumulo/server/client/BulkImporter.java
1355540 
>   /branches/1.4/src/server/src/main/java/org/apache/accumulo/server/master/tableOps/BulkImport.java
1355540 
>   /branches/1.4/src/server/src/main/java/org/apache/accumulo/server/tabletserver/BIFCopyProcessor.java
PRE-CREATION 
>   /branches/1.4/src/server/src/main/java/org/apache/accumulo/server/tabletserver/TabletServer.java
1355540 
>   /branches/1.4/src/server/src/main/java/org/apache/accumulo/server/zookeeper/DistributedWorkQueue.java
PRE-CREATION 
>   /branches/1.4/src/server/src/main/java/org/apache/accumulo/server/zookeeper/IZooReaderWriter.java
1355540 
>   /branches/1.4/src/server/src/main/java/org/apache/accumulo/server/zookeeper/ZooReaderWriter.java
1355540 
>   /branches/1.4/test/system/test4/bulk_import_test.sh 1355540 
> 
> Diff: https://reviews.apache.org/r/5684/diff/
> 
> 
> Testing
> -------
> 
> 
> Thanks,
> 
> kturner
> 
>


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message