accumulo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Eric Newton" <eric.new...@gmail.com>
Subject Re: Review Request 27829: ACCUMULO-1085
Date Mon, 10 Nov 2014 22:18:23 GMT


> On Nov. 10, 2014, 9:38 p.m., Mike Drob wrote:
> > test/src/test/java/org/apache/accumulo/test/AssignmentThreadsIT.java, line 34
> > <https://reviews.apache.org/r/27829/diff/1/?file=757270#file757270line34>
> >
> >     Plkease add a test that only one recovery will happen at a time.

Any idea how I can do that?


> On Nov. 10, 2014, 9:38 p.m., Mike Drob wrote:
> > test/src/test/java/org/apache/accumulo/test/AssignmentThreadsIT.java, line 44
> > <https://reviews.apache.org/r/27829/diff/1/?file=757270#file757270line44>
> >
> >     Is this just chars 0-9 & a-f?

yes; originally I left the splits as binary, but it was really ugly to read the logs.


> On Nov. 10, 2014, 9:38 p.m., Mike Drob wrote:
> > test/src/test/java/org/apache/accumulo/test/AssignmentThreadsIT.java, line 73
> > <https://reviews.apache.org/r/27829/diff/1/?file=757270#file757270line73>
> >
> >     Why go through arithmetic gymnastics instead of just sleep(11000)?

Taking the tablets offline probably takes most of the 11 seconds.  This just extends the wait
to make sure that it's at least 11 seconds, but not any more than that.


> On Nov. 10, 2014, 9:38 p.m., Mike Drob wrote:
> > server/tserver/src/main/java/org/apache/accumulo/tserver/TabletServer.java, line
2904
> > <https://reviews.apache.org/r/27829/diff/1/?file=757267#file757267line2904>
> >
> >     Why are we using a Semaphore over just making the method synchronized? It would
make sense if there was some other code section also checking on the semaphore, but I don't
see it.

For the fairness property. I'm going to have to make changes in this approach because the
metadata table has to be able to recover at all times, and the minor compaction must be included
in the recovery (Thanks to an offline review by [~kturner].)


- Eric


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/27829/#review60681
-----------------------------------------------------------


On Nov. 10, 2014, 9:11 p.m., Eric Newton wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/27829/
> -----------------------------------------------------------
> 
> (Updated Nov. 10, 2014, 9:11 p.m.)
> 
> 
> Review request for accumulo.
> 
> 
> Repository: accumulo
> 
> 
> Description
> -------
> 
> Allow for multiple threads to perform assignment, but use only a single thread for recovery
playback
> 
> 
> Diffs
> -----
> 
>   core/src/main/java/org/apache/accumulo/core/conf/Property.java f59b654 
>   server/tserver/src/main/java/org/apache/accumulo/tserver/TabletServer.java 1e81947

>   server/tserver/src/main/java/org/apache/accumulo/tserver/TabletServerResourceManager.java
ba86522 
>   server/tserver/src/main/java/org/apache/accumulo/tserver/tablet/Tablet.java 9490903

>   test/src/test/java/org/apache/accumulo/test/AssignmentThreadsIT.java PRE-CREATION 
> 
> Diff: https://reviews.apache.org/r/27829/diff/
> 
> 
> Testing
> -------
> 
> Added an IT that verifies that tablet loading goes faster with more threads.
> 
> 
> Thanks,
> 
> Eric Newton
> 
>


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