hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Steve Loughran <ste...@apache.org>
Subject Re: Removing JMM issues
Date Tue, 18 Nov 2008 14:24:47 GMT
Peter Veentjer wrote:
> Ok great,
> 
> I'll make the changes and check them in.

Submit patches, ideally with tests. If Hudson isn't happy, the code 
doesn't get in. Obviously concurrency problems are hard to test -for 
those we'll accept proofs of correctness in the formal notation of your 
choice :)

> 
> But this is just scratching the surface. Atm I'm trying to figure out
> which classes are the core of Hadoop so I can focus on that part
> first. There is so much source (and so little time) ;)

I would look at

* the RPC system
* HDFS

They are the underpinning, and if they have problems, they can be severe.

What may be good is to search for the code for
  -everywhere that uses Sleep(), as it's a sign of spinning on something
  -any/all catches of InterruptedException, especially the places that 
swallow them. It's a bad sign when code discards attempts to interrupt 
the thread, as it usually means some other part of the program wants you 
to stop, not keep looping

Mime
View raw message