Github user dineshjoshi commented on a diff in the pull request:

    https://github.com/apache/cassandra/pull/279#discussion_r227581961
  
    --- Diff: 
test/burn/org/apache/cassandra/utils/memory/LongBufferPoolTest.java ---
    @@ -36,9 +36,34 @@
     
     import static org.junit.Assert.*;
     
    +/**
    + * Long BufferPool test - make sure that the BufferPool allocates and 
recycles
    + * ByteBuffers under heavy concurrent usage.
    + *
    + * The test creates two groups of threads
    + *
    + * - the burn producer/consumer pair that allocates 1/10 poolSize and then 
returns
    + *   all the memory to the pool. 50% is freed by the producer, 50% passed 
to the consumer thread.
    + *
    + * - a ring of worker threads that allocate buffers and either immediately 
free them,
    + *   or pass to the next worker thread for it to be freed on it's behalf.  
Periodically
    + *   all memory is freed by the thread.
    + *
    + * While the burn/worker threads run, the original main thread checks that 
all of the threads are still
    + * making progress every 10s (no locking issues, or exits from assertion 
failures),
    + * and that every chunk has been freed at least once during the previous 
cycle (if that was possible).
    + *
    + * The test does not expect to survive out-of-memory errors, so needs 
sufficient heap memory
    + * for non-direct buffers and the debug tracking objects that check the 
allocate buffers.
    + * (The timing is very interesting when Xmx is lowered to increase garbage 
collection pauses, but do
    + * not set it too low).
    + */
     public class LongBufferPoolTest
     {
         private static final Logger logger = 
LoggerFactory.getLogger(LongBufferPoolTest.class);
    +    final int avgBufferSize = 16 << 10;
    --- End diff --
    
    Constants are typically uppercased with underscores as separators eg. 
`AVG_BUFFER_SIZE`. Do you also need this to be package private? It can be made 
private and static.


---

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to