Re[2]: GC Warning: Header allocation failed: Dropping block.

[Yura Smolsky <info at altervision dot biz>]

Hello, Hans.

BH> It means that the garbage collector's attempt to obtain more memory
BH> from the OS (on Linux it uses sbrk() or mmap()) failed.

BH> By default the GC uses as much memory as it thinks it needs and the
BH> OS will give it.  Try running with the GC_PRINT_STATS environment
BH> variable defined, and see what that says.

will do

BH> Also keep something like
BH> "top" running to see whether you are really running out of
BH> memory/swap space.  (I'm assuming a Linux environment.)

yes. i monitor through "top". it shows that process uses less than
half of total free memory...

BH> In general, for a large Java heap, you are better off running
BH> a recent gcj.  And ideally the collector should be built with
BH> -DLARGE_CONFIG, which I don't think it is by default.

ok. thanks.

>> -----Original Message-----
>> From: java-owner@gcc.gnu.org [mailto:java-owner@gcc.gnu.org] 
>> On Behalf Of Yura Smolsky
>> Sent: Wednesday, February 16, 2005 11:44 AM
>> To: java@gcc.gnu.org
>> Subject: GC Warning: Header allocation failed: Dropping block.
>> 
>> 
>> Hello, java.
>> 
>> I am user of PyLucene lib. It's port of java Lucene library 
>> to Python. And there is PythonThread which is really a 
>> wrapper around a Java/libgcj thread that python is tricked 
>> into thinking it's one of its own.
>> 
>> 
>> My code creates like 10 threads of this type. Each thread has 
>> reference to PyLucene's IndexSearcher object which size is 
>> very huge, like 400-800M.
>> 
>> When It creates many threads of this type and uses object 
>> IndexSearcher then I receive following exceptions: GC 
>> Warning: Header allocation failed: Dropping block or GC 
>> Warning: Out of Memory!  Returning NIL!
>> 
>> 1. What does this errors mean?
>> 2. When I decrease amount of threads to 2 or 3 then these 
>> exception are gone. Why? 3. How can I specify the total size 
>> of memory which can consume gcj threads or the whole process?
>> 
>> Thanks.
>> 
>> Yura Smolsky.
>> 
>> 
>> 




Yura Smolsky.