CMS collection keep working during holiday

Jon Masamitsu Jon.Masamitsu at Sun.COM
Wed Oct 8 22:09:18 UTC 2008


Ken,

One of the differences between the CMS concurrent
collection of the old gen and a full collection
(concurrent mode fail) is the way SoftReferences
are processed when the heap is full of live data.
A full collection will clear all SoftReferences
before it throws an out-of-memory.  If a full
collection is done and the heap is still full
of live data, a full collection will clear all
SoftReferences.  When CMS does a concurrent
collection and the heap is still full, a
concurrent mode failure occurs resulting in
a full collection.  And as with full collections,
all the SoftReferences will be cleared if
necessary.  If you set -XX: SoftRefLRUPolicyMSPerMB=0
SoftReferences will be cleared promptly and should
tell us if we're encountering this difference in
the way SoftReferences are handled.

Jon


On 10/08/08 09:13, Ken-- at newsgroupstats.hk wrote:
> Attached.
> http://www.nabble.com/file/p19882121/histo.txt histo.txt 
> 
> 
> 
> 
> antonios.printezis wrote:
>> jmap -histo too
>>
>> Tony
>>
>> kirk wrote:
>>> Ken-- at newsgroupstats.hk wrote:
>>>> Follow up to the jconsole above. This gif is captured today. You can 
>>>> see the
>>>> first CMS after a FULL GC is working. The 2nd one is working too but the
>>>> memory free from 2nd generation is much less than the first CMS. The 
>>>> memory
>>>> free by 3rd CMS is further decreased.
>>>>
>>>> The 5th CMS cannot free enough memory to make occanpcy in old-gen to 
>>>> less
>>>> than 70%. So, continue non-stopping CMS happen again until a Full GC
>>>> (triggered by Concurrent Mode Failure).
>>>>
>>>> http://www.nabble.com/file/p19872463/jconsole_20081008.jpg
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>   
>>> If GC is failing to recover memory, you either have a memory leak or 
>>> long loitering objects. I'd suggest visualvm or netbeans memory 
>>> profiling using the generations option to find out what objects are 
>>> surviving
>>>
>>> Regards,
>>> Kirk
>> -- 
>> ---------------------------------------------------------------------
>> | Tony Printezis, Staff Engineer   | Sun Microsystems Inc.          |
>> |                                  | MS UBUR02-311                  |
>> | e-mail: tony.printezis at sun.com   | 35 Network Drive               |
>> | office: +1 781 442 0998 (x20998) | Burlington, MA 01803-2756, USA |
>> ---------------------------------------------------------------------
>> e-mail client: Thunderbird (Linux)
>>
>>
>>
>>
> 



More information about the hotspot-gc-dev mailing list