Thanks Andrea,
I have several gzipped data dumps, I'll send you an sftp login info to get them
shortly.
I used Eclipse Memory Analyzer (http://eclipse.org/mat/), which tries to
identify potential leak suspects based on a heap dump.
Here is the current result of jmap, PS Old Gen is at this time about 75% full:
-histo:live
num #instances #bytes class name
----------------------------------------------
1: 26627 496999424 [B
2: 3656028 152263424 [C
3: 4329868 138555776 java.lang.String
4: 1828746 96256040 [Ljava.lang.Object;
5: 2323575 92943000 org.apache.xerces.dom.AttrNSImpl
6: 2226851 71259232 org.apache.xerces.dom.TextImpl
7: 1437451 57498040
org.eclipse.emf.ecore.util.EObjectContainmentEList
8: 419811 57094296
org.eclipse.xsd.impl.XSDElementDeclarationImpl
9: 538143 43051440 javax.imageio.metadata.IIOAttr
10: 600818 38452352 org.apache.xerces.dom.ElementNSImpl
11: 420048 36964224 org.eclipse.xsd.impl.XSDParticleImpl
12: 906590 36263600
org.eclipse.emf.ecore.util.EDataTypeUniqueEList
13: 893506 35740240
org.eclipse.emf.ecore.util.EDataTypeUniqueEList$Unsettable
14: 192649 30271176 <constMethodKlass>
15: 535819 30005864 javax.imageio.metadata.IIOMetadataNode
16: 1214128 29139072 java.util.ArrayList
17: 192649 23142600 <methodKlass>
18: 961124 23066976
org.eclipse.xsd.impl.XSDConcreteComponentImpl$XSDContentsEList
19: 19030 22289344 <constantPoolKlass>
20: 227148 20226896 [Ljava.util.HashMap$Entry;
21: 478857 19154280 org.eclipse.emf.ecore.util.EObjectEList
22: 572791 18329312 java.util.Vector
-heap
Server compiler detected.
JVM version is 17.1-b03
using thread-local object allocation.
Parallel GC with 8 thread(s)
Heap Configuration:
MinHeapFreeRatio = 40
MaxHeapFreeRatio = 70
MaxHeapSize = 3221225472 (3072.0MB)
NewSize = 1310720 (1.25MB)
MaxNewSize = 17592186044415 MB
OldSize = 5439488 (5.1875MB)
NewRatio = 2
SurvivorRatio = 8
PermSize = 268435456 (256.0MB)
MaxPermSize = 536870912 (512.0MB)
Heap Usage:
PS Young Generation
Eden Space:
capacity = 550895616 (525.375MB)
used = 195447488 (186.39324951171875MB)
free = 355448128 (338.98175048828125MB)
35.478134572775396% used
>From Space:
capacity = 192806912 (183.875MB)
used = 0 (0.0MB)
free = 192806912 (183.875MB)
0.0% used
To Space:
capacity = 196214784 (187.125MB)
used = 0 (0.0MB)
free = 196214784 (187.125MB)
0.0% used
PS Old Generation
capacity = 2147483648 (2048.0MB)
used = 1672784152 (1595.2912826538086MB)
free = 474699496 (452.7087173461914MB)
77.8950821608305% used
PS Perm Generation
capacity = 268435456 (256.0MB)
used = 142182176 (135.59548950195312MB)
free = 126253280 (120.40451049804688MB)
52.96698808670044% used
From: Andrea Aime
<[email protected]<mailto:[email protected]>>
Date: Sat, 28 Jan 2012 12:28:55 -0500
To: Gabriel Roldán <[email protected]<mailto:[email protected]>>
Cc: Matt Bertrand
<[email protected]<mailto:[email protected]>>, Justin
Deoliveira <[email protected]<mailto:[email protected]>>,
"[email protected]<mailto:[email protected]>"
<[email protected]<mailto:[email protected]>>
Subject: Re: [Geoserver-users] Geoserver memory usage - possible GWC leak?
On Fri, Jan 27, 2012 at 8:50 PM, Gabriel Roldan
<[email protected]<mailto:[email protected]>> wrote:
I will try with a profiling session later today.
@Justin: any advice on what kind of care should be taken to minimize
the proliferation of XSDSchemaImpl and GridCoverageReader instances?
Normally too many XSDSchemaImpl (hundreds of MB, a few tens is unfortunately
normal) is normally indication of a memory leak in the WFS/schema subsystem.
GridCoverageReader are cached in ResourcePool, as opening them is expensive,
however per se they do not hold much memory, if they do it's via the
raster metadata,
normally caused my many ill structured raster files, like large and
striped geotiffs
(no inner tiling structure).
Bertrand, could you provide more details on your memory analysis?
Do you have a memory dump we can look into, or the output of jmap,
something like:
jmap -histo:live <jvm pid> | head -25
Cheers
Andrea
--
-------------------------------------------------------
Ing. Andrea Aime
GeoSolutions S.A.S.
Tech lead
Via Poggio alle Viti 1187
55054 Massarosa (LU)
Italy
phone: +39 0584 962313
fax: +39 0584 962313
mob: +39 339 8844549
http://www.geo-solutions.it
http://geo-solutions.blogspot.com/
http://www.youtube.com/user/GeoSolutionsIT
http://www.linkedin.com/in/andreaaime
http://twitter.com/geowolf
-------------------------------------------------------
------------------------------------------------------------------------------
Try before you buy = See our experts in action!
The most comprehensive online learning library for Microsoft developers
is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
Metro Style Apps, more. Free future releases when you subscribe now!
http://p.sf.net/sfu/learndevnow-dev2
_______________________________________________
Geoserver-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/geoserver-users
