Debugging Java from Dumps

© 2011 IBM Corporation
Chris Bailey – Java Support Architect
Debugging from Dumps
Debugging Applications and Products from Dumps

Debugging Strategies
■ Two common strategies for debugging:
–
■ Debugging from Trace: Log or print relevant data at runtime
– Typical approach used on distributed platforms
– Pros: History is available, showing how the failure point was reached
Collected data usually has context via the trace point
– Cons: only specified data is collected, often resulting in multiple iterations
ongoing performance overhead to running application
■ Debugging from “Dumps”: Collect all available data on failure
– Used widely on on System Z
– Pros: All available data is collected
No ongoing performance overhead to application
– Cons: Very limited history of how the failure point was reached: only current thread stacks
Limited context around data: only Object or structure types
■ Using a combination of the two provides powerful debugging capabilities
– Note: in-memory trace buffers are in the dumps, so can aid with trace data collection!

Dump Types
■ Three different dump types:
Dump
Format
Approx.
Size on
Disk
Objects,
Classes,
Class
Loaders
Thread
Details
Field
Names
Field and
Array
References
Primitive
Fields
Primitive
Array
Contents
Accurate
GC Roots
Native
Memory
and
Threads
IBM
PHD
20% of
Java heap
size  with
Javacore
*      
HPROF Java heap
size        
System
dump
Java heap
size + 30%        

Dump Generation
■ Interactive Methods
– Using a SIGQUIT or Ctrl-Break:
● When a Ctrl-Break or SIGQUIT (usually generated using "kill -3") is sent to the IBM
runtime
● a "user" event is generated in the IBM dump engine
● can however be configured to generate either a PHD format dump, or a System dump
using the following command line options to the Java application
● -Xdump:heap:events=user (PHD Heap Dump)
● -Xdump:system:events=user (System dump)
– Using the operating system to produce a system dump:
● AIX: gencore (or the destructive kill -6 or kill -11)
● Linux/Solaris: gcore (or the destructive kill -6 or kill -11)
● Windows: userdump.exe
● z/OS: SVCDUMP/Console Dump
– Using IBM Monitoring and Diagnostics Tools for Java - Health Center:
● The Health Center tool provides a menu option to request dumps from a running Java
process
● can request either a PHD or system dump

Dump Generation
■ Interactive Methods
– Using the wsadmin scripting client: (using Jython with the -lang jython option):
● Thread dump:
AdminControl.invoke(AdminControl.completeObjectName("type=JVM,proc
ess=server1,*"), "dumpThreads")
● System dump:
ess=server1,*"), "generateSystemDump")
● Portable heap dump (PHD):
ess=server1,*"), "generateHeapDump")

load Class load class name -Xdump:system:events=load,filter=com/ibm/example/Example
unload Class unload class name -Xdump:system:events=unload,filter=com/ibm/example/Example
throw An exception being thrown exception name -Xdump:system:events=throw,filter=java/net/ConnectException
catch An exception being caught exception name -Xdump:system:events=catch,filter=java/net/ConnectException
systhrow A Java exception is about
Dump Generation
■ Event Based Methods
– Using the IBM dump engine:
● The dump engine provides a large number of events on which you can produce a "heap"
or "system" dump
●
Event Description Filtering Example
●
gpf GPF (Crash) -Xdump:system:events=gpf
●
user ●
User generated signal
●
vmstop ●
VM shutdown, including
●
●
●
●
●
●
●
●
allocation A Java object is allocated size of object -Xdump:system:events=allocate,filter=#5m
●
(SIGQUIT or Ctrl-Break)
-Xdump:system:events=user
call to System.exit()
exit code -Xdump:system:events=vmstop,filter=#0..#10
to be thrown by the JVM
exception name -Xdump:system:events=systhrow,
filter=java/lang/OutOfMemoryError,range=1..4
● Exceptions can also be filtered on throwing method using '#'
– -Xdump:system:events=throw,filter=java/lang/NullPointerException#com/ibm/example/Example.bad

Generating Dumps
■ Event Based Methods
– Using the IBM trace engine:
● the trace engine allows PHD and system dumps to be triggered on method entry or exit
● achieved using the following command line option to the application:
● produces a system dump when the Example.trigger() method is called
– -Xtrace:maximal=mt,trigger=method{com/ibm/example/Example.trigger,sysdump}
● produces a PHD dump when the Example.trigger() method is called
– -Xtrace:maximal=mt,trigger=method{com/ibm/example/Example.trigger,heapdump}
● set a range so that you don't create dumps every time the method is called
– -Xtrace:maximal=mt,trigger=method{com/ibm/example/Example.trigger,sysdump,,5,1}
–
■ Programmatic Methods:
● provide a com.ibm.jvm.Dump class with three methods:
– javaDump(), heapDump() and systemDump()

Generating Dumps
■ Memory Analyzer also provides an "Acquire
Heap Dump" option that allows you to trigger
and load a snapshot dump from a Java process
running on the same machine as Memory
Analyzer.
■ On HotSpot based runtimes the Memory
Analyzer generates the dump using JMap.
■ For the IBM runtimes the dump is generated
using the "late attach" functionality and the the
programmatic API.

0 5000000000 10000000000 15000000000
350000
300000
250000
200000
150000
100000
50000
0
Phil (Linux 64 bit PPC)
Same Disk
RAM Disk
GB Netw ork
Core Size (bytes)
Time to Generate (ms)
Dump Generation
■ Dump writing speed is relatively quick:
– ~10s per GB providing physical memory is available
– Otherwise dictated by disk write speed
– Can be minimized by good configuration of system:
● Memory assigned to file caching
● Use of ramdisks
■ System dumps compress very well:
– Usually to 10% of original size

IBM Monitoring and Diagnostic Tools for Java – Memory Analyzer
■ Originally a tool for identifying memory leaks in Java applications
■ Works with dumps from all major Java vendors: IBM and HotSpot (Oracle, HP-UX, Mac OSX)
■ Provides a number of functions:
– Memory leak detection and footprint analysis
– Memory efficiency analysis
– Thread stack traces with Object references
–
■ Provides extension points to write analysis plugins
– Used by IBM Extensions for Memory Analyzer

IBM Extensions for Memory Analyzer (IEMA)
■ Goal:
build product and subject matter expert knowledge into Memory Analyzer to allow
users and support to visualize and debug IBM SWG product applications
■ Development project that extends Memory Analyzer
– Individual plug-ins for each product
–
■ Adoption from multiple products:
– Java SE
– WebSphere Application Server
– WebSphere ESB
– WebSphere Portal Server
– CICS Transaction Gateway
– 40 query plug-ins/custom reports created
– many more name resolvers (object annotations)
11 © 2010 IBM Corporation

IBM Extensions for Memory Analyzer – WebSphere Application Server
■ WebSphere Application Server Overview:
– Install information
● Version, build level, install directory
– Server information
● Server, Node and Cell name, startup and dump times
– Applications status
● Applications status: started/stopped/leaking
– Memory breakdown
● Applications, HTTP Sessions, WAS Runtime, SIB etc
■ Thread Pool Analysis:
– All threads
● WebSphere thread pools and Java thread groups
– Thread pool configuration
● Min and max sizes, ability to grow
– Current thread count
– Presence of hung threads
■ Plus a number of other reports, including:
– Application Class Loader Leaks and Hung Thread Analysis
– HA Manager status, HTTP Session details, Web Containter Analysis, Cache status, etc

Memory Analyzer Extension Points
■ See:http://wiki.eclipse.org/MemoryAnalyzer/Extending_Memory_Analyzer
■ org.eclipse.mat.api.nameResolver
– Provides readable description of an object in some of MAT's view (like toString())
– Annotates object types with detail about an object
● eg. hostname and port for a Socket object
■ org.eclipse.mat.report.query
– Adds a menu item that executes your plug-in code
– Creates report style output in any form you want: Tree, Text, HTML, Pie Charts, etc.

Memory Analyzer API
■ ISnapshot – Represent one dump
– Each object and class has a unique Integer ID. Most methods will return an int or array of ints.
Then you can call snapshot.getObject with the int to get an IObject representing the item
–
– getGCRoots – List of all GC roots
– getClasses – List of all classes (or search by name)
– getInboundRefererIds – List of incoming references
– getOutboundReferentIds – List of outgoing references
– getHeapSize – Shallow heap size of the object
– getRetainedHeapSize – Retained heap size of the object
–
■ IObject – represents an item in the heap
– getObjectAddress – This is the address of the object in the Java heap
– getClazz – Get the class of an object
– getUsedHeapSize/getRetainedHeapSize – Same as ISnapshot
– getDisplayName – The class, address, and name resolver
– resolveValue – For an IBM system dump or HPROF dump, given a name of a field, find the
object representing that field. This identifier can have periods which separate going down that
tree of items.

Get the String name of the object field
Memory Analyzer Extension Points: nameResolver
import org.eclipse.mat.SnapshotException;
import org.eclipse.mat.snapshot.extension.IClassSpecificNameResolver;
import org.eclipse.mat.snapshot.extension.Subject;
import org.eclipse.mat.snapshot.model.IObject;
@Subject("com.ibm.ws.webcontainer.webapp.WebAppImpl")
public class WebApp implements IClassSpecificNameResolver {
public String resolve(IObject object){
String result = null;
try {
Lookup field value
IObject appName = (IObject)object.resolveValue("applicationName");
IObject contextPath = (IObject)object.resolveValue("contextPath");
if (appName != null && contextPath != null) {
result = appName.getClassSpecificName() + " at " +
contextPath.getClassSpecificName();
}
} catch (SnapshotException e) {
e.printStackTrace();
}
return result;
}
}
■ NameResolvers “annotate” objects with a line of text:
Select object type to “annotate”
Return String with text to annotate

Memory Analyzer API
■ IResult – What a query returns
– TextResult – Plain text or HTML content
– ObjectListResult – Grid of results with in/outbound refs
– SectionSpec – Separate results into sections
● Add QuerySpecs
– PieFactory...build() - Generate a pie chart
– ListResult – Table of items
– CompositeResult – Display results in separate tabs

Find the Hung Threads
Memory Analyzer Extensions: query
import org.eclipse.mat.SnapshotException;
import org.eclipse.mat.snapshot.extension.IClassSpecificNameResolver;
import org.eclipse.mat.snapshot.extension.Subject;
import org.eclipse.mat.snapshot.model.IObject;
Set name of query in menu
@Name("Hung Thread Stacks")
@Category(WASHelper.WAS_CATEGORY + WASHelper.HUNGTHREADS_SUBCATEGORY)
@Help("List any threads that may be hung.nn")
public class HungThreadsList extends BasePlugin {
public IResult execute(IProgressListener listener) throws Exception {
int[] hungThreadIds = HungThreads.getHungThreadIds(snapshot);
return new HungThreadsListOutgoingReferencesTree(
hungThreadIds, listener);
}
■ Queries produce “reports” available from pull down menus:
Set category in menu
Return “IResult” report

Finding the Hung Threads
public static int[] getHungThreadIds(ISnapshot snapshot)
throws SnapshotException {
int[] threadIds = findObjects(snapshot, "java.lang.Thread");
List<Integer> hungThreads = new ArrayList<Integer>();
for (int thr : threadIds) {
IObject thrObj = snapshot.getObject(thr);
if (isThreadHung(thrObj)) {
hungThreads.add(thr);
}
}
int[] hungThreadObjectIds = getArrayIntegers(hungThreads);
return hungThreadObjectIds;
}
■ findObjects() makes it easy to find all objects of a give type
■ Returns a list of ObjectIds (ints) for the hung threads
Find all Threads
Check if the Thread is hung
Return the hung threads

Checking if the Thread is hung
public static boolean isThreadHung(IObject thread) throws SnapshotException {
boolean result = false;
boolean isHung = MATHelper.resolveValueBool(thread, "isHung");
if (isHung != null && isHung.booleanValue()) {
result = true;
}
return result;
}
■ Hung thread status found by checking boolean field “isHung”
Get isHung field
Check value of field
Return field value

Displaying the Hung Threads
public class HungThreadsListOutgoingReferencesTree extends OutgoingReferencesTree {
private static final int sortIndex = COLUMN_INDEX_MAX_BUILTIN + 2;
public HungThreadsListOutgoingReferencesTree(ISnapshot snapshot,
int[] roots, IProgressListener listener)
throws SnapshotException {
super(snapshot, roots, listener);
}
■ Customization used to add additional columns to output report:
– Last Dispatch Time and Estimated Time Hung
■ Displays a Tree view of outgoing references, from the hung threads
Customized
OutgoingReferencesTree

IBM Monitoring and Diagnostic Tools for Java – IDDE
■ Interactive Diagnostic Data Explorer (IDDE)
■ Provides a visual representation of your set of dump files
■ An investigation log editor in which you run commands. This editor provides the following features:
– Command assistance
– Auto-completion of text
– The ability to save commands and output, which you can then send to other people
– Highlighted text and flagging of issues
– The ability to add your own comments
■ An outline view to help you navigate the commands and output in the log editor
■ The ability to create your own commands, as plug-ins, and import plug-ins created by others

Debugging Java from Dumps

More Related Content

What's hot (20)

Similar to Debugging Java from Dumps (20)

More from Chris Bailey (20)

Recently uploaded (20)

Debugging Java from Dumps