Techniques

Nothing ever works exactly as planned and various techniques and tools as well as an understanding of commonly encountered problems are valuable aids to troubleshooting your web applications.

Enable debug logging

Resin provides a number of diagnostic messages using the JDK logging facility. Full debug logging is enabled with an empty name (meaning match all names) and a level of 'all'. Since this will produce many messages, it is useful to put the messages in a a seperate file.

Full debug logging of the server and all applications

The following configuration creates a debug log for each day, and when a problem occurs the debug log is used to help find out what the problem is. Since the log entry is in resin.conf, the log will capture messages from the server itself and all applications. The resulting log messages are found in the file $RESIN_HOME/log/debug.log.

<!-- resin.conf -->
<resin>
  ...
  <log name='' level='finer' path='log/debug.log'
       timestamp="[%H:%M:%S.%s]"
       rollover-period='1D'/>
  ...
</resin>

More specific areas can be targetted with appropriate values for name. The documentation for log provides a table of useful values for name.

Full debug logging of one web application

Often you can narrow down the logging requirements to a single web application. The log entry can be placed in <web-app-root>/WEB-INF/web.xml and only log messages from that web application will go to the log file.

The following configuration creates a debug log for each day in <web-app-root>/WEB-INF/work/debug.log.

<!-- <web-app-root>/WEB-INF/web.xml -->

<web-app>
  ...
  <log name='' level='finer' path='WEB-INF/work/debug.log'
       timestamp="[%H:%M:%S.%s]"
       rollover-period='1D'/>
  ...
</web-app>

Thread dump

If an application seems stuck, or is running out of resources, a thread dump will reveal the state of the server.

Java's thread dumps are a vital tool for server debugging. Because servlets are intrinsically multithreaded, it is very possible to create deadlocks without realizing it, or to have runaway threads that consume resources and cause OutOfMemory exceptions. That's especially true when you start adding third-party software like databases, EJB, and Corba ORBs.

Thread dump using JDK 5 tools

jps and jstack are useful tools included in JDK 5, providing a quick command line method for obtaining stack traces of all current threads.

# jps
12903 Jps
20087 Resin
# jstack 20087
Attaching to process ID 20087, please wait...
Debugger attached successfully.
Client compiler detected.
JVM version is 1.5.0-beta2-b51
Thread 12691: (state = BLOCKED)
 - java.lang.Object.wait(long) (Compiled frame; information may be imprecise)
 - com.caucho.util.ThreadPool.runTasks() @bci=111, line=474 (Compiled frame)
 - com.caucho.util.ThreadPool.run() @bci=85, line=423 (Interpreted frame)
 - java.lang.Thread.run() @bci=11, line=595 (Interpreted frame)


Thread 12689: (state = BLOCKED)
 - java.lang.Object.wait(long) (Compiled frame; information may be imprecise)
 - com.caucho.util.ThreadPool.runTasks() @bci=111, line=474 (Compiled frame)
 - com.caucho.util.ThreadPool.run() @bci=85, line=423 (Interpreted frame)
 - java.lang.Thread.run() @bci=11, line=595 (Interpreted frame)

...

Thread dump by sending a signal

On Unix, "kill -QUIT" will produce a thread dump. On the Linux IBM JDKs, you'll need to signal one of the child processes. On the Linux Sun JDK, you'll need to kill the grandparent process of most of the threads:

unix> ps -ef | grep java
ferg     14243 10792  0 Jan13 pts/0    00:00:00 sh -c /usr/java/bin/java
ferg     14244 14243  0 Jan13 pts/0    00:00:00 java -Dres
ferg     14253 14244  0 Jan13 pts/0    00:00:01 java -Dres
ferg     14254 14253  0 Jan13 pts/0    00:00:00 java -Dres
ferg     14255 14253  0 Jan13 pts/0    00:00:00 java -Dres
ferg     14256 14253  0 Jan13 pts/0    00:00:00 java -Dres
ferg     14257 14253  0 Jan13 pts/0    00:00:00 java -Dres
ferg     14258 14253  0 Jan13 pts/0    00:00:00 java -Dres
unix> kill -QUIT 14244

The Solaris 1.2 JDK has an extra quirk. You need to run Resin in the foreground (i.e. not using "start".) Then you'll need to answer some questions on the console.

Thread dump if signalling doesn't work

You get a thread dump without signalling the process by starting the JVM with some extra arguments to allow a debugger to attach. You can then attach with the debugger at any time to get a thread dump. This technique works on all operating systems.

Unix users (and Cygwin users on Windows) will recognize the opportunity to make a script:

#!/bin/sh
echo -e "suspend\nwhere all\nresume\nquit" | $JAVA_HOME/bin/jdb -connect \
  com.sun.jdi.SocketAttach:hostname=localhost,port=5432

Although no rigorous benchmarking has been done, there appears to be little overhead or performance penalties involved in having the JVM start with the options that allow a debugger to attach.

Understanding the thread dump

In any case, you'll eventually get a trace that looks something like the following (each JDK is slightly different):

Full thread dump:

"tcpConnection-8080-2" daemon waiting on monitor [0xbddff000..0xbddff8c4]
        at java.lang.Object.wait(Native Method)
        at com.caucho.server.TcpServer.accept(TcpServer.java:525)
        at com.caucho.server.TcpConnection.accept(TcpConnection.java:190)
        at com.caucho.server.TcpConnection.run(TcpConnection.java:136)
        at java.lang.Thread.run(Thread.java:484)

"tcpConnection-8080-1" daemon waiting on monitor [0xbdfff000..0xbdfff8c4]
        at java.lang.Object.wait(Native Method)
        at com.caucho.server.TcpServer.accept(TcpServer.java:525)
        at com.caucho.server.TcpConnection.accept(TcpConnection.java:190)
        at com.caucho.server.TcpConnection.run(TcpConnection.java:136)
        at java.lang.Thread.run(Thread.java:484)

"tcpConnection-8080-0" daemon waiting on monitor [0xbe1ff000..0xbe1ff8c4]
        at java.lang.Object.wait(Native Method)
        at com.caucho.server.TcpServer.accept(TcpServer.java:525)
        at com.caucho.server.TcpConnection.accept(TcpConnection.java:190)
        at com.caucho.server.TcpConnection.run(TcpConnection.java:136)
        at java.lang.Thread.run(Thread.java:484)

"tcp-accept-8080" runnable [0xbe7ff000..0xbe7ff8c4]
        at java.net.PlainSocketImpl.socketAccept(Native Method)
        at java.net.PlainSocketImpl.accept(PlainSocketImpl.java:413)
        at java.net.ServerSocket.implAccept(ServerSocket.java:243)
        at java.net.ServerSocket.accept(ServerSocket.java:222)
        at com.caucho.server.TcpServer.run(TcpServer.java:415)
        at java.lang.Thread.run(Thread.java:484)

"resin-cron" daemon waiting on monitor [0xbe9ff000..0xbe9ff8c4]
        at java.lang.Thread.sleep(Native Method)
        at com.caucho.util.Cron$CronThread.run(Cron.java:195)

"resin-alarm" daemon waiting on monitor [0xbebff000..0xbebff8c4]
        at java.lang.Thread.sleep(Native Method)
        at com.caucho.util.Alarm$AlarmThread.run(Alarm.java:268)

"Signal Dispatcher" runnable [0..0]

"Finalizer" daemon waiting on monitor [0xbf3ff000..0xbf3ff8c4]
        at java.lang.Object.wait(Native Method)
        at java.lang.ref.ReferenceQueue.remove(ReferenceQueue.java:108)
        at java.lang.ref.ReferenceQueue.remove(ReferenceQueue.java:123)
        at java.lang.ref.Finalizer$FinalizerThread.run(Finalizer.java:162)

"Reference Handler" daemon waiting on monitor [0xbf5ff000..0xbf5ff8c4]
        at java.lang.Object.wait(Native Method)
        at java.lang.Object.wait(Object.java:420)
        at java.lang.ref.Reference$ReferenceHandler.run(Reference.java:110)

"main" waiting on monitor [0xbfffd000..0xbfffd210]
        at java.lang.Thread.sleep(Native Method)
        at com.caucho.server.http.ResinServer.waitForExit(ResinServer.java:674)
        at com.caucho.server.http.ResinServer.main(ResinServer.java:821)
        at com.caucho.server.http.HttpServer.main(HttpServer.java:95)

There should be one tcp-accept-xxx thread for each <http> and <srun> that Resin's listening for. The tcp-accept-xxx thread should almost always be in socketAccept.

There should be several tcpConnection-xxx-n threads. Each of these is the servlet thread. On a busy server, these can appear anywhere in your code. If several appear in one location, you've likely got some sort of deadlock or at least a slow lock. Idle threads are either in tcpAccept or httpRequest or runnerRequest (for keepalive threads.)

For deadlocks, you'll want to look at the "waiting on monitor" threads and any case where lots of threads are stuck at the same location.

Out of Memory and Garbage Collection

Most memory problems are due to memory leaks in the application program. For example, a cache or a vector that fills up with stale data, or a singleton or static variable which doesn't properly detect a web-app reload. Some more exotic memory issues relate to running out of heap memory or virtual memory when using a large number of threads (> 256).

-verbosegc

-verbosegc is a debugging output from the JVM. It's very convenient for checking basic memory usage and garbage collection time. It's a good idea to use -verbosegc for any production system. When starting Resin, you'll use -J-verbosegc.

[GC 9176K->8647K(9768K), 0.0014790 secs]
[GC 9287K->8668K(9768K), 0.0011120 secs]
[GC 9308K->8668K(9768K), 0.0007810 secs]

The "(9768K)" is the most important number, showing that the maximum allocated heap is about 10 meg. The other numbers show the actual heap used before and after garbage collection.

Check memory usage with a heap dump

If an application gobbles up memory until it throws an OutOfMemory exception, or seems to be spending an excessive amount of time doing garbage collection, a heap dump can help track down the source of the problem.

What you really need to do is get a good CPU and heap profile. The JDK has a simple one (but not very user-friendly), so you don't absolutely need to buy a profiler.

An alternative for simple heap monitoring (garbage collection issues), are the jvmstat tools.

It's a standard java argument, so "java -Xrunhprof:help" will tell you how to start it. For example, you could start Resin (unix or windows) as

> httpd.sh -J-Xrunhprof:heap=sites,cpu=samples

(On Unix, Resin's startup script has a -cpuprof-ascii argument that sets it for you automatically.)

Run a load against it for a decent amount of time (you can even use something like Apache's "ab" for 10 minutes), and then stop the server nicely. i.e. you can't just kill it with ctrl-C, but you need to make it do a clean exit.

Understanding java.hprof.txt heap

Assuming you're being cheap, and using the JDK's heap profiling instead of buying a profiler, you'll need a bit of help interpreting it.

The following is an example of a heap dump running Resin. Here's what you might see skipping down to the "SITES BEGIN" section. (I did mention this is the cheap way of doing things.) As with most profiling, you only want to pay attention to the top 20. Everything else is in the noise. Ignore it.

SITES BEGIN (ordered by live bytes) Tue Jan  9 17:44:33 2001
          percent         live       alloc'ed  stack class
 rank   self  accum    bytes objs   bytes objs trace name
    1 11.87% 11.87%   983520  120 1393320  170  2553 [B
    2  9.89% 21.76%   819600  100 1286772  157  4070 [B
    3  9.09% 30.85%   753756   23 3539376  108  4970 [L<Unknown>;
    4  5.83% 36.68%   483564   59  778620   95  7180 [B
    5  5.74% 42.42%   475368   58  745836   91  7178 [B
    6  4.35% 46.77%   360624   44  696660   85  7182 [B
    7  2.97% 49.74%   245880   30  450780   55  7176 [B
    8  2.37% 52.11%   196704   24  352428   43  7254 [B
    9  1.88% 53.99%   155724   19  262272   32  7174 [B
   10  1.78% 55.77%   147528   18  245880   30  7137 [B
   11  1.53% 57.30%   126988 1063 16973092 129113  3271 [C
   12  1.34% 58.64%   110684 3953 20362832 727244  1213 sun/io/CharToByteISO8859_1
   13  1.25% 59.88%   103320  738  141820 1013  5942 java/lang/Class
   14  1.21% 61.10%   100548   49  221616  108  5003 [L<Unknown>;
   15  1.21% 62.31%   100548   49  221616  108  5005 [L<Unknown>;
   16  1.07% 63.38%    89080 1532 18393580 317347  1340 [B
   17  0.79% 64.18%    65568    8   81960   10  8408 [B
   18  0.79% 64.97%    65552    4   65552    4 27630 [C
   19  0.70% 65.67%    58232   24 1110128  386  5038 [C
   20  0.68% 66.35%    56200  450  116816  980  7186 [C

There are two things to look for. First, if any one class starts chewing up a large number in the "live objs" column, you'll need to take a look. That might be a memory leak.

Second, if some class has a huge number in the "alloc'ed objs" column, you may be wasting lots of garbage collection time that could be easily solved with some caching.

The [C in the class name means a character array. To see what that really means, you'll need to look at the stack trace (3271):

TRACE 3271:
	java/lang/String.<init>(String.java:244)
	com/caucho/util/CharBuffer.close(CharBuffer.java:714)
	com/caucho/vfs/FilesystemPath.normalizePath(FilesystemPath.java:162)
	com/caucho/vfs/FilesystemPath.schemeWalk(FilesystemPath.java:127)

That's part of Resin's VFS code. Maybe in the future it'll make sense to try to reduce that.

You can get a longer trace using the "-prof-depth 10" argument to Resin. (Or the corresponding depth in -Xrunhprof.) That will often give more information.

Understanding java.hprof.txt CPU

The CPU profiling is a little easier to read. On some JDK's you'll need to run it with the JIT disabled.

CPU SAMPLES BEGIN (total = 424614) Tue Jan  9 17:44:33 2001
rank   self  accum   count trace method
   1 21.36% 21.36%   90704  7266 com/caucho/server/http/VirtualHost.logAccess
   2 10.84% 32.20%   46041  7269 java/net/SocketInputStream.socketRead
   3  5.99% 38.19%   25428  1213 java/lang/Class.newInstance0
   4  5.11% 43.31%   21715  7896 com/caucho/util/CharBuffer.toString
   5  4.82% 48.13%   20463  1286 sun/io/CharToByteISO8859_1.convert
   6  3.54% 51.66%   15018  1242 sun/io/CharToByteConverter.<init>
   7  2.68% 54.35%   11388  7241 java/io/PrintWriter.<init>
   8  2.47% 56.82%   10508  7748 com/caucho/server/http/Request.fillCookies
   9  2.27% 59.09%    9650  1214 sun/io/ByteToCharConverter.<init>
  10  1.85% 60.94%    7857  5097 java/lang/String.<init>
  11  1.59% 62.53%    6754  1341 java/lang/String.substring
  12  1.57% 64.10%    6650  1340 java/lang/String.getBytes
  13  0.92% 65.02%    3907  7897 java/lang/String.<init>
  14  0.76% 65.78%    3227  3259 com/caucho/vfs/FilePath.fsWalk
  15  0.75% 66.53%    3195  7895 com/caucho/server/http/Request.fillCookie
  16  0.71% 67.25%    3031  7321 java/lang/String.getBytes
  17  0.71% 67.95%    2996  3270 com/caucho/util/CharBuffer.close
  18  0.68% 68.63%    2892  3271 java/lang/String.<init>
  19  0.66% 69.29%    2782  7318 com/caucho/vfs/FilePath.openWriteImpl
  20  0.61% 69.90%    2604  7320 java/io/FileOutputStream.<init>

You should only pay attention to the top 20 or so. You'll probably need to ignore a few of the top 10, because they're just waiting for a user response. The SocketInputStream.socketRead is an example.

TRACE 7266:
	com/caucho/server/http/VirtualHost.logAccess(VirtualHost.java:487)
	com/caucho/server/http/Application.logAccess(Application.java:1846)
	com/caucho/server/http/Response.finish(Response.java:1345)
	com/caucho/server/http/Request.finish(Request.java:416)

Monitor garbage collection

Run Resin with the additional argument -Xloggc:gc.log where "gc.log" is the name of the log file which will be created in the root Resin directory, e.g. /resin/gc.log.

Once the server has run under load for a while, or when problems start appearing, look in the gc.log file and search for "Full". They will happen fairly infrequently at first, and then towards the end of the file, they will become progressively more frequent until they are happening almost continuously (with little or no change in the size of the heap). Note that the "timestamp" in the first column is the number of seconds since the process started.

Garbage collection logging will have a small imapct on performance but is invaluable in diagnosing garbe collection related problems.

Increase heap memory

Clean up the classpath

Starting Resin with an empty CLASSPATH environmental variable is a first step to eliminating these classes.

win> set CLASSPATH=
win> bin/httpd.exe

unix.sh> export CLASSPATH=""
unix.sh> bin/httpd.sh

If you have manually placed any jars in $RESIN_HOME/lib or you jdk directories, those are possible conflicts as well.

If a new version of Resin s installed overtop of an older version of Resin (i.e. in the same directory), some old jars might be left. It is best to give each version it's own directory.

If the RESIN_HOME environment variable is not set, Resin may be picking up an older version.

You can start httpd with the -verbose option to see the CLASSPATH that is being used when Resin is started.

Monitor the HTTP communication

To monitor just the HTTP headers, the following debug logging can be enabled in $RESIN_HOME/resin.conf:

<resin xmlns="http://caucho.com/ns/resin">

  ...

  <log name='com.caucho.server.http' level='finer' path='log/http.log'/>
  <log name='com.caucho.server.connection' level='finer' path='log/http.log'/>

  ...

</resin>

Intercepting and monitoring the raw data passed between a web browser and Resin can provide valuable insight. This raw data reveals the headers and content that are submitted by the browser, and the headers and content that are returned by Resin.

The Apache Axis jar includes a utility "tcpmon" that can be used to intercept and monitor the communication between a browser and Resin. Using tcpmon, you specify a "listen port" and a "target host" and a "target port". For example, if you usually have Resin running on port 8080, you can start tcpmon with a "listen port" os 9090, and a target host of localhost and a target port of 8080. Now you will use a url http://localhost:9090 in your browser. This causes your browser to use tcpmon. tcpmon will log the request, pass it through to Resin (on port 8080), and log the response before returning it to the browser.

Use an external compiler

Resin's default is to use the `internal' compiler because it is the most likely to be available. Sometimes the internal compiler will start to cause errors, causing exceptions or simply hanging and taking up a thread. The solution is to change the compiler to `javac' in resin.conf:

  <javac compiler="javac" args=""/>

Jikes is another option for the compiler, as discussed in the documentation for javac.

Increase stack size to avoid thread limitations

Each thread is allocated a stack, if the size of the stack is too large, then you may run out of memory when a higher number of threads is reached. Techniques to adress this are discussed in JVM Tuning.

Use netstat to obtain an OS level view of current TCP/IP port usage

The netstat utility is useful for obtaining information from the operating system about the current state of network usage.

unix$ netstat -anp
win> netstat -an

The -a indicates that both listening (waiting for a connection) and non-listening (active) sockets should be displayed. The -n causes the display of port numbers (i.e. 80) instead of port names (i.e. http). The -p causes the display of the process that is using the socket. Since the windows netstat is different than the Unix one, -p is not used on Windows.

Connection state

The connection state is probably the most important information. The man page for netstat has a full description of each state that is displayed.

"LISTEN" or "LISTENING" is an indication that there is a process waiting for connections on the socket.

"TIME_WAIT" indicates that the socket is waiting after close to handle packets still in the network. After a connection is closed, the socket is kept open by the operating system for a short period of time. Even after the close is done, there may be some stray packets on the network that are destined for the connection. The idea of TIME_WAIT is that it keeps the socket open long enough to catch those stray packets, so that they don't get erronously delivered to a new connection opened on the same socket.

Port usage

If Resin is indicating that it cannot bind to a port, it may be an indication that another process is using the port. netstat may help determine the process that is using the port.

Since netstat produces a lot of information, it is helpful to filter the output to limit it to the port of interest. The following example is for port 80: