Tomcat - a minimalistic user's guide

The script which has the most significance for users is tomcat (tomcat.sh/tomcat.bat). The other Tomcat related scripts serve as a simplified single-task oriented entry point to the Tomcat script (set different command line parameters etc.).

A closer look at tomcat.sh/tomcat.bat yields that it performs the following actions:

Starting Tomcat from another directory

By default Tomcat will use TOMCAT_HOME/conf/server.xml for configuration. The default configuration will use TOMCAT_HOME as it's base for the contexts.

You can change this by using the "-f /path/to/server.xml" option, with a different server configuration file and setting the home property of the context manger. You need to set up the required files inside the home:

• A webapps/ directory (if you created one) - all war files will be expanded and all subdirectories added as contexts.
• conf/ directory - you can store a special web.xml and other configuration files.
• logs/ - all logs will go to this directory instead of the main TOMCAT_HOME/logs/.
• work/ - work directories for the contexts.

If the ContextManager.home property in server.xml is relative, it will be relative to the current working directory.

web.xml

A detailed description of web.xml and the web application structure (including directory structure and configuration) is available in chapters 9, 10 and 14 of the Servlet API Spec and we are not going to write about it.

There is however a small Tomcat related "feature" that is related to web.xml. Tomcat lets the user define default web.xml values for all context by putting a default web.xml file in the conf directory. When constructing a new Context, Tomcat uses the default web.xml file as the base configuration and the application specific web.xml (the one located in the application's WEB-INF/web.xml), only overwrite these defaults.

Setting Tomcat to cooperate with the Apache web server

Up until now we have not discussed Tomcat as a server add on, instead we have considered it as a stand-alone container and discussed how it can be used. There are however a few problems with this picture:

1. Tomcat is not as fast as Apache when it comes to static pages.
2. Tomcat is not as configurable as Apache.
3. Tomcat is not as robust as Apache.
4. There are many sites with long time investment in certain web servers, for example, sites that are using CGI scripts/Server API modules/perl/php... We cannot assume that all of them will want to ditch this legacy.

For all these reasons it is recommended that real world sites use a web server, such as Apache, for serving the static content of the site, and use Tomcat as a Servlet/JSP add-on.

We are not going to cover the different configuration in depth, instead we will:

1. Cover the fundamental behavior of the web server.
2. Explain what configuration is needed.
3. Demonstrate this on Apache.

Web server operation

In a nutshell a web server is waiting for client HTTP requests. When these requests arrive the server does whatever is needed to serve the requests by providing the necessary content. Adding a servlet container may somewhat change this behavior. Now the web server needs also to perform the following:

• Load the servlet container adapter library and initialize it (prior to serving requests).
• When a request arrives, it needs to check and see if a certain request belongs to a servlet, if so it needs to let the adapter take the request and handle it.

The adapter on the other hand needs to know what requests it is going to serve, usually based on some pattern in the request URL, and to where to direct these requests.

Things are even more complex when the user wants to set a configuration that uses virtual hosts, or when they want multiple developers to work on the same web server but on different servlet container JVMs. We will cover these two cases in the advanced sections.

The most obvious configuration that one can think of is the identity of the servlet URLs that are under the responsibility of the servlet container. This is clear; someone must know what requests to transmit to the servlet container... Yet there are additional configuration items that we should provide to the web-server/servlet-container combination:

• We also need to provide configuration regarding the available Tomcat processes and on which TCP/IP host/port they are listening.
• We need to tell the web server the location of the adapter library (so it will be able to load it on startup).
• We need to set adapter internal information such as where and how much to log, etc.

All this information must appear either in the web server configuration, or in a private configuration files used by the adapter. The next section will demonstrate how configuration can be implemented on Apache.

Making it on Apache

This section shows you how to configure Apache to work with Tomcat; it tries to provide explanations as well as insight for the configuration directives that you should use. You can find additional information in the jserv install page .

When Tomcat starts up it will automatically generate a configuration file for Apache in TOMCAT_HOME/conf/tomcat-apache.conf. Most of the time you don't need to do anything but include this file (appending "Include TOMCAT_HOME/conf/tomcat-apache.conf") in your httpd.conf. If you have special needs, for example an AJP port other the 8007, you can use this file as a base for your customized configuration and save the results in another file. If you manage the Apache configuration yourself you'll need to update it whenever you add a new context.

Tomcat 3.1: you must restart Tomcat and apache after adding a new context; Apache doesn't support configuration changes without a restart. Also the file TOMCAT_HOME/conf/tomcat-apache.conf is generated when Tomcat starts, so you'll need to start Tomcat before Apache. Tomcat will overwrite TOMCAT_HOME/conf/tomcat-apache.conf each startup so customized configuration should be kept elsewhere.

The Apache-Tomcat configuration uses Apache core configuration directives as well as Jserv unique directives so it may confuse you at first, there are however two things simplifying it:

• In general you can distinguish between the two directive "families" by noting that all the Jserv unique directives start with an "Ajp" prefix.
• The entire Tomcat related configuration is concentrated in a single configuration file named tomcat.conf, or the automatically generated tomcat-apache.conf, so you can look at a single file.

Let's look now at a sample tomcat.conf file.

    
###########################################################
#      A minimalistic Apache-Tomcat Configuration File    #
###########################################################

# Note: this file should be appended or included into your httpd.conf

# (1) Loading the jserv module that serves as Tomcat's apache adapter.
LoadModule jserv_module libexec/mod_jserv.so

# (1a) Module dependent configuration.
<IfModule mod_jserv.c>

# (2) Meaning, Apache will not try to start Tomcat.
ApJServManual on
# (2a) Meaning, secure communication is off
ApJServSecretKey DISABLED
# (2b) Meaning, when virtual hosts are used, copy the mount
# points from the base server
ApJServMountCopy on
# (2c) Log level for the jserv module.
ApJServLogLevel notice

# (3) Meaning, the default communication protocol is ajpv12
ApJServDefaultProtocol ajpv12
# (3a) Default location for the Tomcat connectors.
# Located on the same host and on port 8007
ApJServDefaultHost localhost
ApJServDefaultPort 8007

# (4)
ApJServMount /examples /root
# Full URL mount
# ApJServMount /examples ajpv12://hostname:port/root
</IfModule>
            

As you can see the configuration process was split into 4 steps that will now be explained:

1. In this step we instruct Apache to load the jserv shared-object (or the NT world dll). This is a well known Apache directive. If the loading went well and the module came from a file named mod_jserv.c (1a) we can start with the rest of the Jserv-Tomcat configuration.
2. This step sets various Jserv internal parameters, these parameters:
   • Instruct jserv not to start the Tomcat process. Automatically starting Tomcat is not implemented yet.
   • Disable the secret key challenge/response between Apache and Tomcat. Again, the secret key work is not implemented yet.
   • Instruct jserv to copy the base server mount points (see next section) in case of virtual hosting.
   • Instruct jserv to use the notice log level. Other log levels include emerg, alert, crit, error, warn, info and debug.
3. This step sets the default communication parameters. Basically it says that the default protocol used for the communication is ajpv12 (do not mess with this one) and that the Tomcat process runs on the same machine and listens on port 8007. If you run Tomcat on a machine other than the one used for Apache you should either update your ApJServDefaultHost or use a full URL when mounting contexts (see next). Also, if you configured the Tomcat connectors to use a port other then 8007, you should update your ApJServDefaultPort or use a full URL when mounting contexts.
4. This step mounts a context to Tomcat. Basically it says that all the web server paths that start with /examples go to Tomcat. This ApJServMount example is a rather simple one, in fact ApJServMount can also provide information regarding the communication protocol to be used and the location where the Tomcat process listens, for example:
   ApJServMount /examples ajpv12://hostname:port/root
   mounts the context /examples to a Tomcat process that runs on host "hostname" and listens on port number "port".

Obtaining the Jserv Module (mod_jserv)

As previously stated, we need a web server adapter to sit in Apache and redirect requests to Tomcat. For Apache, this adapter is a slightly modified version of mod_jserv.

You may try to look here and see if there is an already pre-built version of mod_jserv that suites your OS (Usually there is one for NT), however, being a native library you should not expect that yet (too many OS's, not enough developers, life too short...). Moreover, small variations in the way you built Apache/Your specific UNIX® variant may result in dynamic linking errors. You should really try to build mod_jserv for your system (don't panic, it is not that hard!).

Building mod_jserv on UNIX involves the following:

1. Download the source distribution of Tomcat from here.
2. Uncompress it into some directory.
3. Building the module:
   • Change directory into jakarta-tomcat/src/native/apache/jserv/
   • Execute the build command
     apxs -c *.c -o mod_jserv.so
     apxs is part of the Apache distribution and should be located in your APACHE_HOME/bin.

1. Download the source distribution of Tomcat from here.
2. Unzip it into some directory.
3. Building the module:
   • Change directory into jakarta-tomcat\src\native\apache\jserv
   • Add Visual C++ into your environment by executing the script VCVARS32.BAT.
   • Execute the build command
     nmake -f Makefile.win32
     nmake is the Visual C++ make program.

Making Apache serve your context's static files

The previous Apache-Tomcat configuration file was somewhat inefficient, it instructed Apache to send any request for a resource that starts with the /examples prefix to be served by Tomcat. Do we really want that? There are many static files that may be a part of our servlet context (for example images and static HTML), why should Tomcat serve these files?

You may actually have reasons for doing that, for example:

1. You may want to configure Tomcat based security for these resources.
2. You may want to follow users requests for static resources using interceptors.

Having Apache serve the static files requires the following:

1. Instructing Apache to send all servlet requests to Tomcat.
2. Instructing Apache to send all JSP requests to Tomcat.

    
######################################################################
#              Apache-Tomcat Smart Context Redirection               #
######################################################################
LoadModule jserv_module modules/ApacheModuleJServ.dll
<IfModule mod_jserv.c>
ApJServManual on
ApJServDefaultProtocol ajpv12
ApJServSecretKey DISABLED
ApJServMountCopy on
ApJServLogLevel notice

ApJServDefaultHost localhost
ApJServDefaultPort 8007

#
# Mounting a single smart context:
#
# (1) Make Apache know about the context location.
Alias /examples D:\tomcat\webapps\examples
# (2) Optional, customize Apache context service.
<Directory "D:\tomcat\webapps\examples">
    Options Indexes FollowSymLinks
# (2a) No directory indexing for the context root.
#    Options -Indexes
# (2b) Set index.jsp to be the directory index file.
#    DirectoryIndex index.jsp
</Directory>
# (3) Protect the WEB-INF directory from tampering.
<Location /examples/WEB-INF/>
    AllowOverride None
    deny from all
</Location>
# (4) Instructing Apache to send all the .jsp files under the context to the 
# jserv servlet handler. 
<LocationMatch /examples/*.jsp>
    SetHandler jserv-servlet
</LocationMatch>
# (5) Direct known servlet URLs to Tomcat.
ApJServMount /examples/servlet /examples

# (6) Optional, direct servlet only contexts to Tomcat.
ApJServMount /servlet /ROOT
</IfModule>
      

As you can see, the beginning of this configuration file is the same as seen in the previous example. The last step (mounting a context), however, was replaced in a long series of Apache and Ajp configuration directives that will now be explained:

1. This step informs Apache of the context location and aliases it to an Apache virtual directory. This way Apache can serve files from this directory.
2. This optional step instructs Apache more about how to serve the context; for example you can decide if Apache will allow directory indexing (listing) or set a special index file.
3. This step instructs Apache to protect the WEB-INF directory from client access. For security reasons it is important to prevent visitors from viewing the content of the WEB-INF directory, for example web.xml can provide valuable information for intruders. This step blocks the WEB-INF content from visitors.
4. This step instructs Apache to serve all the jsp locations within the context using the jserv servlet handler. The servlet handler redirects these requests based on the default host and port.
5. This step mounts specific servlet URLs to Tomcat. You should note that you should have as many such mount directives as the number of specific servlet URLs.
6. This last step is an example for the addition of servlet only context to Tomcat.

It is easy to see that this configuration is much more complex and error prone then the first example, this however is the price that you should (for now) pay for improved performance.

Configuring for multiple Tomcat JVMs

• When each context serves a different, specific task and runs on a different machine.
• When we want to have multiple developers work on a private Tomcat process but use the same web server.

    
######################################################################
#             Apache-Tomcat with JVM per Context                     #
######################################################################
LoadModule jserv_module modules/ApacheModuleJServ.dll
<IfModule mod_jserv.c>
ApJServManual on
ApJServDefaultProtocol ajpv12
ApJServSecretKey DISABLED
ApJServMountCopy on
ApJServLogLevel notice

ApJServDefaultHost localhost
ApJServDefaultPort 8007

# Mounting the first context.
ApJServMount /joe ajpv12://joe.corp.com:8007/joe

# Mounting the second context.
ApJServMount /bill ajpv12://bill.corp.com:8007/bill
</IfModule>
            

As you can see in the previous example, using several JVMs (even even those that run on different machines) can be accomplished easily by using a full Ajp URL mount. In this full URL we actually specify the host where the Tomcat process is located and it's port.

Had the two Tomcat processes run on the same machine, we would have to configure each of them with different connector ports. For example, assuming that the two JVMs runs on localhost, the Apache-Tomcat configuration should have something that looks like:

    
######################################################################
#      Apache-Tomcat with Same Machine JVM per Context               #
######################################################################
LoadModule jserv_module modules/ApacheModuleJServ.dll
<IfModule mod_jserv.c>
ApJServManual on
ApJServDefaultProtocol ajpv12
ApJServSecretKey DISABLED
ApJServMountCopy on
ApJServLogLevel notice

ApJServDefaultHost localhost
ApJServDefaultPort 8007

# Mounting the first context.
ApJServMount /joe ajpv12://localhost:8007/joe

# Mounting the second context.
ApJServMount /bill ajpv12://localhost:8009/bill
</IfModule>
            

Looking at the above file you can see that we have two explicit Ajp mount points each pointing to a different port on the same machine. It is clear that this configuration requires support from the configuration found in the server.xml files. We will need in these files different <Connector> configurations, for the different Tomcat processes. We will actually need two different server.xml files (lets call them server_joe.xml and server_bill.xml) with different <Connector> entries as shown in the next two samples:

    
<?xml version="1.0" encoding="ISO-8859-1"?>

<Server>
    <!-- Debug low-level events in XmlMapper startup -->
    <xmlmapper:debug level="0" />

    <!--  @@@
        Note, the log files are suffixed with _joe to distinguish
        them from the bill files. 
    -->

    <Logger name="tc_log" 
            path="logs/tomcat_joe.log"
            customOutput="yes" />

    <Logger name="servlet_log" 
            path="logs/servlet_joe.log"
            customOutput="yes" />

    <Logger name="JASPER_LOG" 
        path="logs/jasper_joe.log"
            verbosityLevel = "INFORMATION" />

    <!--  @@@
        Note, the work directory is suffixed with _joe to distinguish
        it from the bill work directory.
    -->
    <ContextManager debug="0" workDir="work_joe" >
        <!-- Context level Setup -->	
        <ContextInterceptor 
		    className="org.apache.tomcat.context.AutoSetup" />
        <ContextInterceptor 
		    className="org.apache.tomcat.context.DefaultCMSetter" />
        <ContextInterceptor 
		    className="org.apache.tomcat.context.WorkDirInterceptor" />
        <ContextInterceptor 
		    className="org.apache.tomcat.context.WebXmlReader" />
        <ContextInterceptor 
		    className="org.apache.tomcat.context.LoadOnStartupInterceptor" />
        <!-- Request processing -->
        <RequestInterceptor 
		    className="org.apache.tomcat.request.SimpleMapper" debug="0" />
        <RequestInterceptor 
		    className="org.apache.tomcat.request.SessionInterceptor" />
        <RequestInterceptor 
		    className="org.apache.tomcat.request.SecurityCheck" />
        <RequestInterceptor 
		    className="org.apache.tomcat.request.FixHeaders" />

        <!-- @@@ This connector uses port number 8007 for it's ajp communication -->
        <Connector 
		    className="org.apache.tomcat.service.SimpleTcpConnector">
            <Parameter 
                name="handler" 
                value="org.apache.tomcat.service.connector.Ajp12ConnectionHandler"/>
            <Parameter name="port" value="8007"/>
        </Connector>

        <!-- @@@ the /jow context -->
        <Context path="/joe" docBase="webapps/joe" debug="0" reloadable="true" > 
        </Context>
    </ContextManager>
</Server>       
            

When looking at server_joe.xml you can see that the <Connector> is configured for port 8007. In server_bill.xml (see next) on the other hand the <Connector> is configured for port 8009.

    
<?xml version="1.0" encoding="ISO-8859-1"?>

<Server>
    <!-- Debug low-level events in XmlMapper startup -->
    <xmlmapper:debug level="0" />

    <!--  @@@
        Note, the log files are suffixed with _bill to distinguish
        them from the joe files. 
    -->

    <Logger name="tc_log" 
            path="logs/tomcat_bill.log"
            customOutput="yes" />

    <Logger name="servlet_log" 
            path="logs/servlet_bill.log"
            customOutput="yes" />

    <Logger name="JASPER_LOG" 
        path="logs/jasper_bill.log"
            verbosityLevel = "INFORMATION" />

    <!--  @@@
        Note, the work directory is suffixed with _bill to distinguish
        it from the joe work directory.
    -->
    <ContextManager debug="0" workDir="work_bill" >
        <!-- Context level Setup -->
        <ContextInterceptor 
		    className="org.apache.tomcat.context.AutoSetup" />
        <ContextInterceptor 
		    className="org.apache.tomcat.context.DefaultCMSetter" />
        <ContextInterceptor 
		    className="org.apache.tomcat.context.WorkDirInterceptor" />
        <ContextInterceptor 
		    className="org.apache.tomcat.context.WebXmlReader" />
        <ContextInterceptor 
		    className="org.apache.tomcat.context.LoadOnStartupInterceptor" />
        <!-- Request processing -->
        <RequestInterceptor 
		    className="org.apache.tomcat.request.SimpleMapper" debug="0" />
        <RequestInterceptor 
		    className="org.apache.tomcat.request.SessionInterceptor" />
        <RequestInterceptor 
		    className="org.apache.tomcat.request.SecurityCheck" />
        <RequestInterceptor 
		    className="org.apache.tomcat.request.FixHeaders" />

        <!-- @@@ This connector uses port number 8009 for it's ajp communication -->
        <Connector className="org.apache.tomcat.service.SimpleTcpConnector">
            <Parameter 
                name="handler" 
                value="org.apache.tomcat.service.connector.Ajp12ConnectionHandler"/>
            <Parameter name="port" value="8009"/>
        </Connector>

        <!-- @@@ the /bill context -->
        <Context path="/bill" docBase="webapps/bill" debug="0" reloadable="true" > 
        </Context>
    </ContextManager>
</Server>       
          

The port configuration is not the only place where the joe and bill configuration differs. We have @@@ marks in the xml files marking the four places where changes had to be made. As you can see, this difference is necessary to avoid the two Tomcat processes from overwriting each other's logs and workspace.

Then we should start the two Tomcat processes using the -f command line option:

Configuring virtual hosting

With the current (Ver3.1) Tomcat, virtual hosting awareness is provided by the web server (Apache/Netscape?). The web server virtual hosting support is used by the Tomcat adapter to redirect requests belonging to a certain virtual host to the JVM(s) containing the contexts of this virtual host. This means that if (for example) we have two virtual hosts (vhost1 and vhost2), we will have two JVMs: one running the contexts of vhost1 and the other running the contexts of vhost2. These JVMs are not aware of each others existence, in fact, they are not aware of the concept of virtual hosting. All the virtual hosting logic is inside the web-server adapter. To make things clearer, lets look at the following sample Apache-Tomcat configuration file:

    
######################################################################
#        Apache Tomcat Virtual Hosts Sample Configuration            #
######################################################################
LoadModule jserv_module modules/ApacheModuleJServ.dll
<IfModule mod_jserv.c>
ApJServManual on
ApJServDefaultProtocol ajpv12
ApJServSecretKey DISABLED
ApJServMountCopy on
ApJServLogLevel notice

ApJServDefaultHost localhost
ApJServDefaultPort 8007

# 1 Creating an Apache virtual host configuration
NameVirtualHost 9.148.16.139

# 2 Mounting the first virtual host
<VirtualHost 9.148.16.139>
ServerName www.vhost1.com
ApJServMount /examples ajpv12://localhost:8007/examples
</VirtualHost>

# 3 Mounting the second virtual host
<VirtualHost 9.148.16.139>
ServerName www.vhost2.com
ApJServMount /examples ajpv12://localhost:8009/examples
</VirtualHost>
</IfModule>
            

As can be seen, steps 1,2 and 3 define two Apache virtual hosts and for each of them, mount the /examples context to a certain ajpv12 URL. Each such ajpv12 URL points to a JVM that contains the virtual host. The configuration of the two JVMs is very similar to the one demonstrated in the previous section, we will need again to use two different server.xml files (one for each virtual host process) and we will need to start the Tomcat processes with the -f command line option. After doing that we will be able to approach Apache, each time with a different host name, and the adapter will redirect us to the appropriate JVM.

The need for improved virtual host support
Having each virtual host implemented by a different JVM is a huge scalability problem. The next versions of Tomcat will make it possible to support several virtual hosts within the same Tomcat JVM.

Real world configuration tips

Modify and customize the batch files

• To set resource limits such as maximum number of descriptors.
• To add new CLASSPATH entries (for example, JDBC drivers).
• To add new PATH/LD_LIBRARY_PATH entries (for example, JDBC drivers DLLs).
• To modify the JVM command line settings.
• Make sure that you are using a specific JVM (out of the two or three JVMs installed on your machine).
• To switch user from root to some other user using the "su" UNIX® command.
• Your pet reason.

Some of these changes can be done without explicit changes to the basic scripts; for example, the Tomcat script can use an environment variable named TOMCAT_OPTS to set extra command line parameters to the JVM (such as memory setting etc.). On UNIX you can also create a file named ".tomcatrc" in your home directory and Tomcat will take environment information such as PATH, JAVA_HOME, TOMCAT_HOME and CLASSPATH from this file. On NT however (and also on UNIX® when the modifications are for something such as the JVM command line) you are forced to rewrite some of the startup script...

Modify the default JVM settings

The default JVM settings in the Tomcat script are very naïve; everything is left for defaults. There are a few things that you should consider to improve your Tomcat performance:

1. Modify your JVM memory configuration. Normally the JVM allocates an initial size for the Java heap and that's it, if you need more then this amount of memory you will not get it.
   Nevertheless, in loaded sites, giving more memory to the JVM improves Tomcat's performance. You should use command line parameters such as -Xms/-Xmx/-ms/-mx to set the minimum/maximum size of the Java heap (and check to see if the performance was improved).
2. Modify your JVM threading configuration. The SUN JDK1.2.2 for Linux comes with support for both, green and native threads. In general native threads are known to provide improved performance for I/O bound applications, green threads on the other hand put less stress on the machine. You should experiment with these two threading models and see which model is better for your site (in general, native threads are better).
3. Select the best JVM for the task. There are several JVM vendors, for example on Linux there are today (21/03/2000) two product level JVMs: the SUN JDK1.2.2 and the IBM JDK1.1.8. If your application does not require a specific JDK functionality, you should benchmark the two JVMs and select the better one. In my (Gal Shachor) internal tests I found the IBM JVM significantly faster than the one created by SUN, you should check that for yourself and make a calculated decision.

Modify your connectors

The Connectors, as configured in Tomcat's default server.xml contains two Connectors configured as in the next server.xml fragment:

        <!-- (1) HTTP Connector for stand-alone operation -->
        <Connector className="org.apache.tomcat.service.SimpleTcpConnector">
            <Parameter
                name="handler"
                value="org.apache.tomcat.service.http.HttpConnectionHandler"/>
            <Parameter
                name="port"
                value="8080"/>
        </Connector>

        <!-- (2) AJPV12 Connector for out-of-processs operation -->
        <Connector className="org.apache.tomcat.service.SimpleTcpConnector">
            <Parameter
                name="handler"
                value="org.apache.tomcat.service.connector.Ajp12ConnectionHandler"/>
            <Parameter
                name="port"
                value="8007"/>
        </Connector>
                

1. Is a Connector that listens on port 8080 for incoming HTTP requests. This connector is needed for stand-alone operation.
2. Is a Connector that listens on port 8007 for incoming AJPV12 requests. This connector is needed for web-server integration (out-of-process servlet integration).

It is clear that a sane Tomcat deployment will use either an out-of-process servlet integration or a stand-alone operation, removing the unnecessary Connector is important.

Use a thread pool in your connectors

Tomcat is a multi-threaded servlet container this means that each request needs to be executed by some thread. By default when a request arrives Tomcat creates a new thread, starts it and has it serve the request. This behavior is problematic for loaded sites because:

• Starting and stopping a thread for every request puts a needless burden on the operating system and the JVM.
• It is hard to limit the resource consumption. If 300 requests arrive concurrently Tomcat will open 300 threads to serve them and allocate all the resources needed to serve all the 300 requests at the same time. This causes Tomcat to allocate much more resources (CPU, Memory, Descriptors...) than it should and it can lead to low performance and even crashes if resources are exhausted.

The solution for these problems is to use a thread pool. Servlet containers that are using a thread pool relieve themselves from directly managing their threads. Instead of allocating new threads; whenever they need a thread they ask for it from the pool, and when they are done, the thread is returned to the pool. The thread pool can now be used to implement sophisticated thread management techniques, such as:

1. Keeping threads "open" and reusing them over and over again. This saves the trouble associated with creating and destructing threads continuously.
   • Usually the administrator can instruct the pool not to keep too many idle threads, freeing them if needed.
2. Setting an upper bound on the number of threads used concurrently. This prevents the resources allocation problem associated with unlimited thread allocation.
   • If the container maxed out to the threads upper limit, and a new request arrives, the new request will have to wait for some other (previous) request to finish and free the thread used to service it.

You can refine the techniques described above in various ways, but these are only refinements. The main contribution of thread pools is thread-reuse and having a concurrency upper bound that limits resource usage.

Using a thread pool in Tomcat is a simple move; all you need to do is to use a PoolTcpConnector in your <Connector> configuration. For example the following server.xml fragment defines ajpv12, pooled Connector:

        <!-- A pooled AJPV12 Connector for out-of-processs operation -->
        <Connector className="org.apache.tomcat.service.PoolTcpConnector">
            <Parameter
                name="handler"
                value="org.apache.tomcat.service.connector.Ajp12ConnectionHandler"/>
            <Parameter
                name="port"
                value="8007"/>
        </Connector>
                

This fragment is very simple and the (default) pool behaviour instructed by it is:

• Upper bound for concurrency of 50 threads.
• When the pool has more then 25 threads standing idle it will start to kill them.
• The pool will start 10 threads on creation, and it will try to keep 10 vacant threads (as long as the upper bound is kept).

The default configuration is suitable for medium load sites with an average of 10-40 concurrent requests. If your site differs you should modify this configuration (for example reduce the upper limit). Configuring the pool can be done through the <Connector> element in server.xml as demonstrated in the next fragment:

        <!-- A pooled AJPV12 Connector for out-of-processs operation -->
        <Connector className="org.apache.tomcat.service.PoolTcpConnector">
            <Parameter
                name="handler"
                value="org.apache.tomcat.service.connector.Ajp12ConnectionHandler"/>
            <Parameter
                name="port"
                value="8007"/>
            <Parameter
                name="max_threads"
                value="30"/>
            <Parameter
                name="max_spare_threads"
                value="20"/>
            <Parameter
                name="min_spare_threads"
                value="5" />
        </Connector>
                

As can be seen the pool has 3 configuration parametes:

• max_threads - defines the upper bound to the for the concurrency, the pool will not create more then this number of threads.
• max_spare_threads - defines the maximum number of threads that the pool will keep idle. If the number of idle threads passes the value of max_spare_threads the pool will kill these threads.
• min_spare_threads - the pool will try to make sure that at any time there is at least this number of idle threads waiting for new requests to arrive. min_spare_threads must be bigger then 0.

You should use the above parameters to adjust the pool behavior to your needs.

Disable servlet auto-reloading

Servlet auto-reloading is really useful for development time. However it is very expensive (in performance degradation terms) and may put your application in strange conflicts when classes that were loaded by a certain classloader cannot co-operate with classes loaded by the current classloader.

So, unless you have a real need for class reloading during your deployment you should turn off the reloadable flag in your contexts.

Start Tomcat from /etc/inittab

Unfortunately the adapters developed for Apache (or for any of the other servers) cannot start Tomcat yet. On UNIX® however, you can use the init table to start Tomcat automatically upon machine startup. FIXME:

Authors

Copyright ©1999 The Apache Software Foundation
Legal Stuff They Make Us Say
Contact Information