Migrating from Resin 3.0 to Resin 4.0

From Resin 3.0

Revision as of 00:01, 13 November 2009 by Emil (Talk | contribs)
Jump to: navigation, search

Contents

Preliminaries

Before installation, there are several changes that you'll need to know in order to migrate to Resin 4.0:

  1. The required JDK version is now Java 1.6. As of this writing, Java 1.5 and earlier have reached the end of service life. Make sure you use the JDK -- the JRE is not sufficient.
  2. The build process for compiling the Resin JNI and installing Resin have changed. Before Resin 4.0, you would install Resin by unpacking the distribution directly in the filesystem. Resin 4.0's configure script now creates makefiles that install all the necessary files in standard locations for Linux/Unix. There is also a .deb package for Ubuntu available. Windows installation is unchanged.
  3. The main configuration file has been renamed from resin.conf to resin.xml.

As a general recommendation for migration, we suggest starting with a clean installation of Resin 4.0, including using the sample configuration. By beginning with the sample configuration and modifying it to match your deployment setup, you'll be much more likely to configure the server correctly. This approach is usually much easier than starting with your existing configuration and modifying to make it compatible with the new version.

Migrating from Resin 3.0

Starting Resin

One of the major changes from Resin 3.0 to Resin 4.0 is the introduction of the Resin Watchdog which changes the way in which Resin is started and managed during runtime. The httpd.sh and httpd.exe executables no longer exist. The Watchdog documentation linked above has in-depth details, but essentially Resin now runs with two Java processes, the main Resin process and the Watchdog. This architecture introduces many advantages for management and reliability, but changes the way that Resin is started and how the Java process is managed.

Example Resin 4.0 command line

java -jar $RESIN_HOME/lib/resin.jar -conf /etc/resin/resin.xml -root-directory=/var/www -server app-a start

JVM arguments

JVM arguments for the main Resin process are now set within the resin.xml file instead of being passed via a script. Any JVM arguments on the command line actually go to the Watchdog process. JVM arguments can be set per server in a cluster, but in general you can set them in a <server-default> tag to apply to all servers.

For example, if you wanted to set the maximum heap size of all the Resin processes to 1GB:

 <server-default>
   ...
   <jvm-arg>-Xmx1g</jvm-arg>

Linux/Unix

It is possible to run Resin on the console if you like for debugging or other purposes using the "console" command (instead of the "start" command as above). Note however that if you intend to run Resin as a long running process, you should not run Resin in console mode in the background. This will discard all logging sent to stdout (see the logging section below).

If you are using Linux, a sample init.d script for Resin is included with the distribution. The make install command installs it to /etc/init.d/resin. Edit the file to set the name of the server, the location of the resin.xml file, the log directory, and the root directory. Check with your distribution's documentation to see how to add the script to one or more of your runtime levels. The script may be useful for Unix as well.

Unix allows only root to bind to ports below 1024. If you use Resin as your webserver (recommended) and bind to port 80, you'll need to start Resin as root. In Resin 4.0, the Resin process can drop privileges as soon as it's bound to all its ports. You can configure the user that Resin uses in the <server> or <server-default> sections:

 <server-default>
   ...
     <resin:if test="${resin.userName == 'root'}">
       <user-name>www-data</user-name>
       <group-name>www-data</group-name>
     </resin:if>


Windows

On Windows, Resin is now started using the resin.exe program instead of httpd.exe.

Logging

For Resin 4.0, the preferred logging method is to use java.util.logging for all output. Standard output and error are still supported, but should be used for special cases only. The <log> tag of Resin 3.x has been replaced by the <log-handler> tag in Resin 4.0. The <log-handler> can be thought of as a control on output logging, directing log messages from certain sources and of certain levels to a file (or other output stream). Conversely, the <logger> tag controls the source of log messages, controlling the logging level of individual packages. <logger> tags use the same syntax in Resin 4.0 as in 3.x, but <log> tags should be converted to <log-handler> tags. Conversion is fairly straight forward:

 <log name="com.caucho" path="stdout:" timestamp="[%H:%M:%S.%s] "/>

becomes

 <log-handler name="com.caucho" level="all" path="stdout:" timestamp="[%H:%M:%S.%s] "/>

This log-handler will send log messages to standard output if you run Resin on the console, but if you run Resin as a daemon process, the output will be in the directory specified by the --log-directory argument passed to the Watchdog on the command line. For example, if you set the log directory to /var/www/log and ran a server with the id "app-a", the output would be in /var/www/log/jvm-app-a.log. For the "default" server (i.e. the server with an empty string as its id), the filename will be jvm-default.log.

Similarly, the Watchdog's log will go to /var/www/log/watchdog-manager.log.

Converting clustered setups

Clustering in Resin 4.0 is drastically different from that of Resin 3.x. Though conversion is required, the resultant configuration file is more understandable and the capabilities are greater. In Resin 4.0, a cluster is a collection of servers that serve a set of virtual hosts. For each virtual host, there are a number of web-apps. The configuration of Resin 4.0 follows this model, with the resin.xml file having roughly the following outline:

 <resin>
   <cluster id="app">
     <server>...</server>
     <server>...</server>
     
     <host id="">
       <web-app>
         ...
       </web-app>
     </host>
   </cluster>
   
   <cluster id="web">
     ...
   </cluster>
 </resin>

Notice that <server> is now a child of <cluster> instead of the other way around (clusters contain a set of servers) and that the <srun> tag has disappeared. <host> tags are now side-by-side <server> tags as well, instead of the <host> being a child of <server>. Think of servers as representing a Resin instance and all the configuration that goes along with that (e.g. JVM arguments, ports, ip addresses, etc.). Hosts are collections of web-apps. A cluster then organizes servers and hosts together to show which instances serve which content.

To convert an existing clustered setup to the Resin 4.0 syntax,

Filters

Rewrite filter

JMS

Hessian

Deprecated features

Custom types

Use CanDI instead of resin:type

Support for Apache 1.x is deprecated

Migrating from Resin 3.1

Starting Resin 4.0

For the most part, Resin 4.0 is started the same way as Resin 3.1. One exception is if you start Resin attached to the terminal. In general, you should run a Resin server using the "start" command, but

Personal tools