blob: b82f9d3859bbb3c8200e3b5555c68b5821654d5a [file] [log] [blame]
<?xml version="1.0"?>
<document>
<copyright>
Copyright 1999-2004 The Apache Software Foundation
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
</copyright>
<properties>
<title>Components</title>
<author email="cmanolache@yahoo.com">Costin Manolache</author>
<author email="jfrederic.clere@fujitsu-siemens.com">Jean-Frederic Clere</author>
<author email="yoavs@apache.org">Yoav Shapira</author>
<date>$Date$</date>
</properties>
<section name="Intro">
<p>
Each component instance has a name, that is used for configuration and at runtime.
Each component has a number of configurable properties. The following rules are used:
<ul>
<li>The name is composed from the type and a local part, separated with a ':' ( example: channel.unixsocket:/tmp/jk.socket ) </li>
<li>The 'type' consist of '.' and ascii characters. It is mapped to a JMX 'domain'. </li>
<li>The local part consists of ascii characters and .:/;
<p>
Note that '=,' are not currently allowed - a future version may support the jmx syntax by using quotes to separate the local part from the property and value ( in .properties mode we must use '=' to separate the value from type, local name and property name ).
</p>
</li>
<li>The property is a simple name, with no dots. </li>
<li>A simple form of substitution is used in values, where $(property) will be replaced with a previously defined setting. If the property has ':' in it, it'll take the value from the object, if not it'll take the value from a global map.
</li>
</ul>
</p>
</section>
<section name="Common properties">
<p>
Common properties for all components
</p>
<p>
<table>
<tr>
<th>Property name</th>
<th>Default</th>
<th>Description</th>
</tr>
<tr>
<td>disabled</td>
<td>0 (false)</td>
<td>"disabled" state for the component, 1=true 0=false</td>
</tr>
<tr>
<td>debug</td>
<td>0 (false)</td>
<td>Debug level for the component, 0=disabled, 1..10 enabled. Higher levels
generate more debug information.</td>
</tr>
<tr>
<td>ver (short for, and formerly known as, version)</td>
<td>0</td>
<td>'Generation' of the component config. Important for runtime reconfiguration.
If you edit the config file or set the shmem properties, you need to also
upgrade the version of the modified component. The config layer will detect
the change and call the setter method.</td>
</tr>
</table>
</p>
</section>
<section name="workerEnv">
<p>This component represent the core jk2, it has the default logger for all other components. Is the central controller, it controls global properties
and provides access to all other objects</p>
<p>
<table>
<tr>
<th>Property name</th>
<th>Default</th>
<th>Description</th>
</tr>
<tr>
<td>logger</td>
<td>logger</td>
<td>Default loger used by jk2 components, can be changed in the config file, normally it defaults to "logger" the Alias for the default logger for the Server/platform.</td>
</tr>
<tr>
<td>sslEnable</td>
<td>1 (true)</td>
<td>Enable handling of SSL</td>
</tr>
<tr>
<td>timing</td>
<td>0</td>
<td>Will jk2 get request timing (needs APR?)</td>
</tr>
<tr>
<td>forwardKeySize</td>
<td>not set</td>
<td>Enable filling of javax.servlet.request.key_size</td>
</tr>
<tr>
<td>forwardURICompat</td>
<td>set</td>
<td>Pass the URI untouched.</td>
</tr>
<tr>
<td>forwardURICompatUnparsed</td>
<td>not set</td>
<td>Parse the URI until the '?'.</td>
</tr>
<tr>
<td>forwardURIEscaped</td>
<td>not set</td>
<td>Pass the URI escaped.</td>
</tr>
<tr>
<td>noRecoveryIfRequestSent</td>
<td>set</td>
<td>No recovery in LB mode if a Tomcat allready received the request.</td>
</tr>
<tr>
<td>noRecoveryIfHeaderSent</td>
<td>set</td>
<td>No recovery in LB mode if a Tomcat allready start to send reply to client.</td>
</tr>
</table>
</p>
<p>Only one of the forwardURI option could be used it replaces the default value.</p>
</section>
<section name="config">
<p>The config component, hold the detail of the conifg system, such config file name, create global defines</p>
<p>
<table>
<tr>
<th>Property name</th>
<th>Default</th>
<th>Description</th>
</tr>
<tr>
<td>file</td>
<td>${serverRoot}/conf/workers2.properties</td>
<td>Location of the workers2.properties file</td>
</tr>
<tr>
<td>debug</td>
<td>0</td>
<td>Set the debug level of the config component</td>
</tr>
<tr>
<td>debugEnv</td>
<td>0</td>
<td>Set the debug level of the hidden env component </td>
</tr>
</table>
</p>
</section>
<section name="uriMap"/>
<section name="shm">
<p>Shared memory descriptor</p>
<p>
<table>
<tr>
<th>Property name</th>
<th>Default</th>
<th>Description</th>
</tr>
<tr>
<td>file</td>
<td>No default value</td>
<td>Name of the file that will be mmapped to use as shared memory, If set to 'anonymous' use the anonymous shered memory</td>
</tr>
<tr>
<td>size</td>
<td>No default value</td>
<td>Deprecated. Size of the file.</td>
</tr>
<tr>
<td>slots</td>
<td>256</td>
<td>Number of shared memory slots. Set to the number of child processes</td>
</tr>
<tr>
<td>useMemory</td>
<td>0</td>
<td>Use process memory instead of shared memory. Useful for single child mpm's</td>
</tr>
</table>
</p>
</section>
<section name="uri">
<p>A uri stores a pattern that is used
to match requests to workers, and asociated properties</p>
<p>If the uri name doesn't have a slash then it is considered as a virtual host
directive. Uri name can have a virtual host name and(or) port associated with. Format
of such a name is <b>hostname</b> or <b>hostname:port</b> where hostname
is virtual server name and the port is vitual server port number. The port number
is used only for the non default server ports.</p>
<p>
Special case is a default server named as <b>[uri:*]</b> that is used when the virtual
host cannot be found inside the configuration. All the uri directives not containing
host name belongs to this default server making global mappings.
</p>
<p>
Addition wild char scheme id <b>[uri:*:port]</b> that is used when you wish to
match any virtual host having specified (non-default) port number, like [uri:*:443].
This will map all the virtual hosts no mather what is their name but that have port number 443.
</p>
<p>
The order how the host names are resolved is :
<ul>
<li>Exact host name and optional non default port number</li>
<li>Alias matching host name and port number</li>
<li>*:port if the port is other then default</li>
<li>Default server</li>
</ul>
</p>
<p>
<table>
<tr>
<th>Property name</th>
<th>Default</th>
<th>Description</th>
</tr>
<tr>
<td>group</td>
<td>lb:lb (The default loadbalancer)</td>
<td>Name of the tomcat group or worker that will process the request corresponding to the uri. This used
to be called 'worker'</td>
</tr>
<tr>
<td>context</td>
<td/>
<td>the context path for this uri component (webapp style).</td>
</tr>
<tr>
<td>servlet</td>
<td/>
<td>Servlet path for this mapping</td>
</tr>
<tr>
<td>alias</td>
<td/>
<td>server name alias. This setting should only be used for
host uris like [uri:myHost:myPort] ( i.e. no /) </td>
</tr>
</table>
</p>
</section>
<section name="vm">
<p>Represents the JVM when used as inprocess container
</p>
<p>
<table>
<tr>
<th>Property name</th>
<th>Default</th>
<th>Description</th>
</tr>
<tr>
<td>JVM</td>
<td>(Autoguess)</td>
<td>JVM to use for this vm</td>
</tr>
<tr>
<td>OPT</td>
<td/>
<td>Option to pass to this vm, this is a multivalued property</td>
</tr>
<tr>
<td>classpath</td>
<td/>
<td>-Djava.class.path 0ption to pass to this vm, this is a multivalued property</td>
</tr>
</table>
</p>
</section>
<section name="channels">
<p>A channel represents a transport protocol, connecting 2
sides for RPC communication. The most common and standard is the tcp socket.
Other important channels are unix socket and jni</p>
<subsection name="channel.un">
<p>
AF_UNIX socket. Only on UNIX like platform. These sockets are faster
than "normal" sockets but they are limited to the machine.
</p>
<p>
<table>
<tr>
<th>Property name</th>
<th>Default</th>
<th>Description</th>
</tr>
<tr>
<td>file</td>
<td>Name of socket</td>
<td>Name of the socket file (It is created by the Tomcat ChannelUn)</td>
</tr>
</table>
</p>
</subsection>
<subsection name="channel.socket">
<p>
Defines a communication transport to a remote Servlet Engine. </p>
<p>The name of the channels should be: channel.socket:HOST:PORT, where HOST and PORT are the
tomcat Ajp location. You could use other names and explicitely set HOST and PORT, but this
is discouraged. In most cases, you don't need to set any other config - just add a line like
[channel.socket:localhost:8009] and all other things will have good defaults.
</p>
<p>
NB: Starting with JK2 2.0.4, APR is mandatory and the channel.socket use APR (previously
called channel.apr).
</p>
<p>
Tomcat Engine must be set with jvmRoute="HOST:PORT", to match the default tomcatId of the channel.
</p>
<p>
<table>
<tr>
<th>Property name</th>
<th>Default</th>
<th>Description</th>
</tr>
<tr>
<td>port</td>
<td>extracted from the component name</td>
<td>Port where Tomcat is listening. It is automatically extracted from the name - you shouldn't have to specify it explicitely.</td>
</tr>
<tr>
<td>host</td>
<td>extracted from the component name</td>
<td>Remote host. You should use the name, no need to override it</td>
</tr>
<tr>
<td>graceful</td>
<td>0</td>
<td>If 1, only requests for existing sessions will be forwarded</td>
</tr>
<tr>
<td>keepalive</td>
<td>0</td>
<td>? </td>
</tr>
<tr>
<td>timeout</td>
<td>0 (infinite)</td>
<td>Socket timeout for sending and receiving</td>
</tr>
<tr>
<td>ndelay</td>
<td>0</td>
<td>If set to 1 Disables the Nagle algorithm for send coalescing</td>
</tr>
<tr>
<td>lb_factor</td>
<td>1</td>
<td>Load balancing factor to use. The lower the lb_factor the more often that tomcat will get requests but see
"level" below.</td>
</tr>
<tr>
<td>level</td>
<td>1</td>
<td>Worker Priority. Valid values are 0-3. The functioning workers with the lowest level
will be checked for the lowest lb_value, and if found will be run. The upper level workers are
only checked upon failure of all workers on ALL of the levels below them. This is very
useful for implementing failover withing a cluster. You could set the tomcat server local
on the same machine as the apache instance to level 0 and all of the other workers to level 1.
This would cause apache to only use the external tomcats when the local tomcat is down.</td>
</tr>
<tr>
<td>group</td>
<td>lb</td>
<td>loadbalanced groups to which this channel and the associated worker will be added, multivalued. You need to set it only if you have an advanced setup with multiple clusters.</td>
</tr>
<tr>
<td>tomcatId</td>
<td>Automatically set to the localname ( host:port )</td>
<td>Must match the JVM route on tomcat the server.xml Engine element, for load balancing</td>
</tr>
<tr>
<td>route</td>
<td>Automatically set to the localname ( host:port )</td>
<td>Same as tomcatId</td>
</tr>
</table>
</p>
</subsection>
<subsection name="channel.jni">
<p>The jni channel, used if tomcat is started inprocess</p>
</subsection>
</section>
<section name="workers">
<p>
For the moment 4 worker types are supported: worker.jni,ajp13,status,lb.
</p>
<subsection name="worker.jni">
<p>worker used in inprocess, holds the details of the Tomcat class to startup, and parameters to pass</p>
<p>There are two predefined jni workers <b>onStartup</b> and <b>onShutdown</b>. Those two workers are executed
during startup and shutdown phase of the connector. Both must exists in the configuration to be able to start
and shutdown Tomcat.
</p>
<p>
<table>
<tr>
<th>Property name</th>
<th>Default</th>
<th>Description</th>
</tr>
<tr>
<td>class</td>
<td>org/apache/jk/apr/TomcatStarter</td>
<td>class that holds the main method called to start tomcat</td>
</tr>
<tr>
<td>ARG</td>
<td/>
<td>Arguments to pass to main method when called</td>
</tr>
<tr>
<td>stdout</td>
<td>NULL</td>
<td>file to redirect Standard output from the java process</td>
</tr>
<tr>
<td>stderr</td>
<td>NULL</td>
<td>file to redirect Standard output from the java process </td>
</tr>
</table>
</p>
</subsection>
<subsection name="ajp13">
<p>Default worker. If a property is in both the worker and the channel, you only need to define it in one place.
The channel passes down the properties to its worker. (at least in the ajp13-channel.socket linkage)</p>
<p>
<table>
<tr>
<th>Property name</th>
<th>Default</th>
<th>Description</th>
</tr>
<tr>
<td>secretkey</td>
<td>NULL</td>
<td>
<b>Magic:</b> The secret key will be set automatically on the associated
worker.
</td>
</tr>
<tr>
<td>tomcatId</td>
<td>Automatically set to the localname ( host:port )</td>
<td>Must match the JVM route on the tomcat server.xml Engine element, for load balancing</td>
</tr>
<tr>
<td>route</td>
<td>Automatically set to the localname ( host:port )</td>
<td>Same as tomcatId</td>
</tr>
<tr>
<td>group</td>
<td>lb</td>
<td>loadbalanced groups to which this channel and the associated worker will be added, multivalued. You need to set it only if you have an advanced setup with multiple clusters.</td>
</tr>
<tr>
<td>lb_factor</td>
<td>1</td>
<td>Load balancing factor to use. The lower the lb_factor the more often that tomcat will get requests but see
"level" below.</td>
</tr>
<tr>
<td>level</td>
<td>1</td>
<td>Worker Priority. Valid values are 0-3. The functioning workers with the lowest level
will be checked for the lowest lb_value, and if found will be run. The upper level workers are
only checked upon failure of all workers on ALL of the levels below them. This is very
useful for implementing failover withing a cluster. You could set the tomcat server local
on the same machine as the apache instance to level 0 and all of the other workers to level 1.
This would cause apache to only use the external tomcats when the local tomcat is down.</td>
</tr>
<tr>
<td>channel</td>
<td/>
<td>Communication channel used by the worker. Use the full name of the channel (everything between the []'s).</td>
</tr>
<tr>
<td>max_connections</td>
<td>0 (unlimited)</td>
<td>Maximum number of currently used endpoints.
If the specified number is reached then the load balancer has the chance
to try another worker. This is very useful in situations when having multiple
servers and you wish to finer grade the lb_factor.
</td>
</tr>
<tr>
<td>connectTimeout</td>
<td>0 (no timeout)</td>
<td>With such timeout set, the web-server will send a CPING request just after physical connect to the remote Tomcat
and will wait for a CPONG reply for the connectTimeout milliseconds, a guarantee that the remote Tomcat is not hang.
Side effect, this round trip add a little delay at connection time and require a recent AJP13 implementation,
with support for CPING/CPONG command.
</td>
</tr>
<tr>
<td>replyTimeout</td>
<td>0 (no timeout)</td>
<td>With such timeout set, the web-server will wait for Tomcat reply to a forwarded request for replyTimeout milliseconds.
Another guarantee that the remote Tomcat is not hang.
Warning, if you have 'normal' long running processes on Tomcat side, you shouldn't use this feature to avoid
invalid errors reports or set the timeout accordingly.
</td>
</tr>
<tr>
<td>prepostTimeout</td>
<td>0 (no timeout)</td>
<td>With such timeout set, the web-server will send a CPING request just before forwarding the request to the remote Tomcat
and will wait for a CPONG reply for the prepostTimeout milliseconds, a guarantee that the remote Tomcat is not hang.
Side effect, this round trip add a little delay in forwarding request and require a recent AJP13 implementation,
with support for CPING/CPONG command.
</td>
</tr>
</table>
</p>
</subsection>
<subsection name="status">
<p>An optional worker that outputs (HTML) pages with useful information to monitor JK2.</p>
<p>To create the worker, add a [status:] or [status:'a name'] section.</p>
<p>The status worker needs no 'local' name, although one can be used if desired.</p>
<p>To access the worker, add a [uri] section with the group= assigned to this worker.</p>
<p>The status worker has the following additional properties:
<table>
<tr>
<th>Property name</th>
<th>Default</th>
<th>Description</th>
</tr>
<tr>
<td>styleMode</td>
<td>0 (None)</td>
<td>Defines if or how the status worker will apply a Style Sheet to returned pages.</td>
</tr>
<tr>
<td>stylePath</td>
<td>NULL</td>
<td>For styleMode=2 or 3, where to find the Style Sheet file.</td>
</tr>
</table>
</p>
<p>The syleMode setting accepts the following values:
<table>
<tr>
<th>styleMode value</th>
<th>Description</th>
</tr>
<tr>
<td>0</td>
<td>No Style settings are used. (Default)</td>
</tr>
<tr>
<td>1</td>
<td>Use the built-in Default Style Sheet. It is presented as an 'Internal Style Sheet' to HTML pages.</td>
</tr>
<tr>
<td>2</td>
<td>Use stylePath as a URI context to an external Style Sheet file to be returned by the Web Server.</td>
</tr>
<tr>
<td>3</td>
<td>Use stylePath as a file-system reference to a Style Sheet file. It is presented as an 'Internal Style
Sheet' to HTML pages.</td>
</tr>
</table>
</p>
<p>If the stylePath has not been set, styleMode=2 and 3 are ignored.</p>
<p>The sylePath setting accepts the following values:
<table>
<tr>
<th>styleMode value</th>
<th>stylePath Description</th>
</tr>
<tr>
<td>2</td>
<td>A URI context to a file (e.g /styles/jkstatus.css). It must begin with '/' and include the file name.</td>
</tr>
<tr>
<td>3</td>
<td>A file-system path (suitable for the OS in use) plus file name. Also ${serverRoot}/conf/'file name'
allows the file to be located in Apache's conf directory.</td>
</tr>
</table>
</p>
<p>If the Style file cannot be found, Mode 2 and 3 record Apache log entries.</p>
<p>For styleMode=2 and 3, changes in the Style Sheet are effective on the next access to the status worker.</p>
</subsection>
<subsection name="lb">
<p>Loadbalanced worker</p>
<p>
<table>
<tr>
<th>Property name</th>
<th>Default</th>
<th>Description</th>
</tr>
<tr>
<td>worker</td>
<td/>
<td/>
</tr>
<tr>
<td>noErrorHeader</td>
<td>1 (true)</td>
<td>If set, jk2 won't touch the headers in case of error and will let for example Apache present the ErrorDocument via mod_alias.
</td>
</tr>
<tr>
<td>noWorkerMsg</td>
<td/>
<td/>
</tr>
<tr>
<td>noWorkerCode</td>
<td>503</td>
<td/>
</tr>
<tr>
<td>hwBalanceErr</td>
<td/>
<td/>
</tr>
<tr>
<td>timeout</td>
<td>0 (disabled)</td>
<td>If all the workers are in the error state, probably by Tomcat
refusing any new connections due to the overload, you can set the timeout forcing lb to wait that some
worker becomes available, instead of immediately returning error to the client. This is very useful in
situations with high peek load. The timeout should be set to the maximum application call time, but
not less then 1 second.
</td>
</tr>
<tr>
<td>attempts</td>
<td>3</td>
<td>Number of attempts that lb will try on each worker before
giving up.
</td>
</tr>
<tr>
<td>recovery</td>
<td>60 (seconds)</td>
<td>Time to wait before retrying to see if the worker came out
of the error state.
</td>
</tr>
<tr>
<td>stickySession</td>
<td>1 (true)</td>
<td>Sessions stick to the same worker, 1=true 0=false
</td>
</tr>
</table>
</p>
</subsection>
</section>
<section name="loggers">
<p>Any connector based on jk2, at least has a default logger, that can be reached using the "logger" alias, the logger used is the more appropiate for the plataform/server combination, Apache2 under in any platform has logger.apache2 as default, IIS on his only platform uses logger.win32, and Any apache 1 install uses logger.file as default.., the config file lets you change that defaults, you can end using logger.file in IIs i.e</p>
<p>The properties shared by all loggers are:
<table>
<tr>
<th>Property name</th>
<th>Default</th>
<th>Description</th>
</tr>
<tr>
<td>level</td>
<td>INFO</td>
<td>Text of the log level. Strings supported: EMERG, ERROR, INFO, DEBUG</td>
</tr>
</table>
</p>
<subsection name="logger.file">
<p>
<table>
<tr>
<th>Property name</th>
<th>Default</th>
<th>Description</th>
</tr>
<tr>
<td>file</td>
<td>${serverRoot}/logs/jk2.log</td>
<td>
Log file. XXX you may be able to change this at runtime,
to implement rolling.
</td>
</tr>
</table>
</p>
</subsection>
<subsection name="logger.win32">
<p>logger used in the IIS server by default, it ends at native Application Event Log.</p>
</subsection>
<subsection name="logger.apache2">
<p>Logger used in Apache2 servers, it normally in ends in error.log </p>
</subsection>
</section>
<section name="How Load Balancing Works">
<p>The lb_factor and level properties combine to deliver a flexible static load balancing solution.
The level property is used to create up to four pools over workers in descending priority and lb_factor
is used to weight the workers within a pool. The lower the level the more likely the worker is to be used
and the lower the lb_factor the more likely the worker is to be used. Here is how the algorithm is
currently implemented:</p>
<p>
(Assume that every worker's lb_value is set to their lb_factor)<br/>
<br/><source>
if (loadbalancer has a route) and (stickysession=1){<br/>
worker= loadbanlancer.getWorkerForRoute(route)<br/>
if worker.hasRedirect<br/>
redirect=worker.redirect<br/>
else if !worker.hasError<br/>
return worker<br/>
}<br/>
<br/>
selectedWorker=null<br/>
foreach lb_level {<br/>
foreach worker in the level {<br/>
if worker.isNotWorking<br/>
continue<br/>
if selectedWorker == null<br/>
selectedWorker = worker<br/>
continue<br/>
if worker.lb_value &lt; selectedWorker.lb_value<br/>
selectedWorker = worker<br/>
continue<br/>
}<br/>
if selectedWorker !=null<br/>
break<br/>
}<br/>
<br/>
Reenable workers in error state if the timeout has passed()<br/>
<br/>
if selectedWorker !=null {<br/>
selectedWorker.lb_value += selectedWorker.lb_factor<br/>
if selectedWorker.lb_value > 255 {<br/>
foreach worker in load_balancer.workers[selectedWorker.level]<br/>
worker.lb_value=worker.lb_factor<br/>
}<br/>
}<br/>
return selectedWorker<br/>
}<br/></source>
</p>
</section>
</document>