The context
Two nodes of a Hazelcast cluster, each on a discrete subnet so multicast is not suitable nor working for node location.
I should like to employ the most minimal XML configuration file, say hazelcast.xml, to configure Hazelcast to use TCP/IP to connect the two nodes. Ideally a directory of the IP addresses of the two nodes.
The question
The Hazelcast docs do a good job of showing how this can be achieved programatically, and how hazelcast.jar/hazelcast-default.xml holds the (considerable) default configuration.
What is unclear is: is any XML configuration I supply overlaid upon the settings within hazelcast-default.xml - or simply used in its stead?
I have both my answers, and should like to share them
Just like the programatic API, the XML configuration overlays the defaults found in hazelcast.jar/hazelcast-default.xml, consequently ...
I can establish a very simple two-member cluster with this hazelcast.xml in the classpath
<hazelcast>
<network>
<join>
<multicast enabled="false"></multicast>
<tcp-ip enabled="true">
<member>192.168.100.001</member> <!-- server A -->
<member>192.168.102.200</member> <!-- server B, on separate subnet -->
</tcp-ip>
</join>
</network>
</hazelcast>
I'm not familiar with hazelcast.conf files.
Mostly used is XML or Programmatic api. For good examples see:
https://github.com/hazelcast/hazelcast-code-samples/tree/master/network-configuration
Example of programmatic:
public class Main {
public static void main(String[] args){
Config config = new Config();
config.getNetworkConfig().getJoin().getTcpIpConfig().addMember("localhost").setEnabled(true);
config.getNetworkConfig().getJoin().getMulticastConfig().setEnabled(false);
HazelcastInstance hz = Hazelcast.newHazelcastInstance(config);
}
}
--
What is unclear is: is any XML configuration I supply overlaid upon the settings within hazelcast-default.xml - or simply used in its stead?
What do you mean? If you use the programmatic API, the rest is not relevant. If you don't provide an explicit Config object while constructing the HazelcastInstance, a defaulting mechanism is used. And eventually it defaults to hazelcast-default.xml.
Related
Is there any ServiceStack utility that can read from custom config sections. ServiceStack has IAppSettings which makes it easy to read from appSettings in a config file. I am wondering if ServiceStack has a similar utility to read from Custom Config Sections.
Thanks
rudrvij
No there isn't, we actively discourage usage of Config Sections which is XML encumbered, in-flexible, non-portable and non-substitutable.
Our preference (when needed) is instead to embed Complex Type configuration in a single AppSetting Text Value which is transparently supported using the human-friendly JSV Format, e.g:
<appSettings>
<add key="RedisConfig"
value="{Host:localhost,Port:6379,Database:1,Timeout:10000}" />
</appSettings>
Which is supported by every App Settings Provider and can be easily serialized into any POCO with:
RedisConfig redisConf = appSettings.Get<RedisConfig>("RedisConf");
After successfully developing an application with multiple ServiceStack services, we are moving to other testing environments, lots of them due to us running a SAAS model (aka multi-tenant). I'd like to reuse some of the base infrastructure services, primarily Redis and RabbitMQ across a few of these environments.
We're using the IAppSetting interface to pull our configuration from multiple sources into one cohesive settings object at run-time, which is then filtered tier. Since tier drives the configuration per environment it made sense to use Tier to prefix any RabbitMQ messages queues, and prefix any generated cache keys that will be used by Redis, thus providing collision protection per environment.
Below is an example:
RabbitMQ => "Some MQ method here" => "mq:qa1.Outbound.inq"
Redis => "Some Redis method here" => "urn:qa1.somePoco:123"
Here is an example configuration and the various enviroments
<appSettings>
<add key="Tier" value="qa1" />
<!--<add key="Tier" value="dev" />-->
<!--<add key="Tier" value="tst" />-->
<!--<add key="Tier" value="stg" />-->
<!--<add key="Tier" value="prod" />-->
</appSettings>
Thank you,
Stephen
Some examples on how to modify Queue Names are in MqNameTests, e.g:
QueueNames.SetQueuePrefix("site1.");
Will add a prefix on QueueNames, e.g:
site1.mq:TestPrefix.inq
Otherwise you can use QueueNames.ResolveQueueNameFn to have complete control over the MQ name, e.g:
QueueNames.ResolveQueueNameFn = (typeName, suffix) =>
"SITE.{0}{1}".Fmt(typeName, suffix.ToUpper());
QueueNames<TestFilter>.In.Print(); // SITE.TestFilter.INQ
Note the same configuration also needs to applied on the client so the same MQ names gets used.
Configuring ServiceStack with AppSettings
ServiceStack is a code-first framework which means all configuration is done in code, but has a rich and versatile configuration model where you can get the behavior your after by reading App Settings in AppHost.Configure():
QueueNames.SetQueuePrefix(AppSettings.Get("Tier","dev"));
Where if Tier doesn't exist in your Web.config (e.g. in Unit Tests) it will use dev otherwise will use the value in your appSettings:
<appSettings>
<add key="Tier" value="qa1" />
</appSettings>
I am writing a REST web service using Jersey, and I'm trying to write a set of unit tests to test the service using the Jersey Test Framework.
However, I use HTTP Authentication and SecurityContext as part of my web service, and I'm having issues setting up JTF to allow me to test these aspects. I can send authentication information in the request, but how do I configure it to know about the different roles and users I wish to set up?
I'm currently using Jetty (via JettyTestContainerFactory), but can switch to different test containers if needed.
The specific configuration I am trying to achieve is two roles, and four users with the combinations of those possible roles (e.g. No roles, role a, role b, roles a and b). The web service will handle giving access to different URLs, so that doesn't need to be specified in the configuration.
I have done this by implementing my own Jetty Test container similar to the one provided by Jersey. We use an embedded Jetty for testing our application in development normally and by creating our own test container based on that embedded Jetty it loads our web application as it would if it was started by a Java main process.
We use a custom Jetty Security Handler configured in a jetty-env.xml file which the embedded Jetty uses to configure the security.
<Set name="securityHandler">
<New class="com.example.DevelopmentSecurityHandler">
<Set name="loginService">
<New class="com.example.DevelopmentLoginService">
<Set name="name">LocalRealm</Set>
<Set name="config">src/main/webapp/WEB-INF/users.properties</Set>
<Call name="start" />
</New>
</Set>
<Set name="authenticator">
<New class="com.example.DevelopmentAuthenticator"></New>
</Set>
<Set name="checkWelcomeFiles">true</Set>
</New>
</Set>
That Jetty env file is loaded by embedded Jetty:
XmlConfiguration configuration = null;
if (jettyEnvFile.exists()) {
try {
configuration = new XmlConfiguration(jettyEnvFile.toURI().toURL());
} catch (Exception e) {
throw new ProcessingException(String.format("Exception loading jetty config from %s", jettyEnvFile));
}
} else {
LOG.warn("No jetty-env.xml found.");
}
The users.properties file referenced in that xml is a simple user to role mapping e.g.
USERNAME=PASSWORD,ROLE_NAME1,ROLE_NAME2
Depending how you configure your Jetty security this may or may not work for you. You can also configure this programmatically, there's lots of examples of embedded Jetty here. The SecuredHelloHandler.java example there could be a good start for you.
For the test container you can basically start by copying org.glassfish.jersey.test.jetty.JettyTestContainerFactory and org.glassfish.jersey.jetty.JettyHttpContainerFactory essentially changing the
public static Server createServer(final URI uri, final SslContextFactory sslContextFactory, final JettyHttpContainer handler, final boolean start)
method to create your version of an embedded Jetty server with security configured however you require.
I am having trouble with configuring a webapp in Tomcat 7. In my WAR file, there is a properties file myApp/WEB-INF/classes/myProps.props, and it contains environment-specific properites. I am trying to override that configuration file on the server, so that the same WAR file will deploy to multiple environments.
I heard there was a way to do this using replacement config files in tomcat/conf/Catalina/myApp. This is the method I am having trouble figuring out.
Also, myApp.war is one of many running on the same Tomcat server, and it does not run as localhost. I want to be able to solve this problem for several of the webapps.
Server version: Apache Tomcat/7.0.23
Server built: Nov 20 2011 07:36:25
Server number: 7.0.23.0
OS Name: Linux
Your tomcat/conf/Catalina/<host> can contain context descriptors that let you configure lots of things including defining "environment entries", which are accessible from Java via JNDI. There are lots of ways to go about using it. Personally, I set an environment entry which is the file system path to my properties file. My app is built to check for this entry, and if it doesn't exist, look for the file on the classpath instead. That way, in dev, we have the dev properties right there on the classpath, but when we build and deploy, we point it to an external file.
There's good documentation for configuring a context on the Tomcat website. See the Defining a Context section on details of how to create the file and where to put it.
As an example, if your host is named myHost and your app is a war file named myApp.war in the webapps directory, then you could create tomcat/conf/Catalina/myHost/myApp.xml with this content:
<Context>
<Environment name="configurationPath" value="/home/tomcat/myApp.properties" type="java.lang.String"/>
</Context>
Then from your code, you'd do a JNDI lookup on java:comp/env/configurationPath (95% certainty here) to get that string value.
I like .properties files instead of
JNDI - why build complex object during program configuration instead of initialization time?
system properties - you can't separately configure several instances of same WAR in single Tomcat
context parameters - they accessible only in javax.servlet.Filter, javax.servlet.ServletContextListener which my be inconvenient
Tomcat 7 Context hold Loader element. According to docs deployment descriptor (what in <Context> tag) can be placed in:
$CATALINA_BASE/conf/server.xml - bad - require server restarts in order to reread config
$CATALINA_BASE/conf/context.xml - bad - shared across all applications
$CATALINA_BASE/work/$APP.war:/META-INF/context.xml - bad - require repackaging in order to change config
$CATALINA_BASE/work/[enginename]/[hostname]/$APP/META-INF/context.xml - nice, but see last option!!
$CATALINA_BASE/webapps/$APP/META-INF/context.xml - nice, but see last option!!
$CATALINA_BASE/conf/[enginename]/[hostname]/$APP.xml - best - completely out of application and automatically scanned for changes!!!
Context can hold custom Loader org.apache.catalina.loader.VirtualWebappLoader (available in modern Tomcat 7, you can add own separate classpath to your .properties), and Parameter (accessed via FilterConfig.getServletContext().getInitParameter(name)) and Environment (accessed via new InitialContext().lookup("java:comp/env").lookup("name")):
<Context docBase="${basedir}/src/main/webapp"
reloadable="true">
<!-- http://tomcat.apache.org/tomcat-7.0-doc/config/context.html -->
<Resources className="org.apache.naming.resources.VirtualDirContext"
extraResourcePaths="/WEB-INF/classes=${basedir}/target/classes,/WEB-INF/lib=${basedir}/target/${project.build.finalName}/WEB-INF/lib"/>
<Loader className="org.apache.catalina.loader.VirtualWebappLoader"
virtualClasspath="${basedir}/target/classes;${basedir}/target/${project.build.finalName}/WEB-INF/lib"/>
<JarScanner scanAllDirectories="true"/>
<Parameter name="min" value="dev"/>
<Environment name="app.devel.ldap" value="USER" type="java.lang.String" override="true"/>
<Environment name="app.devel.permitAll" value="true" type="java.lang.String" override="true"/>
</Context>
If you use Spring and it's XML config:
<context:property-placeholder location="classpath:app.properties"/>
<bean id="dataSource" class="org.apache.commons.dbcp.BasicDataSource" destroy-method="close">
<property name="driverClassName" value="oracle.jdbc.OracleDriver"/>
<property name="url" value="jdbc:oracle:thin:#${db.host}:${db.port}:${db.user}"/>
<property name="username" value="${db.user}"/>
<property name="password" value="${db.pass}"/>
</bean>
With Spring injecting above properties into bean fields are easy:
#Value("${db.user}") String defaultSchema;
instead of JNDI:
#Inject ApplicationContext context;
Enviroment env = context.getEnvironment();
String defaultSchema = env.getProperty("db.user");
Note also that EL allow this (default values and deep recursive substitution):
#Value('${db.user:testdb}') private String dbUserName;
<property name='username' value='${db.user.${env}}'/>
See also:
Adding a directory to tomcat classpath
Can I create a custom classpath on a per application basis in Tomcat
How to read a properties file outside my webapp context in Tomcat
Configure Tomcat to use properties file to load DB connection information
Should you set up database connection properties in server.xml or context.xml
Externalize Tomcat configuration
NOTE With extending classpath to live directory you also allowed to externilize any other configs, like logging, auth, atc. I externilize logback.xmlin such way.
UPDATE Tomcat 8 change syntax for <Resources> and <Loader> elements, corresponding part now look like:
<Resources>
<PostResources className="org.apache.catalina.webresources.DirResourceSet"
webAppMount="/WEB-INF/classes" base="${basedir}/target/classes" />
<PostResources className="org.apache.catalina.webresources.DirResourceSet"
webAppMount="/WEB-INF/lib" base="${basedir}/target/${project.build.finalName}/WEB-INF/lib" />
</Resources>
You can try to place your configuration (properties file) in Apache Tomcat\lib in JAR file and remove it from the web application. When the Tomcat's class loader won't find your config in webapp it will try to find in "lib" directory. So you can externalize your configuration just moving the config to global lib dir (it's shared among other webapps).
I just added a setenv.bat or setenv.sh script in the bin folder of tomcat. Set the classpath variable like
set CLASSPATH=my-propery-folder
Can I supposed to launch One-Jar using JWS? One-JAR provides custom classloader that knows how to load classes and resources from a jars inside an archive whereas in JWS we need to specify each JAR that is being used in resources.
What I supposed to specify in JNLP if I am trying to launch One-Jar -
<?xml version="1.0" encoding="UTF-8"?>
<jnlp spec="1.0+" codebase="" href="">
<information>
<title>Application</title>
<vendor>ABC</vendor>
</information>
<resources>
<!-- Application Resources -->
<j2se version="1.5+"
href="http://java.sun.com/products/autodl/j2se"/>
<jar href="Application.jar" main="true" download="eager" />
</resources>
<application-desc main-class="com.simontuffs.onejar.Boot">
</application-desc>
<update check="background"/>
<security>
<all-permissions/>
</security>
</jnlp>
My Application JAR that is One-JAR contains -
com\simontuffs\onejar\<contains complied classes> like JarClassLoader$1.class etc
lib/<contains all jar>
OneJar.class
main/<my application's jar>
META-INF\MANIFEST.MF\ <contains >
Manifest-Version: 1.0
Ant-Version: Apache Ant 1.8.1
Created-By: One-Jar 0.96 Ant taskdef
Main-Class: com.simontuffs.onejar.Boot
One-Jar-Main-Class: com.application.main.Entry
Name: com/simontuffs/onejar/Boot$3.class
SHA1-Digest: +LPrezs+UEFcE3J7QvumcAEO8Z0=
Name: OneJar.class
SHA1-Digest: 28pzzJWqEpLk1xFwJ/jsAav8LyI=
Name: lib/commons-io-1.4.jar
SHA1-Digest: qHYtB+ds/eI5Ulel2ke6fB29Pc4=
etc..etc..
How to specify these com/simontuffs/onejar/Boot$3.class paths in resource ?
Part of the problem of supporting tools like One-Jar is that loading Jars within Jars requires a custom class-loader. By default JWS will use the usual JRE class-loaders - which do not support it.
There are two possible ways that you might get around that (that I know of).
1. Get access to the custom class-loader.
Get an instance of the custom loader used by One-Jar and set it as the context class-loader. This would require a trusted app., but I get the impression that your app. is trusted.
I have no idea if the One-Jar API provides this loader for your own app.'s use.
2. Use some options when building.
From Options and VM Properties..
one-jar.jar.names
true: Recorded classes are kept in directories corresponding to their jar names.
false: Recorded classes are flattened into a single directory.
Duplicates are ignored (first wins)
The 2nd option sounds like a standard 'fat jar'. That should work with the conventional (default) class-loader used by JWS.