Apache Karaf Features for OSGi Deployment
Apache Karaf Features for OSGi Deployment
When getting started with Apache Karaf, many things appear to happen by magic. This article tries to remove the mystery from one of them.
Join the DZone community and get the full member experience.Join For Free
The CMS developers love. Open Source, API-first and Enterprise-grade. Try BloomReach CMS for free.
I've worked on multiple teams that have successfully used Apache Karaf as an OSGi container. In addition to the usual need to understand OSGi bundles and class resolution in the context of multiple class loaders (which deserves at least one article all by itself), Karaf adds the concept of a "feature" on top of OSGi bundles. While there is detailed documentation on features, I've found that they can be confusing for new users, especially when they are resolved from Maven, since there seems to be so much behind the scenes magic involved.
For everything that Karaf adds to OSGi, the basic unit of installation in the OSGi container is still the bundle. A bundle is a Java Archive (JAR) with some special information in its manifest that identifies it, gives a version, and specifies dependencies. When an OSGi container adds a bundle, it goes through a resolution process to make sure that the bundle's dependencies are met (and that it does not conflict with other installed bundles). However, that resolution process does not include any ability to obtain any dependencies; it just checks to see if they are available and delays or prevents the bundle from starting if a required dependency is missing.
So it is necessary for a user of an OSGi container to identify all of the bundles that need to be available in the container. Unfortunately, this can get unwieldy, as the number of bundles can easily reach the dozens or hundreds.
To address this, Karaf introduces the concept of a feature. A feature is just a group of bundles that should all be installed together. A feature also gets a name and a version. Features are specified in an XML format. For example, here's part of the XML definition for the "core" feature for Apache Camel:
<feature name='camel-core' version='2.16.0' resolver='(obr)' start-level='50'> <feature version='2.5.0'>xml-specs-api</feature> <bundle>mvn:org.apache.camel/camel-core/2.16.0</bundle> <bundle>mvn:org.apache.camel/camel-catalog/2.16.0</bundle> ... </feature>
Notice that a feature can specify a dependency on another feature as well as listing bundles.
To simplify things a little bit, Karaf allows multiple features to be specified in a single XML file. That way, a product like Camel with lots of different pieces can have a single location where all the available features are listed, and users can pick and choose the features they need.
This XML file is called a "feature repository". I've seen this cause confusion with new users of Karaf, because we're used to a repository being a collection of files on a remote system. (For example, this is how Artifactory uses the term.)
A feature repository is just a list of features wrapped in an outer XML element:
<features xmlns="http://karaf.apache.org/xmlns/features/v1.3.0"> <feature name="feature1" version="1.0.0"> ... </feature> <feature name="feature2" version="1.1.0"> ... </feature> </features>
Even though multiple features can be listed in a single feature repository, we install using the name of the individual features inside.
Providing a way to specify dependencies is nice, because it helps to modularize the selection of all the bundles we need. (As one small example, we can shift to new a version of a feature dependency and get the transitive dependencies of that feature without having to figure out which versions have changed.)
But in order to be really useful, there needs to be a way to resolve those dependencies. Otherwise we would still have a manual process of putting all the bundles together so the OSGi container could resolve them. Karaf, recognizing that Maven repositories have become very common for storing dependencies, supports using Maven to find features repositories and bundles.
As a result, to fetch down a feature repository XML file for Apache Camel and get access to all its features, we can just do:
karaf@root()> feature:repo-add mvn:org.apache.camel.karaf/apache-camel/2.18.0/xml/features
Now we have all the Camel features available:
karaf@root()> feature:list | grep camel-core camel-core | 2.18.0 | | Uninstalled | camel-2.18.0 |
And we can install:
karaf@root()> feature:install camel-core karaf@root()> list | grep camel 52 | Active | 50 | 2.18.0 | camel-catalog 53 | Active | 50 | 2.18.0 | camel-commands-core 54 | Active | 50 | 2.18.0 | camel-core 55 | Active | 50 | 2.18.0 | camel-karaf-commands
What is shown in the list is the set of bundles with 'camel' in the name that were installed as part of the feature. There is no rule that the names of bundles or features have to match.
There is much more to know about using Karaf features, but I'm already running long for a single article. Next I'll walk through how to configure Karaf to load features at startup, and control how it uses Maven.
Opinions expressed by DZone contributors are their own.