Building and Installing ======================= For the impatient: $ ./configure $ make $ make check $ make install Alternatively (and recommended), use a remote build directory: $ mkdir build $ cd build $ ../configure $ make $ make check $ make install If you can't seem to find configure then you probably downloaded the source, you'll need to "bootstrap" your development environment with: $ ./bootstrap There are numerous configuration options that might be applicable to you. Check them out via: $ ./configure --help The specific options that might be relevant include: --disable-java don't build Java bindings --disable-python don't build Python bindings --disable-webui don't build webui (requires Python 2.6) --disable-optimize don't try to compile with optimizations --without-included-zookeeper excludes building and using the included ZooKeeper package in lieu of a system installed version (note, however, that no attempt is made to find the package and explicitly setting CPPFLAGS and LDFLAGS as appropriate may be necessary) And the specific environment variables that might be relevant are: JAVA_HOME location of Java Development Kit (JDK) JAVA_CPPFLAGS preprocessor flags for JDK JAVA_LDFLAGS linker flags for JDK PYTHON which Python interpreter to use Java ==== During configuraiton we attempt to determine a suitable location for JDK so that we can build JNI based libraries. If we can not do so, you may need to try setting JAVA_HOME, JAVA_CPPFLAGS, or JAVA_LDFLAGS manually. The resulting libmesos.a/so/dylib will include the Java bindings (JNI) if they are enabled (which they are by default, and can be disabled with the '--disable-java' configuration option). The Mesos JAR can be found in [build]/src (where [build] represents a remote build directory if any). Python ====== We build a Python egg that can be included in the PYTHONPATH when running Python based frameworks. The Python egg is built in [build]/src/python/dist (where [build] represents a remote build directory if any). Running Mesos before Installing =============================== After running 'make' but before running 'make install' you can run any of the binaries but certain environment variables must be set up first (see [build]/bin/mesos-build-env.sh.in). It's advisable to run the binaries by using the wrapper scripts provided in [build]/bin. Note that these scripts also have a GDB variant that starts the binaries in GDB when you want to debug. Examples and Tests ================== After running 'make check' there will be a collection of examples and tests that get built. Check out [build]/src/examples and [build]/src/examples/java and [build]/src/examples/python. In addition, see [build]/bin for scripts that can run the tests and also launch the tests in GDB. Hadoop ====== Included in the distribution is a runnable tutorial on using Hadoop on Mesos (./hadoop/TUTORIAL.sh). Try it out! You can also "build" a self-contained distribution of Hadoop with the necessary Mesos components by doing 'make hadoop-0.20.205.0' or 'make hadoop-0.20.2-cdh3u3' from within [build]/hadoop (this uses the tutorial mentioned above). Installing ========== You can install the binaries and libraries with: $ make install Note that this may require superuser privileges. You can change the installation prefix during configure by using the --prefix option, for example: $ ./configure --prefix=/home/benh/mesos Then installation will occur in /home/benh/mesos. Running a Mesos Cluster ========================== Included in an installation are some "deploy" scripts that can be used to start and stop a Mesos cluster. The scripts can only be used after installation (see the previous section on how to install without superuser privileges), and are installed by default in [prefix]/sbin (unless you set --sbindir or --exec-prefix during configuration). Note that the scripts expect the same installation directories on each host in the cluster. The scripts DO NOT copy any Mesos binaries, libraries, configuration files (or themselves) to other hosts in the cluster (it's expected that you'll do the installation yourself). The scripts assume the presence of a 'masters' and a 'slaves' file in [prefix]/var/mesos/deploy by default (unless you set --localstatedir during configuration). These files should contain a list of hostnames/IPs to be used (one per line) for running master(s) and slave(s) respectively. Configuration options for either the mesos-master or mesos-slave are expected to be set on each host before trying to start a cluster. The deploy scripts assume the configuration file will be located at [prefix]/var/mesos/conf/mesos.conf on all hosts. Note that you'll need to at least set the '--master' option on slave hosts! Options specific to running the "deploy" scripts (e.g., whether to use sudo, what ssh options to use when connecting to each host) can be set in [prefix]/var/mesos/deploy/mesos-deploy-env.sh. Once you've populated the 'masters' and 'slaves' file and set up configuration options on each host as appropriate you can start a cluster by doing: $ [prefix]/sbin/mesos-start-cluster.sh And you can stop the cluster by doing: $ [prefix]/sbin/mesos-stop-cluster.sh ====================================================================== As always, if you run into any issues please email: mesos-dev@incubator.apache.org (Remember to include as much debug information as possible.)