Compiling on Debian

    ArangoDB v3.7 requires at least g++ 9.2 as compiler. Older versions, especially g++ 7, do not work anymore unlike for ArangoDB v3.6 and older.

    Optional

    Login to your AWS account and launch an instance of Debian Testing. I used an ‘m3.xlarge’ since that has a bunch of cores, more than enough memory, optimized network and the instance store is on SSDs which can be switched to provisioned IOPs.

    The Current AMI ID’s can be found in the Debian Wiki: wiki.debian.org/Cloud/AmazonEC2Image/

    Upgrade to the very latest version

    Optional

    Once your EC2 instance is up, login ad and sudo su to become root.

    First, we remove the backports and change the primary sources.list

    Update and upgrade the system. Make sure you don’t have any broken/unconfigured packages. Sometimes you need to run safe/full upgrade more than once. When you are done, reboot.

    1. apt-get update
    2. apt-get install aptitude
    3. aptitude -y update
    4. aptitude -y safe-upgrade
    5. aptitude -y full-upgrade
    6. reboot

    Install build dependencies

    Mandatory

    Before you can build ArangoDB, you need a few packages pre-installed on your system.

    Login again and install them.

    1. sudo aptitude -y install git-core \
    2.     build-essential \
    3.     libjemalloc-dev \
    4.     cmake \
    5.     python2.7 \
    6. sudo aptitude -y install libldap2-dev # Enterprise Edition only

    Download the latest source using git:

    1. unix> git clone git://github.com/arangodb/arangodb.git

    This will automatically clone the devel branch.

    Note: if you only plan to compile ArangoDB locally and do not want to modify or push any changes, you can speed up cloning substantially by using the --single-branch and --depth parameters for the clone command as follows:

    1. unix> git clone --single-branch --depth 1 git://github.com/arangodb/arangodb.git

    Setup

    Switch into the ArangoDB directory

    1. unix> cd arangodb
    2. unix> mkdir build

    In order to generate the build environment please execute

    1. unix> cmake ..
    1. cmake -DCMAKE_C_COMPILER=/opt/bin/gcc -DCMAKE_CXX_COMPILER=/opt/bin/g++ ..

    If you compile on macOS, you should add the following options to the cmake command:

    1. cmake .. -DOPENSSL_ROOT_DIR=/usr/local/opt/openssl -DCMAKE_OSX_DEPLOYMENT_TARGET=10.11

    If you also plan to make changes to the source code of ArangoDB, you may want to compile with the Debug build type:

    1. cmake .. -DCMAKE_BUILD_TYPE=Debug

    The Debug target enables additional sanity checks etc. which would slow down production binaries. If no build type is specified, ArangoDB will be compiled with build type RelWithDebInfo, which is a compromise between good performance and medium debugging experience.

    Other options valuable for development:

    Needed if you plan to make changes to AQL language (which is implemented using a lexer and parser files in arangod/Aql/grammar.y and arangod/Aql/tokens.ll) or if you want to enable runtime assertions. To use the maintainer mode, your system has to contain the tools FLEX and BISON.

    1. -DUSE_BACKTRACE=On

    Use this option if you want to have C++ stacktraces attached to your exceptions. This can be useful to more quickly locate the place where an exception or an assertion was thrown. Note that this option will slow down the produces binaries a bit and requires building with maintainer mode.

    1. -DUSE_OPTIMIZE_FOR_ARCHITECTURE=On

    This will optimize the binary for the target architecture, potentially enabling more compiler optimizations, but making the resulting binary less portable.

    ArangoDB will then automatically use the configuration from file etc/relative/arangod.conf.

    1. -DUSE_FAILURE_TESTS=On

    This option activates additional code in the server that intentionally makes the server crash or misbehave (e.g. by pretending the system ran out of memory) when certain tests are run. This option is useful for writing tests.

    1. -DUSE_GOOGLE_TESTS=off

    This option disables all unit tests based on Googletest.

    1. -DUSE_JEMALLOC=Off

    By default ArangoDB will be built with a bundled version of the JEMalloc allocator. This however will not work when using runtime analyzers such as ASAN or Valgrind. In order to use these tools for instrumenting an ArangoDB binary, JEMalloc must be turned off during compilation.

    For release builds configure CMake with the following options:

      Shared memory

      Gyp is used as makefile generator by V8. Gyp requires shared memory to be available, which may not if you i.e. compile in a chroot. You can make it available like this:

      1. none /opt/chroots/ubuntu_precise_x64/dev/shm tmpfs rw,nosuid,nodev,noexec 0 2
      2. devpts    /opt/chroots/ubuntu_precise_x64/dev/pts    devpts    gid=5,mode=620    0  0

      Compile the programs (server, client, utilities) by executing

      1. make

      in the build subdirectory. This will compile ArangoDB and create the binary executable in file build/bin/arangod.

      To run multiple jobs in parallel pass , e.g. make -j8.

      Starting and testing

      1. unix> build/bin/arangod -c etc/relative/arangod.conf --server.endpoint tcp://127.0.0.1:8529 /tmp/database-dir

      This will start up the ArangoDB and listen for HTTP requests on port 8529 bound to IP address 127.0.0.1. You should see the startup messages similar to the following:

      If it fails with a message about the database directory, please make sure the database directory you specified exists and can be written into.

      Use your favorite browser to access the URL

      1. http://127.0.0.1:8529/

      This should bring up ArangoDB’s web interface.

      Re-building ArangoDB after an update

      To stay up-to-date with changes made in the main ArangoDB repository, you will need to pull the changes from it and re-run make.

      Normally, this will be as simple as follows:

      1. unix> git pull
      2. unix> (cd build && make)

      From time to time there will be bigger structural changes in ArangoDB, which may render the old Makefiles invalid. Should this be the case and make complains about missing files etc., the following commands should fix it:

      1. unix> rm -rf build/* 
      2. unix> cd build && cmake .. <cmake options go here>
      3. unix> (cd build && make)

      Note that the above commands will run a full rebuild of ArangoDB and all of its third-party components. That will take a while to complete.

      In a local development environment it is not necessary to install ArangoDB somewhere, because it can be started from within the source directory as shown above.

      If there should be the need to install ArangoDB, execute the following command:

      1. (cd build && sudo make install)

      The server will by default be installed in

      1. /usr/local/sbin/arangod

      The configuration file will be installed in

      1. /usr/local/etc/arangodb3/arangod.conf

      The database will be installed in

      1. /usr/local/var/lib/arangodb3

      The ArangoShell will be installed in

      1. /usr/local/bin/arangosh

      You should add an arangodb user and group (as root), plus make sure it owns these directories:

      1. useradd -g arangodb arangodb
      2. chown -R arangodb:arangodb /usr/local/var/lib/arangodb3-apps/
      3. chown -R arangodb:arangodb /tmp/database-dir/

      Note: The installation directory will be different if you use one of the precompiled packages. Please check the default locations of your operating system, e. g. and /var/lib.