gdp / README-server-install.md @ master
INSTALLING A GLOBAL DATAPLANE SERVER FROM SOURCE CODE
These instructions are relevant only if you are installing the
server code from the source tree. If you have installed the
server code from a package, or if it is already installed, read
Generally speaking, we are moving toward distributions using Docker images (first choice) and Debian packages. Installations from source should only be used if you are collaborating with us on the base system.
These steps assume you are building and installing the code from the source tree. If you are using a Debian package, see the instructions with it and skip to the next section.
Installing Step by Step
- Run the script
adm/gdp-setup.shto set up your system so that the source code can be compilied. In particular, this installs all the third party packages you should need.
/etc/gdp.conf.shif you want to override any of the defaults.
- Run the script
adm/init-gdp-server.shfrom the root of the GDP source tree. This creates the
gdpuser, creates the system directories, etc. Note that this has certain dependencies on Berkeley. See below for details.
- Compile the source tree using
makefrom the source root. Details can be found in
Install the header files, libraries, and binaries using
make install. You will probably need to be root for this to work. If you changed the setting for
gdp.conf.shyou will also have to set
LOCAL_ROOTon the command line, e.g., if you have set
GDP_ROOTto /usr/local/gdp, use:
make install LOCAL_ROOT=/usr/local/gdp
If you want to change defaults, create
/etc/gdp.conf.sh. This is read as a shell file, and should
consist of NAME=value definitions. There are several parameters
that can be changed, the most useful of which is
There are three cases for
/usrmeans that everything will be installed in the "normal" system directories:
/etc/gdp. Note that files in
/etcare in a subdirectory. This is the default.
/opt/localmeans that everything will be installed in subdirectories of
GDP_ROOTwith the exception of configuration files, which will be stored in
- Otherwise everything is installed in subdirectories of
the indicated directory. For example, if
/home/gdpthen all installation will go into that directory instead of the system directories, i.e.,
Other locations can be set explicitly (rather that deriving from
GDP_ETC: the location for configuration files.
GDP_LOG_DIR: where system logs are stored.
Initializing the Server
adm/init-gdp-server.sh should do most of the steps
necessary for initializing a server. This:
- Creates a user and group named
gdp:gdpif it doesn not exist.
- Creates directories needed by the GDP. See below for details.
- Creates initial parameter files.
This script does have some Berkeley dependencies that may require
editing some files after you have run the script. These will
be located in the directory
gdp) is set to the
list of hosts to search to find a GDP router. This will be
initialized with the names of routers at Berkeley. If you are
running your own router you will probably want to change this.
that your domain name is "eecs.berkeley.edu". You should change
this to match your domain. The convention is to use the
reversed domain name of the host running
.gdplogd appended. For example, if your
server.foo.example.com, the value of that parameter
Proceed to the documenet
doc/gdp-server-admin.md for details about