How to setup a OpenDap server.
Setup a server OS
To setup a centos system you can take a clean computer or a clean virtual machine (using for example vmware). Download centos burn a dvd and insert it into your computer. Follow the installation guide and setup the operating system. Any linux/unix system will be useable. If you want to work with windows you can best use cygwin to install linux compatible software.
Install required libraries & tools
(sudo) yum install subversion
The following libraries are required:
- java (1.6 or higher) follow the centos java guide to install. I noticed that the jpackage distribution isn't updated anymore. Also I had to apply a patch for this error in the java rpm
- ant (http://ant.apache.org) this is used for building java software.
The following need c,c+,fortran (all prefered from the gcc (gcc, g+, gfortran). Make sure you have up to date versions of automake, autoconf, etc... (should be if you use a recent distribution). For these I usually install from source because there might be some options I need later, like --with-hdf, --with-dap or --with-netcdf-4
- hdf5 (1.8.2) Download hdf5 and ./configure && make && make install.
- netcdf (4.0) Download netcdf and ./configure && make && make install.
To install Hyrax follow the Hyrax installation guide.
To configure Hyrax follow the Hyrax configuration guide
There are some alternative opendap servers you could use:
- Thredds is a pure java based implementation. It's easier to setup and has some nice features which makes it easy to work with ensembles. I noticed some errors when querying large datasets if it does not have enough memory. The main difference with the hyrax server is the architecture. The hyrax server has a native (c++) backend which does the data input/output. I think hyrax is more efficient for large datasets but I haven't tested this properly.
- Pydap is a pure python based implementation. I haven't tested this.