We recommend running long builds from the bio/scripts/project_build script. This is a perl program that reads a project.xml file and loads each source in turn. This makes multiple calls to ant to avoid memory problems encountered when running many Java task sequentially from ant. It also has the option of dumping the production database during the build and recovering from these dumps in case of problems.
Note
This script requires the Expect and XML::Parser::PerlSAX Text::Glob perl modules - install with: sudo cpan -i XML::Parser::PerlSAX Expect Text::Glob
Run the build script from the mine directory:
$ ../bio/scripts/project_build -b -v server_name /some/dump/location/dump_file_prefix
The server_name is hostname of the machine where the pg_dump command should be run. If you are running project_build on the same machine as PostgreSQL then you should specify localhost as the server name. If the PostgreSQL server is on a remote machine, give its hostname. In that case the script will try to run pg_dump on the remote machine using ssh. This makes dumping a little faster and allows for the case where /some/dump/location/dump_file_prefix is only visible on the remote machine.
Dumps are performed when a source has dump=true in its project.xml definition:
<source name="uniprot-malaria" type="uniprot" dump="true">
<property name="uniprot.organisms" value="36329"/>
<property name="src.data.dir" location="/data/flyminebuild/malaria/uniprot/7.7/36329"/>
</source>
In this example, the dump will be made immediately after the uniprot-malaria source has been ‘’successfully’’ merged.
Once all sources are integrated project_build will run any post-processing steps (also configured in the project.xml).
It is also possible to run individual integrate and post-process steps separately, see below.
The project_build script accepts the following flags:
| -v | is passed to ant to make it run in verbose mode, ant output can be seen in pbuild.log |
| -l | attempt to restart by reading the last dump file (see note below) |
| -b | run build-db before starting build and drop any existing backup databases (created when using the -t flag) |
| -V | set the release number to pass to ant (as -Drelease=release_number) [1]. |
Dump files take the name dump_file_prefix.final.
Running project_build with ‘’‘-l‘’’ will reload the latest dump (if any) with dump_file_prefix and restart the build from that point.
Note
You must use the full path to the dump file, e.g. /some/dump/location/dump_file_prefix
Before starting the build process you will need to set up the appropriate properties and then initialise your database with this command:
$ cd MINE_NAME/dbmodel/
$ ant build-db
Warning
Running the build-db target will drop the current database and create a new, blank database.
To run a data source, run this command in the MINE_NAME/integrate/ directory, specifying the source name (as it appears in project.xml):
$ ant -v -Dsource=malaria-gff
Most sources have multiple stages in retrieving data, to run just one stage use:
$ ant -v -Dsource=malaria-gff -Daction=[retrieve|load]
The stages are:
Most sources do not have a translate step so retrieve will write to the common-tgt-items database.
| [1] | this allows multiple sets of properties file. eg. passing ‘’‘-V test’’ causes the build system to look for MINE_NAME.properties.test instead of the default file MINE_NAME.properties |