To install the TraP you must:

  1. Install and configure a suitable back-end database;
  2. Optionally, install and configure MongoDB as a pixel store;
  3. Install the core pipeline dependencies;
  4. Build and install the TraP itself.

Some details on each of these steps is provided below.

Note, though, that the overall procedure is complex, and correctly configuring all the various components can be difficult. It is possible instead to use Vagrant to quickly and easily set up a virtual machine which provides a fully configured and working ready-to-go pipeline and supporting tools. This is a quick and easy way to get up and running for testing purposes or when simplicity is preferable to ultimately high performance. Refer to the Vagrant TraP repository for details and instructions.

Back-end Database

The TraP supports two database management systems for use as the pipeline database: MonetDB and PostgreSQL. Both are available for common operating systems and package managers: pick one and install it.

A complete description of configuring your database management system is beyond the scope of this guide: refer to its documentation for details. Some brief notes on things to look out for follow.


Ensure that the access rights to your server are set appropriately, for example to trust connections from whichever machine(s) will be used to run the TraP. This is done by editing pg_hba.conf.


To be able to administer MonetDB databases, you need to be a member of the monetdb group.

To issue remote management commands, such as database creation, you need to both enable this functionality and set a passphrase:

monetdbd set control=yes ${dbfarm}
monetdbd set passphrase=${myphassphrase} ${dbfarm}

Pixel Store

Optionally, the pixel contents of all images processed (but not the metadata) can be saved to a MongoDB database for future reference (e.g. via the Banana web interface). This naturally requires that a MongoDB daemon is installed and configured to accept requests from TraP clients.

Core Dependencies

To build the TraP, you will need:

In addition to the above, to run the TraP you will need:

To work with the pipeline database, you will need at least one of:

To work with the pixel store you will need:

Most of these dependencies should be easily satisfied by operating system-level package management or through the Python Package Index, and we strongly suggest you take advantage of that convenience rather than building everything from source. The most notable exceptions are pyrap and casacore: these are not commonly packaged in mainstream distributions. They can be compiled from source, or users of Ubuntu-based distributions might find the SKA South Africa Launchpad page useful.


See also the note on casacore measures data, which can often cause confusing errors if out-of-date or incorrectly configured.

Build and Install

Once all dependencies have been satisfied, building should be straightforward:

$ git clone
$ cd tkp
$ python install

Following installation, including setting up and configuring the database, follow the test procedure to ensure that everything is working and ready for use.

Distributed processing via Celery

If you wish to run a TraP job across multiple machines, you may optionally also install a Celery broker (at least version 3.0); see the Celery website for further details of the Celery package.

Multiple different options for Celery brokers are available; refer to the Celery documentation for details. We have had success with RabbitMQ.

This functionality is currently not well supported and should be considered experimental.