Manage Galaxy servers

gravity, galaxy
pip install gravity==0.8.3


gravity - Galaxy Server Administration

A process manager (supervisor) and management tools for Galaxy servers.

Installing this will give you an executables, galaxy which is used to manage Galaxy. A virtualenv will automatically be created for your Galaxy server. It's a good thing.


Python 2.7 is required. Sadly, Python 3 won't work, because supervisor doesn't support it, although work is in progress.

To install:

pip install gravity

To make your life easier, you are encourged to install into a virtualenv, and to make that trivial, you are encouraged to use virtualenv-burrito.



Add this section to your galaxy.ini, reports_wsgi.ini, tool_shed_wsgi.ini to set:

instance_name = string  # override the default auto-generated instance name
galaxy_root = /path     # if galaxy is not at ../ or ./ from the config file
virtualenv = /path      # override default auto-generated virtualenv path (it
                        #   will be created if it does not exist)
log_dir = /path         # where to log galaxy server output. for uwsgi you
                        #   probably want to use uwsgi's logto option though
uwsgi_path = /path      # explicit path to uwsgi, otherwise it will be found on
                        #   $PATH. Or, the special value `install`, which will
                        #   cause it to be installed into service's virtualenv

Potentially useful information, tricks, etc.:

  • Unless you set different state dirs with --state-dir or $GRAVITY_STATE_DIR, there will only be one supervisord for all of your galaxy instances. But they will be separated out by a generated instance_name. You can override this with instance_name in [galaxy:server].
  • To put configs (galaxy configs, galaxy + reports, whatever) into the same instance, set their instance_name to the same string in each config's [galaxy:server]. This puts them into a single supervisor group, which may be what you want, although note that any start/stop/restart/etc. is performed on the entire group, which may not be what you want.
  • The config manager generally views things in terms of config files. If you change the virtualenv or galaxy_root in a config file, it will not change that value for all services in the instance (supervisor group), it will only change it for the services started from that config.
  • Anything you drop in to $GRAVITY_STATE_DIR/supervisor/supervisord.conf.d will be picked up by supervisord on a galaxy supervisorctl update or just galaxy update
  • The job_conf.xml parsed corresponds to the galaxy config, it'll check the path in job_config_file in [app:main] or default to galaxy_root/config/job_conf.xml if that file exists. If handlers in job_conf.xml have a corresponding [server:] in galaxy.ini, they will be started using Paste. If there is not a corresponding [server:] they will be started as a "standalone" server with galaxy_root/lib/galaxy/


Use galaxy -h for help. Subcommands also support -h, e.g. galaxy add -h.


Register a Galaxy, Reports, or Tool Shed server config with the process manager, create a virtualenv, create supervisor configs, and update. Does not start.


List config files registered with the process manager.


List known instances and services.

get /path/to/galaxy.ini

Show stored configuration details for the named config file.

rename /path/to/old.ini /path/to/new.ini

Use this if you move your config.

remove /path/to/galaxy.ini remove instance_name

Deregister a Galaxy et. al. server config., or all configs referencing the supplied instance_name.

start [instance_name]
stop [instance_name]
restart [instance_name]

Roughly what you'd expect. If instance_name isn't provided, perform the operation on all known instances.

If you call start from the root (or from 1 subdirectory deep) of a Galaxy source tree, config/galaxy.ini if it exists, or else config/galaxy.ini.sample will automatically be registered with galaxy add and then galaxy start will start the newly added server.

reload [instance_name]

The same as restart but uWSGI master processes will only receive a SIGHUP so the workers restart but the master stays up.

graceful [instance_name]

The same as reload but Paste servers will be restarted sequentially, and the next one will not be restarted until the previous one is up and accepting requests.


Figure out what has changed in configs, which could be:

  • changes to [galaxy:server]
  • adding or removing [server:] sections
  • adding or removing a [uwsgi] section
  • adding or removing handlers in job_conf.xml

This will perform the operation for all registered configs, which may cause unintended service restarts.

Any needed changes to supervisor configs will be performed and then supervisorctl update will be called. You will need to do a galaxy start after this to start any newly added instances (or possibly even old instances, since adding new programs to a group in supervisor causes the entire group to be stopped).

Update is called automatically for the start, stop, restart, reload, and graceful subcommands.

supervisorctl [subcommand]

Pass through directly to supervisor


Stop supervisord