An Ansible role for installing and managing Galaxy servers. Despite the name confusion, Galaxy bears no relation to Ansible Galaxy.
This role has the same dependencies as the git module. In addition, Python virtualenv is required (as is pip, but pip will be automatically installed with virtualenv). These can easily be installed via a pre-task in the same play as this role:
- hosts: galaxyservers
pre_tasks:
- name: Install Dependencies
apt:
name: "{{ item }}"
become: yes
when: ansible_os_family == 'Debian'
with_items:
- git
- python-virtualenv
- name: Install Dependencies
yum:
name: "{{ item }}"
become: yes
when: ansible_os_family == 'RedHat'
with_items:
- mercurial
- python-virtualenv
roles:
- galaxyproject.galaxy
If your git
executable is not on $PATH
, you can specify its location with the git_executable
variable. Likewise
with the virtualenv
executable and corresponding galaxy_virtualenv_command
variable.
Not all variables are listed or explained in detail. For additional information about less commonly used variables, see the defaults file.
Many variables control where specific files are placed and where Galaxy writes data. In order to simplify configuration,
you may select a layout with the galaxy_layout
variable. Which layout you choose affects the required variables.
If using any layout other than root-dir
:
galaxy_server_dir
: Filesystem path where the Galaxy server code will be installed (cloned).
If using root-dir
:
galaxy_root
: Filesystem path of the root of a Galaxy deployment, the Galaxy server code will be installed in to a subdirectory of this directory.
Layout control
galaxy_layout
: available layouts can be found in the vars/ subdirectory and possible values include:root-dir
: Everything is laid out in subdirectories underneath a single root directory.opt
: An FHS-conforming layout across multiple directories such as/opt
,/etc/opt
, etc.legacy-improved
: Everything underneath the Galaxy server directory, as withrun.sh
.legacy
: The default layout prior to the existence ofgalaxy_layout
and currently the default so as not to break existing usage of this role.custom
: Reasonable defaults for custom layouts, requires setting a few variables as described in vars/layout-custom.yml
Either the root-dir
or opt
layout is recommended for new Galaxy deployments.
Options below that control individual file or subdirectory placement can still override defaults set by the layout.
New options for Galaxy 18.01 and later
galaxy_config_style
(default:ini-paste
): The type of Galaxy configuration file to write,ini-paste
for the traditional PasteDeploy-style INI file, oryaml
for the YAML format supported by uWSGI.galaxy_app_config_section
(default: depends ongalaxy_config_style
): The config file section under which the Galaxy config should be placed (and the key ingalaxy_config
in which the Galaxy config can be found. Ifgalaxy_config_style
isini-paste
the default isapp:main
. Ifgalaxy_config_style
isyaml
, the default isgalaxy
.galaxy_uwsgi_yaml_parser
(default:internal
): Controls whether theuwsgi
section of the Galaxy config file will be written in uWSGI-style YAML or real YAML. By default, uWSGI's internal YAML parser does not support real YAML. Set tolibyaml
to write real YAML, if you are using uWSGI that has been compiled with libyaml. see YAML Syntax below and unbit/uwsgi#863 for details.
YAML Syntax
To override the default uWSGI configuration, place your uWSGI options under the uwsgi
key in the galaxy_config
dictionary explained below. Note that the defaults are not merged with your config, so you should fully define the
uwsgi
section if you choose to set it.
uWSGI's internal YAML parser expects YAML in an non-standards-conforming syntax:
- Quotes are not supported: all values are read as strings
- Multiple overlapping keys are used to specify multiple instances of an option
- Despite appearing to be a YAML dictionary, the order of the dictionary keys in the config is preserved
Note that regardless of which galaxy_uwsgi_yaml_parser
you use, galaxy_config.uwsgi
should be written in real
YAML because Ansible parses it with libyaml, which does not support the uWSGI internal parser's invalid syntax. This
role will automatically convert the proper YAML to uWSGI-style YAML as necessary.
In order to specify repeititve uWSGI options, the value of the option should be a list. For example, multiple mules for Galaxy handlers are specified like so:
galaxy_config:
uwsgi:
# a few example options
socket: 127.0.0.1:4001
master: true
# define mules
mule:
- lib/galaxy/main.py
- lib/galaxy/main.py
farm: job-handlers:1,2
This role will convert the above to the proper uWSGI-style representation in galaxy.yml
:
uwsgi:
farm: job-handlers:1,2
master: true
mule: lib/galaxy/main.py
mule: lib/galaxy/main.py
socket: 127.0.0.1:4001
The value of galaxy_config.uwsgi
can be either a hash (dictionary) of option/value pairs, or a list of one-item
option/value hashes. In the former case, the options will be written to galaxy.yml in sorted order, since hashes do not
maintain order. In the latter case, order is preserved, and allows for use of uWSGI's configuration
logic, for which the order matters.
When specifying control logic such as if-*
and for
, the value of the option is a list where:
- The first list member is the conditional
- The remaining list members are the option/values in the control block
- Do not specify the closing
endif
orendfor
, the role will do this for you
For example, to use if-exists
as is commonly done for uWSGI Zerg Mode, use:
galaxy_config:
uwsgi:
# a few example options
- socket: 127.0.0.1:4001
- master: true
# define master FIFOs
- master-fifo:
- /srv/galaxy/var/zerg-new.fifo
- /srv/galaxy/var/zerg-run.fifo
- /srv/galaxy/var/zerg-old.fifo
# control block
- if-exists:
- /srv/galaxy/var/zerg-run.fifo
- hook-accepting1-once: writefifo:/srv/galaxy/var/zerg-run.fifo 2q
- hook-accepting1-once: writefifo:/srv/galaxy/var/zerg-new.fifo 1
The role converts this to the following galaxy.yml
contents:
uwsgi:
socket: 127.0.0.1:4001
master: true
master-fifo: /srv/galaxy/var/zerg-new.fifo
master-fifo: /srv/galaxy/var/zerg-run.fifo
master-fifo: /srv/galaxy/var/zerg-old.fifo
if-exists: /srv/galaxy/var/zerg-run.fifo
hook-accepting1-once: writefifo:/srv/galaxy/var/zerg-run.fifo 2q
endif: null
hook-accepting1-once: spinningfifo:/srv/galaxy/var/zerg-new.fifo 1
Feature control
Several variables control which functions this role will perform (all default to yes
except where noted):
galaxy_create_user
(default:no
): Create the Galaxy user. Running as a dedicated user is a best practice, but most production Galaxy instances submitting jobs to a cluster will manage users in a directory service (e.g. LDAP). This option is useful for standalone servers. Requires superuser privileges.galaxy_manage_paths
(default:no
): Create and manage ownership/permissions of configured Galaxy paths. Requires superuser privileges.galaxy_manage_clone
: Clone Galaxy from the source repository and maintain it at a specified version (commit), as well as set up a [virtualenv][virtualenv] from which it can be run.galaxy_manage_static_setup
: Manage "static" Galaxy configuration files - ones which are not modifiable by the Galaxy server itself. At a minimum, this is the primary Galaxy configuration file,galaxy.ini
.galaxy_manage_mutable_setup
: Manage "mutable" Galaxy configuration files - ones which are modifiable by Galaxy (e.g. as you install tools from the Galaxy Tool Shed).galaxy_manage_database
: Upgrade the database schema as necessary, when new schema versions become available.galaxy_fetch_dependencies
: Fetch Galaxy dependent modules to the Galaxy virtualenv.galaxy_build_client
: Build the Galaxy client application (web UI).galaxy_client_make_target
(default:client-production-maps
): Set the client build type. Options include:client
,client-production
andclient-production-maps
. See Galaxy client readme for details.galaxy_manage_errordocs
(default:no
): Install Galaxy-styled 413 and 502 HTTP error documents for nginx. Requires write privileges for the nginx error document directory.
Galaxy code and configuration
Options for configuring Galaxy and controlling which version is installed.
galaxy_config
: The contents of the Galaxy configuration file (galaxy.ini
by default) are controlled by this variable. It is a hash of hashes (or dictionaries) that will be translated in to the configuration file. See the Example Playbooks below for usage.galaxy_config_files
: List of hashes (withsrc
anddest
keys) of files to copy from the control machine.galaxy_config_templates
: List of hashes (withsrc
anddest
keys) of templates to fill from the control machine.galaxy_local_tools
: List of local tool files or directories to copy from the control machine, relative togalaxy_local_tools_src_dir
(default:files/galaxy/tools
in the playbook).galaxy_local_tools_dir
: Directory on the Galaxy server where local tools will be installed.galaxy_dynamic_job_rules
: List of dynamic job rules to copy from the control machine, relative togalaxy_dynamic_job_rules_src_dir
(default:files/galaxy/dynamic_job_rules
in the playbook).galaxy_dynamic_job_rules_dir
(default:{{ galaxy_server_dir }}/lib/galaxy/jobs/rules
): Directory on the Galaxy server where dynamic job rules will be installed. If changed from the default, ensure the directory is on Galaxy's$PYTHONPATH
(e.g. in{{ galaxy_venv_dir }}/lib/python2.7/site-packages
) and configure the dynamic rules plugin injob_conf.xml
accordingly.galaxy_repo
(default:https://github.com/galaxyproject/galaxy.git
): Upstream Git repository from which Galaxy should be cloned.galaxy_commit_id
(default:master
): A commit id, tag, branch, or other valid Git reference that Galaxy should be updated to. Specifying a branch will update to the latest commit on that branch. Using a real commit id is the only way to explicitly lock Galaxy at a specific version.galaxy_force_checkout
(default:no
): Ifyes
, any modified files in the Galaxy repository will be discarded.galaxy_clone_depth
(default:null
): Depth to use when performing git clone. Leave unspecified to clone entire history.
Path configuration
Options for controlling where certain Galaxy components are placed on the filesystem.
galaxy_venv_dir
(default:<galaxy_server_dir>/.venv
): The role will create a [virtualenv][virtualenv] from which Galaxy will run, this controls where the virtualenv will be placed.galaxy_config_dir
(default:<galaxy_server_dir>
): Directory that will be used for "static" configuration files.galaxy_mutable_config_dir
(default:<galaxy_server_dir>
): Directory that will be used for "mutable" configuration files, must be writable by the user running Galaxy.galaxy_mutable_data_dir
(default:<galaxy_server_dir>/database
): Directory that will be used for "mutable" data and caches, must be writable by the user running Galaxy.galaxy_config_file
(default:<galaxy_config_dir>/galaxy.ini
): Galaxy's primary configuration file.
User management and privilege separation
galaxy_separate_privileges
(default:no
): Enable privilege separation mode.galaxy_user
(default: user running ansible): The name of the system user under which Galaxy runs.galaxy_privsep_user
(default:root
): The name of the system user that owns the Galaxy code, config files, and virtualenv (and dependencies therein).galaxy_group
: Common group between the Galaxy user and privilege separation user. If set andgalaxy_manage_paths
is enabled, directories containing potentially sensitive information such as the Galaxy config file will be created group- but not world-readable. Otherwise, directories are created world-readable.
Access method control
The role needs to perform tasks as different users depending on which features you have enabled and how you are
connecting to the target host. By default, the role will use become
(i.e. sudo) to perform tasks as the appropriate
user if deemed necessary. Overriding this behavior is discussed in the defaults file.
Error documents
galaxy_errordocs_dir
: Install Galaxy-styled HTTP 413 and 502 error documents under this directory. The 502 message uses nginx server side includes to allow administrators to create a custom message in~/maint
when Galaxy is down. nginx must be configured separately to serve these error documents.galaxy_errordocs_server_name
(default: Galaxy): used to display the message "galaxy_errdocs_server_name
cannot be reached" on the 502 page.galaxy_errordocs_prefix
(default:/error
): Web-side path to the error document root.
Miscellaneous options
galaxy_restart_handler_name
: The role doesn't restart Galaxy since it doesn't control how Galaxy is started and stopped. Because of this, you can write your own restart handler and inform the role of your handler's name with this variable. See the examplesgalaxy_admin_email_to
: If set, email this address when Galaxy has been updated. Assumes mail is properly configured on the managed host.galaxy_admin_email_from
: Address to send the aforementioned email from.
None
Install Galaxy on your local system with all the default options:
- hosts: localhost
vars:
galaxy_server_dir: /srv/galaxy
connection: local
roles:
- galaxy
Once installed, you can start with:
$ cd /srv/galaxy
$ sh run.sh
Install Galaxy as per the current production server best practices:
- Galaxy code (clone) is "clean": no configs or mutable data live underneath the clone
- Galaxy code and static configs are privilege separated: not owned/writeable by the user that runs Galaxy
- Configuration files are not world-readable
- PostgreSQL is used as the backing database
- The 18.01+ style YAML configuration is used
- Two job handler mules are started
- When the Galaxy code or configs are updated by Ansible, Galaxy will be restarted using
supervisorctl
- hosts: galaxyservers
vars:
galaxy_config_style: yaml
galaxy_layout: root-dir
galaxy_root: /srv/galaxy
galaxy_commit_id: release_18.09
galaxy_separate_privileges: yes
galaxy_create_user: yes
galaxy_manage_paths: yes
galaxy_user: galaxy
galaxy_privsep_user: gxpriv
galaxy_group: galaxy
galaxy_restart_handler_name: Restart Galaxy
postgresql_objects_users:
- name: galaxy
password: null
postgresql_objects_databases:
- name: galaxy
owner: galaxy
galaxy_config:
uwsgi:
socket: 127.0.0.1:4001
buffer-size: 16384
processes: 1
threads: 4
offload-threads: 2
static-map:
- /static/style={{ galaxy_server_dir }}/static/style/blue
- /static={{ galaxy_server_dir }}/static
master: true
virtualenv: "{{ galaxy_venv_dir }}"
pythonpath: "{{ galaxy_server_dir }}/lib"
module: galaxy.webapps.galaxy.buildapp:uwsgi_app()
thunder-lock: true
die-on-term: true
hook-master-start:
- unix_signal:2 gracefully_kill_them_all
- unix_signal:15 gracefully_kill_them_all
py-call-osafterfork: true
enable-threads: true
mule:
- lib/galaxy/main.py
- lib/galaxy/main.py
farm: job-handlers:1,2
galaxy:
database_connection: "postgresql:///galaxy?host=/var/run/postgresql"
pre_tasks:
- name: Install Dependencies
apt:
name: ['git', 'python-psycopg2', 'python-virtualenv']
become: yes
roles:
# Install with:
# % ansible-galaxy install galaxyproject.postgresql
- role: galaxyproject.postgresql
become: yes
# Install with:
# % ansible-galaxy install natefoo.postgresql_objects
- role: natefoo.postgresql_objects
become: yes
become_user: postgres
- role: galaxyproject.galaxy
handlers:
- name: Restart Galaxy
supervisorctl:
name: galaxy
state: restarted
Academic Free License ("AFL") v. 3.0
This role was written and contributed to by the following people: