|Version 10 (modified by 4 years ago) (diff),|
Fabric administration scripts
This page describes how to use the Fabric-based administration scripts to manage services running on twistedmatrix.com. The temporary location for the administration scripts is at: https://github.com/twisted-infra/braid
Getting the tools
The first step is to install the needed libraries and tools. I'm using virtualenv and virtualenvwrapper in this guide, you are free to install everything on your stock python.
mkvirtualenv env_name workon env_name pip install fabric git clone git://github.com/GaretJax/twisted-setup-temp.git tmadmin # May be svn or bzr as well, location is temporary cd tmadmin
You can now list all available tasks by calling:
Collection of utilities to automate the administration of Twisted's infrastructure. Use this utility to install, update and start/stop/restart services running on twistedmatrix.com. Available commands: dns.install dns.restart dns.start dns.stop dns.update ...
Before being able to run any task, we need to configure the tool to work with Twisted's infrastructure and your login user. All configuration directives are contained in the
config.py file. The base installation ships with a sample configuration file in
config.py.sample. Copy it over and start editing it:
cp config.py.sample config.py vim config.py
The two main directives to change are the
USER ones. The first contains a list of hosts you want to administer, while the second one is, well, your ssh user to access the previously defined hosts. All found uppercase variables are lowercased and set as attributes of Fabric's
env global variable. This means that any configuration directive recognized by Fabric can be defined in the
config.py file as well.
Note: your user has to be at least in the
service-admin group on the servers you want to manage in order to be able to start/stop/restart and update services. If you want to be able to install new services or give permissions to other users to manage them, your user has to be able to gain root access as well.
Server configuration conventions
Each service has its own directory under
/srv. The structure of the service specific directory reflects the Filesystem Hierarchy Standard (e.g. pid files in
/srv/<service>/var/run/<service>.pid, log files in
Users, groups and privileges
Each service runs as its own system user and owns his root directory (i.e.
/srv/<service>). Each service user has to be part of the
service group. To operate on a service (start/stop, update,...) a given user has to be part of the
service-admin group. Each user in this group has permissions to
sudo to every user in the
As the service user is a system user, login for this user is disabled.
For each service, a System V style init script is provided in
/srv/<service>/etc/init.d/<service>. This script is owned by the service user and has mode
0755. A symlink is provided in
/etc/init.d, this allows the script to be managed by
init(8) and allows updates to be applied to the service directory by a service administrator user.
For convenience, a runner which simplifies the creation of new init scripts is available at
/usr/bin/twistd-service. A minimal example of an init script for an hypothetical DNS service is shown below:
#!/usr/bin/twistd-service ### BEGIN INIT INFO # Provides: dns # Required-Start: $named $network $time # Required-Stop: $named $network # Should-Start: # Should-Stop: # Default-Start: 2 3 4 5 # Default-Stop: 0 1 6 # Short-Description: Twisted DNS server # Description: dns is a DNS server based on Twisted Names ### END INIT INFO SERVICE='dns' ARGS='dns'
This script is a complete System V compatible init script to run
twistd from a specific virtualenv with the correct pid and logfile arguments under the correct user. It provides the
zap commands. It can be run by the root user, the service specific user or any user in the
service-admin group and it is started with the correct user.
How to start/stop/restart services
Each service has its own Fabric namespace. Actions are available as part of each namespace. For example, the
dns service can be started, stopped, and restarted as follows:
How to update existing services
Similarly as done for managing the running states, an
update task lives in each service namespace. It can be run as follows:
Note that a restart is still needed after updating a service.
How to install new services
A service which was just added to the fabfile can be installed by running its
Note, however, that while the previous actions did not require root privileges (being member of the
service-admin group was sufficient), installing a new service requires to be able to
root. This is needed to create the necessary users, install additional packages and create the base environment.
How to give service administration privileges to another user
An existing user which does not currently have service administration tasks can be added to the
service-admin group by executing the
make_service_admin task and supplying a username: