David Shrewsbury ae8952dea7 Add support for version discovery to worker.

The worker now supports a DISCOVER message that can be sent to it
to discover which version of the JSON message format that it supports.
The Gearman function name that it registers is now simply the hostname
and no longer includes 'lbaas-<VERSION>-' as a prefix.

Change-Id: Ib34ae424e5461b3b1f8df601a6e7a1fd99c48c05

2012-11-26 14:25:05 -05:00

3.1 KiB

Raw Blame History

Description

Purpose

A Python-based Gearman worker that handles messages for the Gearman job queue sharing the same name as the local hostname. The messages that it receives are JSON objects describing a load balancer, and returns this same JSON object, but with status fields added to describe the state of the LB.

Installation

Installing the Required Tools

You must have Python setuptools installed. On Ubuntu:

$ sudo apt-get install python-setuptools

Now you may install the Libra toolset:

$ sudo python setup.py install

The worker also needs some packages installed in order to be used with HAProxy. The commands below will install them on Ubuntu:

$ sudo apt-get install haproxy
$ sudo apt-get install socat

The Ubuntu default is to have HAProxy disabled. You will need to edit the file /etc/default/haproxy and set ENABLED to 1 if you want HAProxy to actually start (hint: you do).

Edit /etc/sudoers

The worker needs to be able to run some commands as root without being prompted for a password. It is suggested that you run the worker as the haproxy user and haproxy group on Ubuntu systems. Then add the following line to /etc/sudoers:

%haproxy ALL = NOPASSWD: /usr/sbin/service, /bin/cp, /bin/mv, /bin/rm, /usr/bin/socat

The above lets everyone in the haproxy group run those commands as root without being prompted for a password.

Configuration File

It can be easier to give options via a configuration file. See the sample configuration file etc/sample_libra.cfg for an example and further documentation. Use the --config <libra_worker.py -c> option to specify the configuration file to read.

Running the Worker

The worker can run in either daemon or non-daemon mode. Daemon mode requires escalated privileges so that it can behave like a proper daemon. Non-daemon mode (--nodaemon <libra_worker.py -n> option) is useful for testing.

Basic commands:

# Getting help
$ libra_worker -h

# Start up as a daemon running as the `haproxy` user and
# connecting to the local Gearman job server.
$ sudo libra_worker --user haproxy --group haproxy --server 127.0.0.1:4730

# Start up with debugging output in non-daemon mode
$ libra_worker --debug --nodaemon

NOTE: When running the worker in daemon mode, you must make sure that the directory where the PID file will be (--pid <libra_worker.py -p> option) and the directory where the log files will be written (--logfile <libra_worker.py -l> option) exists and is writable by the user/group specified with the --user <libra_worker.py --user> and --group <libra_worker.py --group> options. Also, the Python module used to start the daemon process does not like it when the PID file already exists at startup.

IF THE WORKER IMMEDIATELY EXITS WHEN STARTED IN DAEMON MODE, AND NO ERROR MESSAGES ARE IN THE LOG, ONE OF THESE REASONS IS THE MOST LIKELY CAUSE!

You can verify that the worker is running by using the sample Gearman client in the bin/ directory:

$ bin/client.py

3.1 KiB Raw Blame History