streampunk.cc Ansible config

This playbook will install autoradio on a cluster of machines, along with the necessary infrastructure (etcd, logging, monitoring and alerting).

Configuration

There are some mandatory configuration variables that need to be set in group_vars/all/vars.yml (which ships with defaults that are only appropriate for the Vagrant test environment). They are:

• public_domain - the DNS domain to use to host the streaming infrastructure
• admin_emails - a list of email addresses that will receive alerts and other administrative spam
• smtp_server - SMTP server to use for outbound email
• smtp_auth_user - username for SMTP authentication
• smtp_auth_password - password for SMTP authentication

Networking

The playbook supports split networking scenarios, where the IP address used by the public is different than the one used by peers to communicate among themselves.

For every host, the Ansible variables public_ip and peer_ip must be defined. There are two ways to do this:

• you can specify these values explicitly in the Ansible inventory for each host, or
• Ansible can auto-detect the IP addresses if you specify the network interfaces used for public/peer networking with the public_network_interface and internal_network_interface variables.

By default, public_network_interface and internal_network_interface are both set to ens6, which happens to work well with the test Vagrant setup.

Inventory Groups

If you are using the default site.yml playbook, there are some host groups that will need to be defined in the Ansible inventory:

• etcd - hosts that will be running etcd
• monitor - hosts that will be running the monitoring stack

You should have at least 3 hosts in the etcd group, and at least 2 in the monitor group (for redundancy).

Testing

This repository contains a script to build dynamic, ephemeral test environments using virtual machines, so that different combinations of Ansible / autoradio / other packages can be tested before deploying them to production.

$ test.sh my-test

The above command will generate an Ansible configuration in the my-test directory, it will start some VMs, and it will run Ansible on them deploying the configuration from this repository.

The test environment can be customized by setting environment variables when calling test.sh:

• DIST: select the Debian distribution to use (default "buster")
• LIBVIRT_USER, LIBVIRT_HOST: use libvirt-over-SSH, needs SSH passwordless authentication to be set up.
• APT_PROXY: set to the host:port address of apt-cacher-ng. When using this directive, ensure that the apt-cacher-ng configuration has a permissive PassThroughPattern to allow access to https:// repositories.
• VM: select the VM manager to use, one of "vagrant" (default) or "vmine" (used in CI setups).
• VERBOSE: set to a non-empty value to run ansible-playbook with increased verbosity.
• AUTORADIO_SRC: if set, points at a local copy of the autoradio source repository which will be compiled and installed instead of using the default packages from the autoradio Debian repository.

So, for example, if you have a copy of the autoradio source repository with some local changes that you want to test, you can use:

$ AUTORADIO_SRC=/path/to/git.autistici.org/ale/autoradio test.sh my-test

and the test environment in my-test will be set up to build autoradio from your local sources.

Check the hosts.ini file in the my-test directory to find out the IP addresses of the virtual machines (which are randomly generated).

Test stream

The test-source.yml Ansible playbook can be used to quickly setup a stream and an associated source for testing purposes. Use the stream_name Ansible variable to set the name of the stream to be created:

$ ansible-playbook -i hosts.ini \
    -e stream_name=/test.ogg \
    test-source.yml

Then it's easy to create test loads with, e.g., autoradio's radiobench utility:

$ go run git.autistici.org/ale/autoradio/cmd/radiobench \
    -n 100 http://10.236.82.10/test.ogg

will start 100 listeners on the stream just created above.