Configure agents

Inmanta agents can be started automatically (auto-started agents) or manually (manually-started agents). This section describes how both types of agents can be set up and configured. Inmanta agents only run on Linux.

Auto-started agents

Auto-started agents always run on the Inmanta server. The Inmanta server manages the full lifecycle of these agents.

Configuring auto-started agents via environment settings

Auto-started agents can be configured via the settings of the environment where the auto-started agent belongs to. The following options are configurable:

• autostart_agent_map

• autostart_agent_deploy_interval

• autostart_agent_deploy_splay_time

• autostart_agent_repair_interval

• autostart_agent_repair_splay_time

• autostart_on_start

The autostart_agent_map requires an entry for each agent that should be autostarted. The key is the name of the agent and the value is either local: for agents that map to the Inmanta server or an SSH connection string when the agent maps to a remote machine. The SSH connection string requires the following format: ssh://<user>@<host>:<port>?<options>. Options is a ampersand-separated list of key=value pairs. The following options can be provided:

Option name	Default value	Description
retries	10	The amount of times the orchestrator will try to establish the SSH connection when the initial attempt failed.
retry_wait	30	The amount of second between two attempts to establish the SSH connection.
python	python	The Python3 interpreter available on the remote side. This executable has to be discoverable through the system PATH.

Auto-started agents start when they are required by a specific deployment or when the Inmanta server starts if the autostart_on_start setting is set to true. When the agent doesn’t come up when required, consult the troubleshooting documentation to investigate the root cause of the issue.

Configuring the autostart_agent_map via the std::AgentConfig entity

The std::AgentConfig entity provides functionality to add an entry to the autostart_agent_map of a specific environment. As such, the auto-started agents can be managed in the configuration model.

Special Requirements for remote std::File, std::Package, std::Service and exec::Run

When using the agents built-in ssh capability, to perform actions over ssh on remote hosts, the following requirements must be met:

• The Inmanta server should have passphraseless SSH access on the machine it maps to. More information on how to set up SSH connectivity can be found at Step 7: Configure ssh of the inmanta user

• The remote machine should have a Python 2 or 3 interpreter installed. The binary executed by default is python.

• The user to log into the remote machine should either be root or have the ability to do a passwordless sudo. To enable passwordless sudo for the user username, add a file to /etc/sudoers.d/ containing username ALL=(ALL) NOPASSWD: ALL. It is advisable to use a safe editor such as visudo or sudoedit for this. For more details, go here.

Manually-started agents

Manually started agents can be run on any Linux device, but they should be started and configured manually as the name suggests.

Requirements

The following requirements should be met for agents that don’t map to the host running the agent process (i.e. The managed device is remote with respect to the Inmanta agent and the agent has to execute I/O operations on the remote machine using self._io):

• The Inmanta agent should have passphraseless SSH access on the machine it maps to. More information on how to set up SSH connectivity can be found at Step 7: Configure ssh of the inmanta user

• The remote machine should have a Python 2 or 3 interpreter installed. The binary executed by default is python.

Step 1: Installing the required Inmanta packages

In order to run a manually started agent, the inmanta-oss-agent package is required on the machine that will run the agent.

sudo tee /etc/yum.repos.d/inmanta-oss-stable.repo <<EOF
[inmanta-oss-stable]
name=Inmanta OSS stable
baseurl=https://packages.inmanta.com/public/oss-stable/rpm/el/\$releasever/\$basearch
gpgcheck=1
gpgkey=https://packages.inmanta.com/public/oss-stable/gpg.A34DD0A274F07713.key
repo_gpgcheck=1
enabled=1
enabled_metadata=1
EOF

sudo dnf install -y inmanta-oss-agent

Step 2: Configuring the manually-started agent

The manually-started agent can be configured via a /etc/inmanta/inmanta.d/*.cfg config file. The following options configure the behavior of the manually started agent:

• config.state-dir

• config.agent-names

• config.environment

• config.agent-map

• config.agent-deploy-splay-time

• config.agent-deploy-interval

• config.agent-repair-splay-time

• config.agent-repair-interval

• config.agent-reconnect-delay

• config.server-timeout

• agent_rest_transport.port

• agent_rest_transport.host

• agent_rest_transport.token

• agent_rest_transport.ssl

• agent_rest_transport.ssl-ca-cert-file

The agent will follow the pip configuration defined in the project.yml. Make sure it can access the pip index configured by the project (See the pip documentation for netrc for more information on how to setup authentication).

The config.agent-map option can be configured in the same way as the autostart_agent_map for auto-started agents.

Step 3: Starting the manually-started agent

Finally, enable and start the inmanta-agent service:

sudo systemctl enable inmanta-agent
sudo systemctl start inmanta-agent

The logs of the agent are written to /var/log/inmanta/agent.log. When the agent doesn’t come up after starting the inmanta-agent service, consult the troubleshooting documentation to investigate the root cause of the issue.