Reboot after provisioning is done

This page will show you an example of how to run a post-provisioning reboot with Rudder. It can be useful for example if your initial configuration applied by Rudder contains items that require a reboot to be applied (like firmware installation or upgrade).

You can use a similar technique to reboot nodes later in their life cycle too.

Detect nodes that need a reboot

We need a mechanism to decide when the provisioning is over and we want to reboot the node.

In order to do so, we can rely on the node’s state. First configure you default state to initializing in Settings → General:

Use initalizing state for new nodes

We will use this state to tag the nodes that are currently in provisioning state. Once provisioning is over, we’ll switch the node to the enabled state.

We now need to choose a criterion to detect the end of the initial configuration. A simple one is to consider reaching 100% compliance as marking the end of the provisioning.

To switch the state automatically in this case, you can run this script on your root server (for example every minute, as a cron job):


set -e
export PATH="/opt/rudder/bin/:${PATH}"

token=$(cat /var/rudder/run/api-token)
mkdir -p /var/log/rudder/jobs/

initializing_nodes=$(curl --silent -k --header "X-API-Token: ${token}" '\[\{"objectType":"node","attribute":"state","comparator":"eq","value":"initializing"\}\]' | jq -r '.data.nodes[].id')

for node in ${initializing_nodes}; do
  compliance=$(curl --silent -k --header "X-API-Token: ${token}" "${node}?level=0" | jq '.data.nodes[0].compliance')
  if [ "${compliance}" = "100" ]; then
    echo "$(date): ${node} has ${compliance}% compliance, switching to 'enabled' status" >> /var/log/rudder/jobs/provisioning.log
    curl --silent -k --header "X-API-Token: ${token}" --request POST "${node}?state=enabled" >/dev/null
    echo "$(date): ${node} has ${compliance}% compliance, skipping for now" >> /var/log/rudder/jobs/provisioning.log

Reboot the nodes

We will create a dedicated technique, with a Command execution once method. This method executes a command and creates a flag to prevent it from running again.

Make sure to specify the unique id parameter as it will be the name of the flag. If it changes, the command will be executed again.

We use reboot -r +1 to reboot the one after one minutes, which gives enough time to the agent to finish its run and send to reports to the server.

Technique to reboot a machine once

We will need to apply this configuration to the enabled nodes, so let’s create the group:

Group of enabled machines

Then create a directive based on this technique, and link it to the Enabled group with a dedicated Post-provisioning reboot rule.

Now when the script will change a node’s status it will:

  • Enter the Enabled nodes group

  • Get the Reboot once directive

  • Download its policies, and set the reboot

  • Send a "repaired" log for the reboot component

  • Reboot

  • From now on, it will send a "compliant" report for the reboot, as it has been done

← Synchronize files from external git repositories