Machine policies

Machine policies are groups of settings that can be applied to Tentacle and SSH endpoints to modify their behavior. They can be used to:

  • Customize the interval between health checks.
  • Run custom health check scripts.
  • Ignore machines that are unavailable during health checks.
  • Configure how Calamari and Tentacles, and SSH Targets are updated.
  • Recover from communication errors with Tentacle.
  • Automatically delete machines.

You can access the machine policies by navigating to Infrastructure ➜ Machine policies.

Getting Started - Machine Policies

Health check

Octopus periodically runs health checks on deployment targets and workers to ensure that they are available and running the latest version of Calamari.

The health status of a deployment target can be set by custom health check scripts. Deployment targets can have four health statuses:

  • Healthy
  • Healthy with warnings
  • Unhealthy
  • Unavailable

A healthy deployment target completes a health check without any errors or warnings. A deployment target that is healthy with warnings completes a health check but encounters a non-critical failure during the health check. An unhealthy deployment target completes a health check but encounters a critical failure while running the health check script. An unavailable deployment target is not contactable by Octopus during a health check.

You can review the health status of you deployment targets by navigating to Infrastructure ➜ Deployment Targets.

Health status of deployment targets

Health check step

It’s also possible to run a health check as part of a deployment or runbook using the built-in health check step.

This step allows a deployment target that was created in the currently executing deployment to be confirmed as healthy and then added to the running deployment for subsequent steps.

Similarly, it allows you to confirm that the Tentacle service on a deployment target is running prior to attempting to perform an action against it.

Initial health check

After installing and configuring a new Tentacle, you need to run a health check and can upgrade the version of Calamari.

  1. From the Infrastructure tab, select deployment targets.
  2. Click the … overflow menu and select Check Health. If you’ve installed multiple Tentacles, it will check all of your Tentacles (if you’d rather check only one Tentacle, select that Tentacle from the deployment targets section, click Connectivity and then Check health).

The first time you complete a health check on a Tentacle or SSH Target, you will see health warnings and that Calamari needs to be installed.

Learn more about Calamari.

Octopus will automatically push the latest version of Calamari with your first deployment, but you can do the following to install Calamari:

  1. From the Infrastructure tab, select deployment targets.
  2. Click the … overflow menu and select Upgrade Calamari on Deployment Targets.

Health check interval

You can set the “Time between checks” to control how frequently automatic health checks run.

Health check type

You can configure health checks to run scripts on deployment targets, or just check that a connection can be established with the deployment target. When the “Run health check scripts” option is selected you will also have the opportunity to customize the PowerShell and Bash scripts that will be executed during the health check. The “Only perform connection test” option is recommended if you are using the raw scripting feature.

Custom health check scripts

Machine policies allow the configuration of custom health check scripts for Tentacle and SSH deployment targets. While we do not expose the full underlying script that runs during health checks, we give you an entry point to inject your own custom scripts. For example, here is the default custom health check script for PowerShell that checks disk space:

Default PowerShell health check script

$freeDiskSpaceThreshold = 5GB
Try {
	Get-WmiObject win32_LogicalDisk -ErrorAction Stop  | ? { ($_.DriveType -eq 3) -and ($_.FreeSpace -ne $null)} |  % { CheckDriveCapacity @{Name =$_.DeviceId; FreeSpace=$_.FreeSpace} }
} Catch [System.Runtime.InteropServices.COMException] {
	Get-WmiObject win32_Volume | ? { ($_.DriveType -eq 3) -and ($_.FreeSpace -ne $null) -and ($_.DriveLetter -ne $null)} | % { CheckDriveCapacity @{Name =$_.DriveLetter; FreeSpace=$_.FreeSpace} }
	Get-WmiObject Win32_MappedLogicalDisk | ? { ($_.FreeSpace -ne $null) -and ($_.DeviceId -ne $null)} | % { CheckDriveCapacity @{Name =$_.DeviceId; FreeSpace=$_.FreeSpace} }
}

The function CheckDriveCapacity informs you about how much space is available on your deployment target’s local hard disk and will write a warning if the free disk space is less than this threshold. You can add additional PowerShell to this script to customize your health checks as you wish, modify or remove the disk space checking altogether. It’s entirely up to you! Just remember, you can copy and paste the original script above back into your machine policy if you run into any problems and wish to get back to the default behavior.

Set the status

A health check script can set the status of a target by returning a non-zero exit code or by writing a service message during the health check. PowerShell based deployment targets can use Write-Warning, Write-Error and Fail-HealthCheck to convey a healthy with warnings or unhealthy status:

PowerShell health check service messages

# For setting a health status of Healthy with Warnings:
Write-Warning "This is a warning"
# For setting a health status of Unhealthy:
Write-Error "This is an error"
Fail-HealthCheck "This is an error"

Bash targets do not include a disk space check by default like PowerShell targets do. As such, there is no default Bash script listed in your machine policy for Bash targets by default. However, you may write your own, or choose to add additional Bash script to run against your Bash targets during health checks. Again, it’s entirely up to you. Unless you select the Only perform connection test option, there are some system prerequisites that are included as part of the standard health check.

Bash deployment targets can use echo_warning, echo_error and fail_healthcheck to convey a healthy with warnings or unhealthy status:

Bash Health Check Service Messages

# For setting a health status of Healthy with Warnings:
echo_warning "This is a warning"
# For setting a health status of Unhealthy:
echo_error "This is an error"
fail_healthcheck "This is an error"

Agent-Level variables

When using a custom health check script, the script execution through Calamari is bypassed. This results in some behavioral differences compared with the normal scripting in Octopus that you would be accustomed to. You can still use the standard # variable substitution syntax, however since this is replaced on the server, environment variables from your target will not be available through Octopus variables.

Ignore machines that are unavailable during health checks

By default, health checks fail if any deployment targets are unavailable during the health check. Machine policies offer an option to ignore machines if they are unavailable during a health check:

By selecting Unavailable machines will not cause health checks to fail, any deployment targets that Octopus cannot contact during a health check will be skipped and the health check marked as successful. If the target is contactable but encounters an error or warning, the usual health check behavior will proceed (i.e., a warning will be reported or the health check will fail with an error).

Configure how Calamari and Tentacle are updated

Brand new Tentacle and SSH endpoints require the installation of Calamari to perform a deployment. Also, if Calamari is updated, the Octopus Server will push the update to Tentacle and SSH endpoints. When there is a Tentacle update, Octopus can automatically update Tentacle endpoints. Machine policies allow the customization of when Calamari and Tentacle updates occur.

By default, Calamari will be installed or updated when a machine is involved in a deployment. The other two options will update Calamari:

  • the first time a machine is added to Octopus and then subsequently when it is involved in a deployment.
  • any time Octopus detects Calamari is out of date (after health checks for example).

Tentacle can be toggled to manually or automatically update. If Automatically is selected, Octopus will start a task to update Tentacles whenever Octopus detects that there is a pending Tentacle upgrade (after health checks for example). Conversely, Octopus will not automatically update Tentacle but instead will display a prompt to begin a Tentacle update on the Deployment Targets and Environments screens.

Maximum number of concurrent upgrades

There is a limit to the number of concurrent upgrades possible when choosing Always keep Calamari up to date. This ensures that upgrades do not adversely effect the performance of your Octopus Server.

The number of concurrent upgrades will be double the Octopus Server’s logical processor count which is a minimum of 2 and will not exceed 32.

Tentacle update account

You can select a username/password account to perform automatic Tentacle updates. When no account is selected, the account that the Tentacle service is running as will attempt to perform Tentacle updates. If this account is not an Administrator it will not have enough permission to perform Tentacle updates. In that scenario you will need to create a username/password account for a user with administrative rights to install software on your machines and select it from the drop down.

Note: This option can not be used when Tentacle is running as Local System.

Recover from communication errors with Tentacle

Recovering from communication errors with Tentacle is available from Octopus 2023.4 onwards.

You can configure whether Octopus Deploy should re-attempt failed communications with a Tentacle for a set duration when a network error occurs. This is particularly useful to reduce deployment failures that occur when Tentacle is on an unstable network connection.

  • With this setting Disabled, if a network error occurs while Octopus Deploy is communicating with a Tentacle, the deployment, runbook run, or health check will fail.

  • With this setting Enabled, if a network error occurs, Octopus will re-attempt communication with the Tentacle for the amount of time configured. However, continued network errors will cause a failure.

Using this setting with a Tentacle that runs on ephemeral storage will introduce the possibility that a script could run twice. For example, when running Tentacle in Docker without a mount. It is recommended to run Tentacle on a persistent file system.

Retry durations

You can configure the amount of time allowed for Octopus to re-attempt failed communication with Tentacle. There are two duration configurations available:

  • Deployment or Runbook Run
  • Health Check

Step Retries and execution timeouts

If you would like to retry a particular step within the deployment process for other types of temporary or transient errors, that can be configured separately.

Automatically delete machines

Machine policies can be configured to automatically remove unavailable machines after a time period. When a health check runs, Octopus detects if machines are unavailable (cannot be contacted). When the Automatically delete unavailable machines option is set, Octopus checks how long a machine has been unavailable. If the specified time period has elapsed, the machine is permanently deleted from Octopus.

Assign machine policies to machines

Assign a machine policy to a machine by selecting a machine from the Environments screen and using the Policy drop down to select the machine policy:

Machine policy can also be set from the command line by using the —policy argument:

Setting machine policy

Tentacle.exe register-with --instance "Tentacle" --server "http://YOUR_OCTOPUS" --apiKey="API-YOUR_API_KEY" --role "web-server" --environment "Staging" --comms-style TentaclePassive --policy "Transient machines"

Help us continuously improve

Please let us know if you have any feedback about this page.

Send feedback

Page updated on Wednesday, October 4, 2023