Workers are machines that can execute tasks that don't need to be run on the Octopus Server or individual deployment targets.
You can manage your workers by navigating to Infrastructure ➜ Worker Pools in the Octopus Web Portal:
Workers are useful for the following scenarios:
- Publishing to Azure websites.
- Deploying AWS CloudFormation templates.
- Deploying to AWS Elastic Beanstalk.
- Uploading files to Amazon S3.
- Backing up databases.
- Performing database schema migrations
- Configuring load balancers.
Where steps run
The following step types and configurations run on a worker:
- Any step that runs a script (usually user supplied) or has a package that has an execution plan of
Octopus Server on behalf of roles,
Worker Pool on behalf of roles.
- Any steps that run on a Cloud Region, an Azure Target, or any target that isn’t a Tentacle, an SSH Target, or an Offline Drop.
- All AWS, Terraform, and Azure steps.
The following steps always run inside the Octopus Server process (and do not run user-supplied code):
- Manual intervention
- Import certificate
A worker receives instruction from the Octopus Server to execute a step, it executes the step using Calamari and returns the logs and any collected artifacts to the Octopus Server.
There are two kinds of workers you can use in Octopus:
Octopus works out-of-the-box without setting up workers. You can run all deployment processes, run script steps on the built-in worker, deploy to Azure and run AWS and Terraform steps, without further setup. The built-in worker is available in a default Octopus set up, and Octopus workers are designed so that, if you aren't using external workers, none of your deployment processes need to be worker aware.
The choices of built-in worker, built-in worker running in a separate account, and external workers enable to you harden your Octopus Server and scale your deployments.
Migrating to Workers
Octopus workers also provides a smooth path to move off the built-in worker, and thus off running scripts on the Octopus Server, and onto external workers, without updating any deployment processes. Learn about how to use the default worker pool to move steps off the Octopus Server.
The Octopus Server has an built-in worker that can deploy packages, execute scripts, and perform tasks that don't need to be performed on a deployment target. The built-in worker is configured by default, however, the built-in worker can be disabled by navigating to Configuration and selecting Disable for the Run steps on the Octopus Server option.
The built-in worker is executed on the same machine as the Octopus Server. When the built-in worker is needed to execute a step, the Octopus Server spawns a new process and runs the step using Calamari. The spawned process is either under the server's security context (default) or under a context configured for the built-in worker.
Adding a worker to the default worker pool will disable the built-in worker, and steps will no longer run on the Octopus Server.
Learn about the security implications and how to configure the built-in worker.
An External Worker is either a Tentacle or an SSH machine that has been registered with the Octopus Server as a worker. The setup of a worker is the same as setting up a deployment target as a Windows Tentacle target or an SSH target, except that instead of being added to an environment, a worker is added to a worker pool.
Using external workers allows delegating work to a machine other than the Octopus Server. This can make the server more secure and allow scaling. When Octopus executes a step on an external worker, it's the external worker that executes Calamari and no user-provided script executes on the Octopus Server itself.
Workers have machine policies, are health checked, and run Calamari, just like deployment targets.
Registering an External Worker
Once the Tentacle or SSH machine has been configured, workers can be added using the Web Portal, the Octopus Deploy REST API, the Octopus.Clients library or with the tentacle executable. Only a user with the
ConfigureServer permission can add or edit workers.
Registering Workers in the Octopus Web Portal
- Navigate to Infrastructure ➜ Workers and click ADD WORKER.
- Select WINDOWS or SSH CONNECTION and click the card for the type of worker you want to configure.
You can choose between:
- Register a Worker as a Listening Mode.
- Register a Worker as a Polling Mode.
- Register a Worker with an SSH Connection.
Register a Worker as a Listening Tentacle
Before you can configure your Windows servers as Tentacles, you need to install Tentacle Manager on the machines that you plan to use as Tentacles.
Tentacle Manager is the Windows application that configures your Tentacle. Once installed, you can access it from your start menu/start screen. Tentacle Manager can configure Tentacles to use a proxy, delete the Tentacle, and show diagnostic information about the Tentacle.
Start the Tentacle installer, accept the license agreement, and follow the prompts.
When the Octopus Deploy Tentacle Setup Wizard has completed, click Finish to exit the wizard.
When the Tentacle Manager launches, click GET STARTED.
On the communication style screen, select Listening Tentacle and click Next.
In the Octopus Web Portal, navigate to the Infrastructure tab, select Deployment Targets and click ADD DEPLOYMENT TARGET ➜ WINDOWS, and select Listening Tentacle.
Copy the Thumbprint (the long alphanumerical string).
Back on the Tentacle server, accept the default listening port 10933 and paste the Thumbprint into the Octopus Thumbprint field and click Next.
Click INSTALL, and after the installation has finished click Finish.
Back in the Octopus Web Portal, enter the hostname or IP address of the machine the Tentacle is installed on, i.e.,
10.0.1.23, and click NEXT.
Add a display name for the deployment target (the server where you just installed the Listening Tentacle).
Select which worker pool the deployment target will be assigned to and click SAVE.
After you have saved the new worker, you can navigate to the worker pool you assigned the worker to, to view its status.
Register a Worker as a Polling Tentacle
It is not currently possible to configure a worker as a Polling Tentacle with the Tentacle Manager, please Registering Workers with the Tentacle Executable.
Register a Worker with an SSH Connection
To register a worker with an SSH Connection, see the instructions for configuring SSH deployment targets.
Registering Workers with the Tentacle executable
Tentacle workers can also register with the server using the Tentacle executable (version 3.22.0 or later), for example:
.\Tentacle.exe register-worker --instance MyInstance --server "https://example.com/" --comms-style TentaclePassive --apikey "API-CS0SW5SQJNLUBQCUBPK8LZY3KYO" --workerpool "Default Worker Pool"
TentacleActive instead of
TentaclePassive to register a polling Tentacle worker.
The Tentacle executable can also be used to deregister workers, for example:
.\Tentacle.exe deregister-worker --instance MyInstance --server "https://example.com/" --apikey "API-CS0SW5SQJNLUBQCUBPK8LZY3KYO"
Recommendations for External Workers
We highly recommend setting up external workers on a different machine to the Octopus Server.
We also recommend running external workers as a different user account to the Octopus Server.
It can be advantageous to have workers on the same local network as the server to reduce package transfer times.
Default pools attached to cloud targets allow co-location of workers and targets, this can help make workers specific to your targets as well as making the Octopus Server more secure by using external workers.
Multiple projects run simultaneously on Workers
Many workers may be running in parallel and a single worker can run multiple actions in parallel.
The task cap determines how many tasks (deployments or system tasks) can run simultaneously. The Octopus System Variable
Octopus.Action.MaxParallelism controls how much parallelism is allowed in executing a deployment action. It applies the same to deployment targets as it does to workers. For example, if
Octopus.Action.MaxParallelism is at its default value of 10, any one deployment action will being deploying to at most 10 deployment targets simultaneously, or have no more than 10 concurrent worker invocations running. Parallel steps in a deployment can each reach their own
MaxParallelism. Coupled with multiple deployment tasks running, up to the task cap, you can see the number of concurrent worker invocations can grow quickly.
External workers and the built-in worker have the same behavior in this regard and in that Workers can run many actions simultaneously and can run actions from different projects simultaneously. Note that this means the execution of an action doesn't have exclusive access to a worker, which could allow one project to access the working folder of another project.
Note that if external workers are added to the default pool, then the workload is shared across those workers: a single external worker will be asked to perform exactly the same load as the built-in worker would have been doing, two workers might get half each, etc.
Workers in HA setups
In an HA Octopus setup, each node has a task cap and can invoke a built-in worker locally, so for a 4-node HA cluster, there are 4 built-in workers. Therefore if you move to external workers, it's likely you'll need to provision workers to at least match your server nodes, otherwise, you'll be asking each worker to do the sum of what the HA nodes were previously doing.