Every job that can touch your Spacelift-managed infrastructure is called a Run. There are four main types of runs, and each of them warrants a separate section.
Three of them are children of Stacks:
task, which is a freeform command you can execute on your infrastructure;
proposed run, which serves as a preview of introduced changes;
tracked run, which is a form of deployment;
There's also a fourth type of run - module test case. Very similar to a tracked run, it's executed on a Terraform module.
In Spacelift, each run is executed on a worker node, inside a Docker container. We maintain a number of these worker nodes (collectively known as the public worker pool) that are available to all customers, but also allow individual customers to run our agent on their end, for their exclusive use. You can read more about worker pools here.
Regardless of whether you end up using a private or a public worker pool, each Spacelift run involves a handover between spacelift.io (which we like to call the mothership) and the worker node. After the handover, the worker node is fully responsible for running the job and communicating the results of the job back to the mothership.
Regardless of the type of the job performed, some phases and terminal states are common. We discuss them here, so that we can refer to them when describing various types of runs in more detail.
Queued means that no worker has picked up the run yet. There are two possible reasons for that: there are no available workers, or the run is not eligible for processing.
Let's start with the latter - more common - scenario. Spacelift serializes all state-changing operations to the Stack. Both tracked runs and tasks have the capacity to change the state, they're never allowed to run in parallel. Instead, each of them gets an exclusive lock on the stack, blocking others from starting.
If you run or task is currently blocked by something else holding the lock on the stack, you'll see the link to the blocker in the header:
The other scenario is just running out of available workers. If you're using the public worker pool, you can track its availability on our status page. In particular, you should look at a system metric called Public worker queuing time. It indicates how long has the oldest run spent in the queue since becoming eligible for processing:
If you're using private workers, please refer to the worker pools documentation for troubleshooting advice.
Queued is a passive state meaning no operations are performed while a run is in this state. When a run is eligible for processing and a worker is available to pick it up, the state will automatically transition to Preparing. The user can cancel the run while it's still queued, transitioning it to the terminal Canceled state.
Canceled state means that the user has manually stopped a Queued run or task even before it had the chance to be picked up by the worker.
Canceled is a passive state meaning no operations are performed while a run is in this state. It's also a terminal state meaning that no further state can follow it.
The preparing state is the first one where real work is done. At this point both the run is eligible for processing and there's a worker node ready to process it. The preparing state is all about the handover and dialog between spacelift.io and the worker.
Here's an example of one such handover:
Note that Ground Control refers to the bit directly controlled by us, in a nod to late David Bowie. The main purpose of this phase is for Ground Control to make sure that the worker node gets everything that's required to perform the job, and that it can take over the execution.
Once the worker is able to pull the Docker image and use it to start the container, this phase is over and the initialization phase begins. If the process fails for whatever reason, the run is marked as failed.
The last phase where actual work is done and which is common to all types of run is the initialization. This phase is handled exclusively by the worker and involves running pre-initialization hooks and vendor-specific initialization process. For Terraform stacks it would mean running
terraform init, in the right directory and with the right parameters.
Important thing to note with regards to pre-initialization hooks and the rest of the initialization process is that all these run in the same shell session, so environment variables exported by pre-initialization hooks are accessible to the vendor-specific initialization process, but are gone and cannot be retrieved afterwards. This is often the desired outcome when working with external secret managers like HashiCorp Vault.
If this phase fails for whatever reason, the run is marked as failed. Otherwise, the next step is determined by the type of the run being executed.
If a run transitions into the failed state means that something, at some point went wrong and this state can follow any state. In most cases this will be something related to your project like:
errors in the source code;
pre-initialization checks failing;
plan policies rejecting your change;
In rare cases errors in Spacelift application code or third party dependencies can also make the job fail. These cases are clearly marked by a notice corresponding to the failure mode, reported through our exception tracker and immediately investigated.
Failed is a passive state meaning no operations are performed while the run is in this state. It's also a terminal state meaning that no further state can supersede it.
Finished state means that the run was successful, though the success criteria will depend on the type of run. Please read the documentation for the relevant run type for more details.
Finished is a passive state meaning no operations are performed while a run is in this state. It's also a terminal state meaning that no further state can supersede it.
Some types of runs in some phases may safely be interrupted. We allow sending a stop signal from the GUI and API to the run, which is then passed to the worker handling the job. It's then up to the worker to handle or ignore that signal.
Generally speaking, Spacelift tries to balance convenience and safety and defaults to only protecting your infrastructure state. Tracked runs cannot be stopped when they're applying changes and tasks cannot be stopped once they start executing the user-supplied command.
Stopped state indicates that a run has been stopped while Initializing or Planning, either manually by the user or - for proposed changes - also by Spacelift. Proposed changes will automatically be stopped when a newer version of the code is pushed to their branch. This is mainly designed to limit the number of unnecessary API calls to your resource providers, though it saves us a few bucks on EC2, too.
Here's an example of a run manually stopped while Initializing:
Stopped is a passive state meaning no operations are performed while a run is in this state. It's also a terminal state meaning that no further state can supersede it.