Skip to content

Approval policy»

Info

Please note, we currently don't support importing rego.v1. For more details, refer to the note in the introduction section.

The approval policy allows organizations to create sophisticated run review and approval flows that reflect their preferred workflow, security goals, and business objectives. Without an explicit approval policy, anyone with write access to a stack can create a run (or a task). An approval policy can make this process more granular and contextual.

Runs can be reviewed when they enter one of three states - queued, unconfirmed, or pending review. Please note if a stack has autodeploy enabled, then the approval policy will not be evaluated here, and you should use a plan policy to warn which will force the stack into an unconfirmed state and the approval policy will get evaluated as a safeguard.

When a queued run needs approval, it will not be scheduled before that approval is received, and if it is of a blocking type, it will block newer runs from scheduling, too. A queued run that's pending approval can be canceled at any point.

Here's an example of a queued run waiting for a human review - note how the last approval policy evaluation returned an Undecided decision. There's also a Review button next to the Cancel button:

Review can be positive (approve) or negative (reject):

With a positive review, the approval policy could evaluate to Approve, thus unblocking the run:

When an unconfirmed run needs approval, you will not be able to confirm it until that approval is received. The run can, however, be discarded at any point:

In principle, the run review and approval process is very similar to GitHub's Pull Request review, the only exception being that it's the Rego policy (rather than a set of checkboxes and dropdowns) that defines the exact conditions to approve the run.

Tip

If separate run approval and confirmation steps sound confusing, don't worry. Just think about how GitHub's Pull Requests work - you can approve a PR before merging it in a separate step. A PR approval means "I'm OK with this being merged." A run approval means "I'm OK with that action being executed."

Rules»

Your approval policy can define the following boolean rules:

  • approve: the run is approved and no longer requires (or allows) review;
  • reject: the run fails immediately.

While the 'approve' rule must be defined in order for the run to be able to progress, it's perfectly valid to not define the 'reject' rule. In that case, runs that look invalid can be cleaned up (canceled or discarded) manually.

It's also perfectly acceptable for any given policy evaluation to return 'false' on both 'approve' and 'reject' rules. This only means that the result is yet 'undecided', and more reviews will be necessary to reach a conclusion. A perfect example would be a policy that requires 2 approvals for a given job - the first review is not yet supposed to set the 'approve' value to 'true'.

Info

Users must have write or admin access to the stack to be able to approve changes.

How It Works»

When a user reviews the run, Spacelift persists their review and passes it to the approval policy, along with other reviews, plus some information about the run and its stack. The same user can review the same run as many times as they want, but only their newest review will be presented to the approval policy. This mechanism allows you to change your mind, very similar to Pull Request reviews.

Data Input»

Info

Note that this is just an example meant for informational purposes (JSON doesn't support comments by design). You can get a sample using the policy workbench.

This is the schema of the data input that each policy request will receive:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
{
  "reviews": { // run reviews
    "current": { // reviews for the current state
      "approvals": [{ // positive reviews
        "author": "string - reviewer username",
        "request": { // request data of the review
          "remote_ip": "string - user IP",
          "timestamp_ns": "number - review creation Unix timestamp in nanoseconds"
        },
        "session": { // session data of the review
          "login": "string - username of the reviewer",
          "name": "string - full name of the reviewer",
          "teams": ["string - names of teams the reviewer was a member of"]
        },
        "state": "string - the state of the run at the time of the approval"
      }],
      "rejections": [/* negative reviews, see "approvals" for schema */]
    },
    "older": [/* reviews for previous state(s), see "current" for schema */]
  },
  "run": { // the run metadata
    "based_on_local_workspace": "boolean - whether the run stems from a local preview",
    "branch": "string - the branch the run was triggered from",
    "changes": [
      {
        "action": "string enum - added | changed | deleted",
        "entity": {
          "address": "string - full address of the entity",
          "name": "string - name of the entity",
          "type": "string - full resource type or \"output\" for outputs",
          "entity_vendor": "string - the name of the vendor",
          "entity_type": "string - the type of entity, possible values depend on the vendor",
          "data": "object - detailed information about the entity, shape depends on the vendor and type"
        },
        "phase": "string enum - plan | apply"
      }
    ],
    "command": "string or null, set when the run type is TASK",
    "commit": {
      "author": "string - GitHub login if available, name otherwise",
      "branch": "string - branch to which the commit was pushed",
      "created_at": "number - creation Unix timestamp in nanoseconds",
      "hash": "string - the commit hash",
      "message": "string - commit message",
      "exist_on_tracked_branch": "boolean - true if commit with this hash exist on tracked branch"
    },
    "created_at": "number - creation Unix timestamp in nanoseconds",
    "creator_session": {
      "admin": "boolean - is the current user a Spacelift admin",
      "creator_ip": "string - IP address of the user who created the session",
      "login": "string - username of the creator",
      "name": "string - full name of the creator",
      "teams": ["string - names of teams the creator was a member of"],
      "machine": "boolean - whether the run was initiated by a human or a machine"
    },
    "drift_detection": "boolean - is this a drift detection run",
    "flags": ["string - list of flags set on the run by other policies"],
    "id": "string - the run ID",
    "runtime_config": {
      "before_init": ["string - command to run before run initialization"],
      "project_root": "string - root of the Terraform project",
      "runner_image": "string - Docker image used to execute the run",
      "terraform_version": "string - Terraform version used for the run"
    },
    "state": "string - the current run state",
    "triggered_by": "string or null - user, trigger policy, or dependent stack that triggered the run. For a dependent stack, this is set to the stack ID that triggered it.",
    "type": "string - type of the run",
    "updated_at": "number - last update Unix timestamp in nanoseconds",
    "user_provided_metadata": [
      "string - blobs of metadata provided using spacectl or the API when interacting with this run"
    ]
  },
  "stack": { // the stack metadata
    "administrative": "boolean - is the stack administrative",
    "autodeploy": "boolean - is the stack currently set to autodeploy",
    "branch": "string - tracked branch of the stack",
    "labels": ["string - list of arbitrary, user-defined selectors"],
    "locked_by": "optional string - if the stack is locked, this is the name of the user who did it",
    "name": "string - name of the stack",
    "namespace": "string - repository namespace, only relevant to GitLab repositories",
    "project_root": "optional string - project root as set on the Stack, if any",
    "repository": "string - name of the source GitHub repository",
    "state": "string - current state of the stack",
    "terraform_version": "string or null - last Terraform version used to apply changes",
    "worker_pool": {
      "id": "string - the worker pool ID, if it is private",
      "labels": ["string - list of arbitrary, user-defined selectors, if the worker pool is private"],
      "name": "string - name of the worker pool, if it is private",
      "public": "boolean - is the worker pool public"
    }
  }
}

Examples»

Tip

We maintain a library of example policies that are ready to use or that you could tweak to meet your specific needs.

If you cannot find what you are looking for below or in the library, please reach out to our support and we will craft a policy to do exactly what you need.

Two approvals and no rejections to approve an Unconfirmed run»

In this example, each Unconfirmed run will require two approvals - including proposed runs triggered by Git events. Additionally, the run should have no rejections. Anyone who rejects the run will need to change their mind in order for the run to go through.

Info

We suggest requiring more than one review because one approval should come from the run/commit author to indicate that they're aware of what they're doing, especially if their VCS handle is different than their IdP handle. This is something we practice internally at Spacelift.

1
2
3
4
5
6
7
8
package spacelift

approve { input.run.state != "UNCONFIRMED" }

approve {
  count(input.reviews.current.approvals) > 1
  count(input.reviews.current.rejections) == 0
}

Here's a minimal example to play with.

Two to approve, two to reject»

This is a variation of the above policy, but one that will automatically fail any run that receives more than one rejection.

1
2
3
4
5
package spacelift

approve { input.run.state != "UNCONFIRMED" }
approve { count(input.reviews.current.approvals) > 1 }
reject  { count(input.reviews.current.rejections) > 1 }

Here's a minimal example to play with.

Require approval for a task command not on the allowlist»

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
package spacelift

allowlist := ["ps", "ls", "rm -rf /"]

# Approve when not a task.
approve { input.run.type != "TASK" }

# Approve when allowlisted.
approve { input.run.command == allowlist[_] }

# Approve with two or more approvals.
approve { count(input.reviews.current.approvals) > 1 }

Info

Options for input.run.type include PROPOSED, TRACKED, TASK, TESTING, DESTROY

Here's a minimal example to play with.

Combining multiple rules»

Usually, you will want to apply different rules to different types of jobs. Since approval policies are attached to stacks, you will want to be smart about how you combine different rules. Here's how you can do that in a readable way, combining two of the above approval flows as an example:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
package spacelift

# First, let's define all conditions that require explicit
# user approval.
requires_approval { input.run.state == "UNCONFIRMED" }
requires_approval { input.run.type == "TASK" }

# Then, let's automatically approve all other jobs.
approve { not requires_approval }

# Autoapprove some task commands. Note how we don't check for run type
# because only tasks will the have "command" field set.
task_allowlist := ["ps", "ls", "rm -rf /"]
approve { input.run.command == task_allowlist[_] }

# Two approvals and no rejections to approve.
approve {
  count(input.reviews.current.approvals) > 1
  count(input.reviews.current.rejections) == 0
}

Here's a minimal example to play with.

Role-based approval»

Sometimes you want to give certain roles but not others the power to approve certain workloads. The policy below approves an unconfirmed run or a task when either a Director approves it, or both DevOps and Security roles approve it:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
package spacelift

# First, let's define all conditions that require explicit
# user approval.
requires_approval { input.run.state == "UNCONFIRMED" }
requires_approval { input.run.type == "TASK" }
approve           { not requires_approval }

approvals := input.reviews.current.approvals

# Let's define what it means to be approved by a director, DevOps and Security.
director_approval { approvals[_].session.teams[_] == "Director" }
devops_approval   { approvals[_].session.teams[_] == "DevOps" }
security_approval { approvals[_].session.teams[_] == "Security" }

# Approve when a single director approves:
approve { director_approval }

# Approve when both DevOps and Security approve:
approve { devops_approval; security_approval }

Here's a minimal example to play with.

Require private worker pool»

You might want to ensure that your runs always get scheduled on a private worker pool, and do not fall back to the public worker pool.

You could use an Approval policy similar to this one to achieve this:

1
2
3
4
5
6
7
package spacelift

# Approve any runs on private workers
approve { not input.stack.worker_pool.public }

# Reject any runs on public workers
reject { input.stack.worker_pool.public }

Here's a minimal example to play with.

You probably want to auto-attach this policy to some, if not all, of your stacks.