esc

Docs

Approvals and policies

How approvals open, who can decide them, the Default policy, freeze windows, break-glass and approving from chat.

How approvals open

  • From a test run. When a test pipeline passes, DeployNX opens an approval for that build.
  • From a green stage build. On GitHub connected through the one-click App, each new green stage build opens a promote-to-prod approval instantly. On Bitbucket, GitLab and GitHub repositories connected with a token, DeployNX checks every 30 minutes by default (or at the interval your organization sets in Settings › General) and opens it automatically, and Check now on Releases › Promoting looks right away. Either way, the repository needs Auto-detect green stage builds turned on in its Stage → prod promote settings. A newer request closes the older one it replaces.

Each approval shows the evidence an approver needs: the test result, the stage build and its health, the release tag, the commit and who asked. Approvals expire if nobody decides them: after 3 days by default, or after the time your organization sets in Settings › General, for requests made after the change.

Who can decide

Role Can request Can approve stage Can approve prod
Owner Yes Yes Yes
Admin Yes Yes Yes
Member Yes Yes No

That is the default. In Settings › Environments, each environment sets who approves releases to it: members, admins or owners. A production-tier environment can’t be set to members. Set to admins, it leaves the decision to the Default policy’s Owners and admins approve prod rule, so if you switch that rule off, members can approve it too. An environment can also name one of your BizStats Hub teams as its approver team (teams are kept on BizStats Hub; Settings › Teams lists them): then only the team’s members approve releases there, and they still need the role.

By default, nobody approves their own production request unless no one else can approve there (a rule in the Default policy). In Settings › General or Policies you can make that never or always. A decided approval can’t be decided again, and rejecting one needs a reason. Scutor, the assistant, can’t approve anything.

The Default policy

Every organization has a Default policy that can’t be deleted. It requires:

  • one approval before production;
  • an approver who is not the requester, unless no one else can approve there;
  • a healthy release in the environment before production on the repository’s route (Stage, unless the repository maps other environments before Production);
  • no force-completed runs going to production.

You can switch individual rules off from Policies, and scope the healthy-environment and force-completed rules to the tiers or environments you pick. If either of the first two is off, a prod release no longer needs a named approver who isn’t the requester. Each request shows how many of the policy’s requirements it meets.

Two more rules are off until you turn them on, both for production unless you pick other tiers or environments:

  • Ships only from: the branches and tags a release may come from, such as main, release/* or v*. A request for a release from another branch shows the rule blocking it, and approving or deploying it is refused.
  • Deploys only: the days and hours deploys may run, in your organization’s time zone (Settings › General). Outside the window, approving or deploying is refused with the time the next window opens; a request can still wait for it.

Rollbacks are exempt from both. Step-up re-authentication for production is planned.

More than one approval

In Settings › Environments, Approvals sets how many different people approve each release to an environment, from one to five. A request keeps waiting until the last approval, which ships it. Each person approves once, any rejection rejects the request, and the inbox and the decision room show who approved and how many more are needed. Settings warns when fewer people can approve there than you ask for.

Freeze windows

Use Freeze now with a reason and a thaw time, or schedule a freeze in advance. A freeze covers what you pick: all production, all non-production, whole tiers or particular environments. While a freeze is active, releases to those environments are held and everyone sees a banner. Held requests become ready again after the thaw. Rollbacks stay allowed during a freeze.

Break-glass

In an incident, open a break-glass session for one environment with an incident link and a reason, for up to four hours, and type the environment’s short name (production for Production) to confirm. On that environment it bypasses what you choose: freezes and the deploy schedule, the approval requirement (all the approvals it asks for), the allowed source branches, or any mix, so a prod release can go out without an approver. It alerts your approver channels and leaves a review item that an owner or admin must close.

Notifications

DeployNX sends approval requests, failures and release updates to Slack, Google Chat and email. Choose channels and routing for your organization in Settings › Integrations, check the delivery log, and send a test message. Each person can set their own email preferences. In Settings › General, owners and admins can have the people who can approve a request reminded once, by email and in your chat channels, when it has waited a number of hours, and choose how long the delivery log is kept.

Approve in chat

Slack and Google Chat messages carry Approve and Reject buttons with the same checks as the app: the approval must still be pending, your role must allow it, the Default policy’s no-self-approval rule applies, and rejecting needs a reason. If your chat account isn’t linked to DeployNX yet, the button opens the approval in the app instead.

NextReleases and rollback