Managing Deployments

DevRamps provides several operations for controlling active and past deployments. You can cancel, retry, rollback, or roll forward from the DevRamps dashboard.

Permissions

All deployment operations (cancel, retry, rollback, roll forward, stop, resume) are available to any member of your DevRamps organization. Role-based access control for deployment operations is planned for a future release.

All deployment operations are recorded in the events audit trail, including who performed the action and the reason provided.

Cancel a Deployment

Cancel an in-progress stage deployment to stop it immediately.

When to use: A deployment is running and you realize the code has a problem, or you need to stop it for any reason.

How it works:

1. Navigate to the pipeline in the dashboard.
2. Click on the in-progress stage.
3. Click Cancel.
4. The stage transitions to Cancelling, then Cancelled.
5. Any running step is terminated. Remaining steps are skipped.
6. The next stage does not start.

What happens to infrastructure: If a Terraform step was in the middle of applying, the partial apply state is preserved. The next deployment will reconcile from the current state.

Retry a Failed Stage

Re-run a failed stage with the same code revision.

When to use: A step failed due to a transient issue (network timeout, rate limit, temporary service outage) and you want to try again without pushing new code.

How it works:

1. Navigate to the failed deployment in the dashboard.
2. Click on the failed stage.
3. Click Retry.
4. The stage re-executes from the beginning with all steps.
5. If successful, the pipeline continues to the next stage.

Note: Retry re-runs all steps in the stage, not just the failed one. This ensures infrastructure and deployment state are consistent.

Emergency Rollback

Roll back to a previously deployed revision.

When to use: A deployment succeeded but the new version is causing issues in production. You want to revert to the last known good version.

How it works:

1. Navigate to the pipeline in the dashboard.
2. Click Emergency Deploy. (Despite the name, this button is used for both rollback and roll forward operations.)
3. Select Rollback as the deployment type.
4. Choose the target revision (the revision you want to roll back to).
5. Optionally specify steps to skip during rollback.
6. Provide a reason for the rollback (recorded in the audit trail).
7. Click Execute.

What happens:

• If the target stage is currently deploying, the deployment is cancelled first.
• The stage status changes to Rolling Back.
• The selected revision's steps execute against the stage. All deployment steps re-run (not just the failed one).
• Infrastructure synthesis (Terraform) is skipped by default during rollback to keep it fast and avoid unintended infrastructure changes.
• The stage and the next stage are blocked from automatic promotions until you manually unblock.

Requirements: The target revision must have been previously deployed to the Build stage (artifact builds must exist).

Emergency Roll Forward

Deploy a newer revision to a specific stage, skipping the normal promotion flow.

When to use: You've pushed a fix and want to deploy it directly to the affected stage without waiting for it to progress through earlier stages.

How it works:

1. Navigate to the pipeline in the dashboard.
2. Click Emergency Deploy.
3. Select Roll Forward as the deployment type.
4. Choose the target revision.
5. Provide a reason.
6. Click Execute.

What happens:

• Similar to rollback, but deploys a newer revision instead of an older one.
• The stage status changes to Rolling Forward.
• The stage is blocked from automatic promotions until manually unblocked.

Stop a Pipeline

Pause all deployments for a pipeline entirely.

When to use: You need to halt all deployment activity -- for example, during a maintenance window or incident investigation.

How it works:

1. Navigate to the pipeline in the dashboard.
2. Click Stop Pipeline.
3. Provide a reason.
4. All in-progress deployments continue to completion, but no new deployments start.
5. New pushes are acknowledged but not deployed.

Resume a Pipeline

Resume a stopped pipeline.

How it works:

1. Navigate to the stopped pipeline.
2. Click Resume Pipeline.
3. Any queued deployments begin executing.
4. New pushes trigger deployments as normal.

Summary of Operations

Operation	Stage State Required	Result
Cancel	In Progress, Rolling Forward, Rolling Back	Stage → Cancelled
Retry	Failed	Stage re-executes from the beginning
Rollback	Any (cancels if in progress)	Stage → Rolling Back → Rolled Back
Roll Forward	Any (cancels if in progress)	Stage → Rolling Forward → Rolled Forward
Stop Pipeline	Any	Pipeline paused, no new deployments
Resume Pipeline	Stopped	Pipeline resumes, queued deployments start
Patch Deploy	Not in patch mode	Deploys specific commit, pauses normal deployments
Release Patch	In patch mode	Exits patch mode, resumes normal deployments