Rescans and one-time scans
You just merged a fix and want proof it worked. Or a new CVE dropped and you want to know now, not after the next discovery. There are several buttons for this, and picking the wrong one means either waiting on work you didn't need or missing what you wanted.
Which one do you need?
| You want to | Use | From |
|---|---|---|
| Check for new vulnerabilities and scan new snapshots | Rescan | A repository or image |
| Run every scanner again, regardless of what ran before | Force rescan | A repository or image |
| Scan a resource that tracks no refs | Scan now | The resource |
| Scan one resource from the queue, ahead of everything else | Adhoc scan | Scan Center, Running view |
| Run a failed or cancelled scan again | Retry | Scan Center, History view |
| Stop a running or waiting scan | Cancel | Scan Center, Running view |
Start with Rescan. It does the least work that answers the question.
Prerequisites
scans:runto rescan, force rescan, scan now, or retry. For a resource, you need it on the organization that owns it, or on one whose product uses it.scans:cancelto cancel a scan, anddiscovery:cancelto cancel discovery.
Rescan
- Open a repository or image.
- Select Rescan.
Dwaar queues one run per tracked ref at high priority, and decides per ref:
| The ref's snapshot | What runs |
|---|---|
| Had a completed full scan | Vulnerability re-check |
| Has never been fully scanned | Full scan |
| Full scan left out a category that's on now | Full scan |
| Already has a run queued or running | Nothing. It's skipped |
You should see a message such as "main: vulnerability re-check · release/5.3: full scan". If every ref is already busy or hasn't resolved yet, the message says there's nothing to rescan.
A re-check only compares the last full scan's SBOM with current vulnerability data. That's what you want when a new CVE drops. It won't notice a fix you just merged until discovery has seen the new commit, so run discovery first, from Upcoming in Scan Center.
Force rescan
Force rescan runs every enabled scanner again on every tracked ref, even where a full scan exists. The scanners' own caches still apply, so it isn't a guarantee that each tool repeats all its work.
Reach for it when a result looks wrong and you want a clean run. Don't make it a habit, because it's the slowest answer to most questions.
Scan now
Repositories and images track refs. Resource kinds that don't track refs offer Scan now instead, which queues a one-time full scan ahead of the queue.
Two kinds offer no scan action at all: packages, and the cloud resources of an AWS account, Vercel team, Railway workspace or Cloudflare account, because their connection's posture check covers them.
Adhoc scan
This is Rescan's cousin in Scan Center, for when you're watching the queue.
- Open Scan Center, on the Running view.
- Select Adhoc scan.
- Pick a resource.
- Select Scan now.
The drawer tells you what will run: a full scan of the default ref at its current snapshot, with every scanner enabled for that resource type. No schedule is created, and discovery keeps planning its scans as before.
Retry
In the History view, a failed or cancelled run has a Retry action. It queues a new run of the same snapshot and mode. Read the run's error first. Retrying only helps when the problem was temporary.
Cancel
- Open Scan Center, on the Running view.
- Select the cancel action on the run.
- Confirm.
The scan stops and is marked cancelled, and you can retry it from History. The same action cancels a discovery in progress.
What you should see
Whichever you used, the run appears in the Running view and on the resource's Scans tab. When it completes, findings update, including moving anything no longer seen to Fixed.
Try this next
Prove a fix end to end once, so you trust it later. Merge a small dependency bump, run the connection's discovery from Upcoming, and watch the new snapshot get a full scan. Then open the findings list and see the old finding go to Fixed.