Discovery and scans
Most scanners make you configure a schedule per tool, per repository. Then a repository appears that nobody added, and it goes unscanned for a year.
Dwaar has exactly one schedule: each connection's discovery frequency. Discovery lists what the connection can see, and that listing is what plans the scanning. A new repository gets picked up because discovery found it, and nobody had to remember.
What a discovery run does
For each tracked ref on each resource, a discovery plans one of two things, depending on the triggers set on the connection's Scanning tab:
| Situation | What gets queued | Trigger shown |
|---|---|---|
| A snapshot that was never fully scanned | A full scan | discovery |
| An unchanged snapshot | A vulnerability re-check, at low priority | schedule |
The second row is the clever part. Vulnerability databases change even when your code does not, so an unchanged commit can still have a new problem. A re-check matches the last full scan's SBOM against current data without running the scanners again.
What a run is
A scan is one run on one snapshot of a resource: a commit of a tracked branch or tag, or an image index or manifest of an image tag.
There is no per-scanner run and no per-scanner cadence. Scan Center keeps the tool, version and on or off state per category, and each run keeps the choices it was queued with.
| Scan mode | What it does |
|---|---|
| Full scan | Runs every enabled scanner on the snapshot. |
| Vulnerability re-check | Runs only the data pipeline, on the last full scan's SBOM. |
| Posture check | Checks a connection's organization settings, AWS account, Vercel team, Railway workspace or Cloudflare account. It has no resource. |
A run moves through stages in order: queued, starting, fetching, analyzing, processing, and then completed, failed or cancelled.
When a run is waiting
A queued run with a blocked reason is waiting because your tenant's running scans are at the cap. It will start on its own, so there is nothing to fix.
Try this next
Push a commit to a tracked branch, then watch for the new snapshot after the connection's next discovery. Seeing one go through the queue makes this page concrete.