Skip to content
Notifications
Clear all

Rolled out Trivy to 150 developers - what broke in our K8s pipeline

1 Posts
1 Users
0 Reactions
2 Views
(@harryj)
Estimable Member
Joined: 6 days ago
Posts: 82
Topic starter   [#19391]

We just finished a company-wide rollout of Trivy for container scanning in our CI/CD pipelines. Overall, it's been great for catching real vulns early. But scaling to 150 devs exposed a few painful snags in our Kubernetes deployment flows.

The main issues we hit:
* **Timeout failures in CI**: Our larger Node.js microservices hit Trivy's default 5m timeout during the `--security-checks vuln,secret` scan. Had to bump it up significantly.
* **False positives on distroless**: The "missing" package manager in distroless base images caused some confusing `UNKNOWN` types and flagged packages that weren't actually present.
* **Pipeline speed complaints**: The biggest one. Developers on older, smaller runner pods saw their build times jump by 3-4 minutes. We're now looking at caching the vulnerability DB or using a client-server setup.

Had to adjust our Trivy action config quite a bit to get it stable. Anyone else run into scaling pains with Trivy in a big dev environment? Specifically around pipeline performance or distroless images?

~hj


Automate the boring stuff.


   
Quote