Scanning the container images deployed to my cluster used to be manual. Now it happens automatically every night. 🐱🏍
I use Trivy to scan container images. I wrote about scanning my GitOps repo for images earlier here. Basically, there's a fancy grep that matches
image: (<name>) and that name is sent to Trivy.
My GitOps repo is on GitHub at kasuboski/k8s-gitops. It seemed natural to run the scan periodically using GitHub Actions. The scan will happen on every push and every night.
I needed a way to exclude an image that wasn't able to scan on an x86 host. The one liner from my previous post needed a
grep -v to exclude certain patterns.
I had a lot of trouble getting
ack configured in a runner. I ended up making a docker image that downloads trivy, finds the images, and scans them.
This image has its own repo kasuboski/trivy-scan-dir. If you just want to scan a repo you can run
docker run -it --rm -v /path/to/yaml:/gitops -e EXCLUDED='no/scan also/noscan' kasuboski/trivy-scan-dir.
To run this in a workflow, add the below step.
- name: Scan Images uses: docker://kasuboski/trivy-scan-dir:latest env: EXCLUDED: 'no/scan also/noscan'
My full workflow can be found in kasuboski/k8s-gitops. It triggers on
push to yaml files.
Workflow Dispatch lets you run the workflow manually from the GitHub Actions UI. This was really convenient for testing. The cron schedule runs every morning at 4:03am.
This workflow has alerted me to multiple vulnerabilities. If the workflow fails, I get an email and then can look into updating the image.
The results even look pretty decent in the GitHub app so I can tell which images I need to be worried about. An example failing run is shown below.
I still want to add something in cluster to enforce only the images I want are running. Finding the images to scan also needs to be more robust. For instance, some images only show up once manifests are rendered.