Cloudland 2023
Series: [blog k8s]
Cloudland this year was quite some time ago, but better late then never:
Michael Friedrich spoke about Observability for Efficient DevSecOps Pipelines. While there is (obviously) some GitLab promotion, some issues to me seem common with other pipelines:
- slow pipelines
- unnecessary blocking/sequentiality
- missing cache for common transferred static data
- container registry, blobs, external artefacts
- observability principles need to be applied to CI/CD pipelines, too.
- telemetry!
- Don’t try to do everything in one step. Try to break into building blocks which can fail independent of each other.
- word of warning: sometimes sensitive data ends up in CI/CD pipeline logs. Build your policies around this
- a hint for transformations with OpenTelemetry, e.g. for long term analysis from cold storage
- eBPF might help again some supply chain attacks
- friendly reminder: Hashicorp Vault supports time based tokens