case study / Faradah (consulting) · 2025–now
Six apps, one secure foundation
Six applications moved from scattered providers onto one GCP foundation with reusable Terraform and keyless CI/CD, cutting cloud costs by more than 60%.
- Client
- Faradah (consulting)
- Period
- 2025–now
- Role
- DevOps Consultant
- Result
- 60%+ lower cloud costs
- Stack
- GCP, Terraform, GitHub Actions, Workload Identity Federation, PAM/IAP, Secret Manager, WAF
In short
- Moved 6 applications from different providers and habits onto one Terraform-managed GCP foundation.
- Replaced per-repository pipelines with central, reusable GitHub Actions workflows.
- Removed every stored cloud key from CI/CD with GitHub OIDC and Workload Identity Federation.
- Made server access private (IAP only) and temporary (Privileged Access Manager).
- Cut cloud infrastructure costs by more than 60%.
Context
Faradah is a Saudi company building several internal and client products. Like many teams that grow one product at a time, each application had arrived with its own provider, deployment script, credentials, backup method and monthly bill. Everything ran, but running it as a portfolio was slow and risky. I work with them as a freelance DevOps consultant.
Constraints
- Applications were spread across different cloud providers with inconsistent CI/CD, access and cost patterns.
- Products could not all be moved at once or forced to become technically identical.
- Long-lived cloud keys and SSH access had to go, without slowing delivery.
Decisions
A shared platform contract
Standardizing did not mean one project or one network for everything. It meant every product gets the same set of capabilities from Terraform:
- project foundation, networking and load balancing behind a WAF,
- compute and managed database patterns,
- public and private storage with automated backups and retention,
- central secrets in Google Secret Manager,
- monitoring, audit metrics and security alerts,
- keyless CI/CD identity, IAP-only administration and time-bound privileged access.
Scale, domains, data and availability stay configurable per product. The security and delivery baseline does not.
Migrate in stages
Each product followed the same path: inventory the running application, data, domains, credentials and deployment process; recreate the foundation in Terraform; set up image builds and deployments; move secrets into Secret Manager; check backups, monitoring and access; run old and new side by side where needed; cut over only after application and operational checks pass; retire the old resources after an agreed observation period. Every migration made the next one faster.
Central GitHub Actions workflows
Instead of every repository carrying its own YAML, one central repository holds reusable workflows for:
- building Docker images with layer caching and pushing them to Artifact Registry,
- deploying Docker Compose workloads to GCE,
- reaching private VMs through OS Login and IAP,
- pulling environment configuration from Secret Manager and restarting services in a controlled way.
Application repositories call these with their own inputs. A fix to authentication, caching or deployment checks is made once and every product gets it.
No stored cloud keys
The workflows authenticate with GitHub OIDC and Google Workload Identity Federation:
- GitHub Actions issues a short-lived OIDC token for an approved repository and workflow.
- Google validates it through a Workload Identity Pool provider.
- The workflow impersonates a dedicated service account that can only do the required build or deploy steps.
No service-account JSON key is stored in GitHub, repository access is limited with attribute conditions, and audit logs show which identity did what. The same identity reaches the private deploy target through OS Login and IAP, so no SSH keys live in the repositories either.
Private, temporary access
Firewall rules block public SSH and only allow the IAP proxy path to approved instances. Google Privileged Access Manager grants administrative IAM roles for a limited time, and the request, grant and activity are logged. Access now depends on identity and IAM, not on who can reach a network port, and no product needs a public bastion.
Cost and security feedback
With everything on one foundation, costs became comparable. I reviewed compute sizing, database tiers, storage, networking and unused resources, and cut what did not affect availability. The platform also raises signals for health failures, database pressure, changes to firewall and IAM-sensitive resources, secret access patterns and public endpoint availability.
Results
- 6 applications on one standard GCP foundation, all managed with Terraform.
- Cloud infrastructure costs reduced by more than 60%.
- Zero long-lived cloud keys in CI/CD; every deployment uses short-lived identity.
- Server access is private through IAP and temporary through Privileged Access Manager.
- Backups, secrets and monitoring are part of the foundation for every product, not an afterthought.