Muyideen Morenigbade

Engineering decisions

• #Each lab is its own directory with its own working code, not a written summary of what was done — the shell scripts, Terraform configs and Ansible playbooks run as committed.
• #The AWS capstone splits infrastructure into a dedicated VPC, RDS and EFS rather than one flat template — the same separation of concerns that shows up in the projects below.
• #Terraform work builds toward a reusable EC2 and security-group module rather than stopping at one-off resource blocks.
• #Monitoring is set up twice — Node Exporter on a bare Linux host, then as a Kubernetes DaemonSet — so the same tool is exercised in both environments it actually runs in.

Engineering decisions

• #Cloud login is a composite action, so OIDC federation is configured once and inherited — consuming repositories store no long-lived cloud credentials.
• #plan, apply and destroy are `workflow_call` workflows rather than copied YAML, so a fix to the pipeline reaches every repository that references it.
• #Consumers pin a tag rather than tracking the default branch, so a module change never lands in someone else’s apply unannounced.
• #Dependabot watches the action and provider versions the modules pin, so upgrades arrive as reviewable pull requests.

Engineering decisions

• #The CRD carries an OpenAPI schema, so an invalid spec is rejected by the API server rather than crashing the controller at reconcile time.
• #Child objects are created with owner references, so deleting a WebApp garbage-collects its Deployment, Service and HPA without the controller being involved.
• #A kopf timer re-reads pod health on an interval and syncs it to `status`, so the resource stays truthful between watch events.
• #RBAC is scoped to the resources the controller actually touches rather than granted cluster-admin.
• #pytest runs with coverage in CI before the image is built and published.

Engineering decisions

• #The pipeline’s last step is a git commit, not a cluster call — CI never holds kubectl credentials.
• #The deployed tag is a line in a committed manifest, so cluster state is readable from the repository and a rollback is a git revert.
• #Application secrets come from Key Vault rather than pipeline variables.
• #The cluster is composed from Terraform modules rather than one flat root configuration, so environments differ by inputs instead of by copies.

Engineering decisions

• #EKS and IAM are separate modules, so node and service-account roles can be reviewed on their own instead of buried in the cluster definition.
• #Delivery is an ArgoCD sync manifest, so the pipeline stops at the registry and the cluster pulls its own changes.
• #The workload is a real containerised Django app rather than a placeholder image, so the build and rollout paths are actually exercised.