DevOps Continuous Delivery and Continuous Deployment Syllabus

In Spinnaker, what is the 'bake' stage?

A stage that produces an immutable machine image (e.g., an AWS AMI) from a source artifact, typically using HashiCorp Packer. The baked image is then deployed, supporting immutable infrastructure.

What is a 'server group' in Spinnaker?

The basic deployable unit representing a set of identical running instances (e.g., an auto-scaling group) created from a single image/configuration. Multiple server groups make up a cluster.

What deployment strategies does Spinnaker provide out of the box?

Red/Black (Spinnaker's name for blue-green), rolling red/black, canary (via automated canary analysis), and highlander. These automate safe progressive rollouts.

What is Spinnaker's Automated Canary Analysis (ACA / Kayenta)?

A component (Kayenta) that statistically compares metrics from a canary deployment against a baseline and automatically scores the canary to decide whether to promote or roll back the release.

What is Argo CD?

An open-source, declarative GitOps continuous delivery tool for Kubernetes. It continuously monitors a Git repository of desired manifests and synchronizes the cluster state to match, treating Git as the single source of truth.

What deployment model does Argo CD follow: push or pull?

A pull-based model. The Argo CD controller runs inside the cluster and pulls the desired state from Git, rather than an external CI server pushing changes into the cluster. This is a core GitOps principle.

What is GitOps?

An operational framework where the entire desired system state (infrastructure and applications) is declaratively stored in Git, and an automated agent continuously reconciles the live environment to match Git. Git becomes the single source of truth and the audit log.

In Argo CD, what does it mean for an application to be 'OutOfSync'?

The live state of resources in the cluster does not match the desired state declared in the Git repository. Argo CD detects the drift and can either alert or automatically sync to reconcile it.

What is the difference between manual sync and automated sync in Argo CD?

Manual sync requires a user to trigger reconciliation after detecting drift; automated sync makes Argo CD automatically apply Git changes to the cluster as soon as drift is detected, enabling hands-off continuous deployment.

What is 'self-heal' in Argo CD?

An automated-sync option where Argo CD automatically reverts any out-of-band changes made directly to the cluster, restoring the state defined in Git and preventing configuration drift.

What is the 'App of Apps' pattern in Argo CD?

A pattern where a single parent Argo CD Application declaratively manages multiple child Applications, allowing an entire suite of apps and their configurations to be bootstrapped and managed from one Git-defined root.

Which Kubernetes manifest tools does Argo CD natively support?

Plain Kubernetes YAML/manifests (directories), Helm charts, Kustomize, and Jsonnet, plus custom config-management plugins.

What is Octopus Deploy?

A commercial deployment automation and release management tool that handles the deployment and operations side of CD across many environments and technologies (Windows, Linux, Kubernetes, cloud), commonly paired with a separate CI build server.

How does Octopus Deploy typically fit with a CI server like Jenkins or TeamCity?

The CI server handles building and testing to produce an artifact/package; Octopus Deploy then takes that package and manages its deployment, promotion across environments, and release orchestration. CI builds, Octopus deploys.

In Octopus Deploy, what is a 'project'?

A container that defines how an application is deployed: it holds the deployment process (steps), variables, and configuration used to create and deploy releases of that application.

In Octopus Deploy, what is a 'release' versus a 'deployment'?

A release is a snapshot/version of the deployment process and packages, including version numbers and variables, that is created once. A deployment is the act of executing that release into a specific environment (e.g., Dev, then Test, then Prod).

What is a 'lifecycle' in Octopus Deploy?

A defined set of rules controlling the order in which a release is promoted through environments (e.g., Dev → Test → Staging → Production) and the conditions/approvals required to advance between phases.

What is the role of the 'Tentacle' agent in Octopus Deploy?

Tentacle is a lightweight agent installed on target machines that the Octopus Server communicates with to execute deployment steps. It supports Listening mode (server connects to target) and Polling mode (target connects to server).

What is a deployment strategy?

A defined approach for releasing a new version of software to replace or coexist with the old version, designed to manage the trade-offs between deployment speed, risk, downtime, and resource cost.

Describe the 'recreate' (basic/highlander) deployment strategy and its main drawback.

The old version is fully stopped and then the new version is started. It is simple and uses no extra resources, but it causes downtime during the switch and offers no easy instant rollback.

Describe the rolling (ramped) deployment strategy.

The new version is gradually rolled out by incrementally replacing instances of the old version one batch at a time until all instances run the new version. It avoids downtime and extra full-capacity cost but creates a period where both versions serve traffic and rollback is slower.

What is Blue-Green Deployment?

A strategy using two identical production environments, 'Blue' (current live) and 'Green' (new version). The new release is deployed and tested on the idle environment, then traffic is switched over all at once, enabling near-zero downtime and instant rollback.

In Blue-Green Deployment, how is a rollback performed?

By switching the router/load balancer to point traffic back to the previous environment (the old Blue/Green), which is still running and unchanged. Rollback is therefore near-instantaneous and low-risk.

What is the main resource cost/disadvantage of Blue-Green Deployment?

It requires running two full production-capacity environments simultaneously, roughly doubling infrastructure cost during the cutover. It can also be complex when handling database schema changes and stateful data shared between environments.

How is traffic switched in Blue-Green Deployment?

Typically by reconfiguring a load balancer, router, or DNS/service to redirect all user traffic from the old environment to the new one in a single cutover, so users move atomically from Blue to Green.

What is a Canary Release?

A strategy that gradually rolls out a change to a small subset of users/servers (the 'canary') first, monitors its behavior, and then incrementally increases traffic to the new version if it performs well, or rolls back if problems appear.

Why is the canary release named after a 'canary'?

It alludes to the canary in a coal mine: a small, expendable early indicator of danger. The small canary group surfaces problems with a new release before they affect the entire user base.

What is the key difference between Blue-Green and Canary deployments?

Blue-Green switches all traffic at once between two full environments. Canary shifts traffic gradually in increasing percentages to the new version, exposing only a fraction of users at a time and allowing fine-grained, metric-driven rollout.

If a canary release routes 5% of traffic to the new version, how is the traffic split expressed, and what guides increasing it?

The split is $95\%$ baseline to $5\%$ canary, i.e., a ratio of $\frac{5}{100}$ of users on the new version. Increases (e.g., to $25\%$, then $50\%$, then $100\%$) are guided by monitoring error rates, latency, and other health metrics meeting acceptance thresholds.

What is A/B testing and how does it differ from a canary release?

A/B testing routes users to different versions to compare business/behavioral metrics and make a product decision; the goal is measuring user response. A canary release routes a small fraction to a new version to validate stability/safety before full rollout; the goal is risk reduction. They use similar traffic-splitting mechanics but have different objectives.