Automation & Orchestration

Why is automation important in data centers? What steps are involved in deploying a single Virtual Machine (VM)

Because many problems are caused by human error (manual mistakes). Even a “simple” task like deploying a VM has many steps and can easily go wrong.

From the data center perspective:

1. Choose a server to run the VM

2. Configure the hypervisor (software that runs VMs, e.g., like VirtualBox but in data centers)

3. Assign an IP address to the VM

4. Configure the network so traffic reaches the VM

5. Choose a remote disk server and allocate storage

6. Configure the hypervisor to send requests from the VM to the storage server

Which data center tasks can be automated?

1. Create and deploy virtual resources (VMs, containers)

2. Monitoring & accounting → measure system load (CPU, memory usage)

3. Optimization → balance load, reduce network traffic

4. Software updates (OS, apps, libraries)

5. Security management → firewalls, passwords, secrets

👉 Basically: deployment, monitoring, scaling, updates, security.

What are the 6 levels of automation?

• Level 0: Manual (no automation)

• Level 1: Automated setup/configuration

• Level 2: Automated monitoring (measure system)

• Level 3: Automated analysis & prediction (detect trends)

• Level 4: Automated root cause detection
  (Root cause = real underlying reason for a problem)

• Level 5: Automated problem fixing (self-healing)

👉 Higher level = system becomes more autonomous.

What is Infrastructure as Code (IaC)? What is Immutable Infrastructure? Benefits of IaC?

Managing infrastructure (servers, networks, storage) using code files, not manual setup.

You treat infrastructure like software:

• Version control (Git)

• Code review

• Testing

Instead of patching (modifying) a running server, you replace it with a new updated one (image).

Benefits of IaC?

• Consistency → no configuration drift
  (Drift = servers slowly becoming different over time)

• Reproducibility → identical Dev/Test/Prod environments

• Speed → automated provisioning

Tools:

• Terraform → provisioning (creating resources)

• Ansible → configuration management

What are containers and why are they popular?

• Lightweight isolated environments

• Use OS-level isolation (Linux features like namespaces & cgroups)

• Docker is most common

• Package application + required libraries together

👉 Advantage: “It works on my machine” problem disappears.

Why do we need orchestration? What does orchestration software do?

Because containers:

• Start quickly

• Have short lifetimes

• Often exist in many copies

Manually managing them is too complex.

Orchestration Software:

• Automates deployment

• Dynamic scaling (automatically increase/decrease instances)

• Coordination across servers

• Resilience (automatic recovery)

👉 If cloud = computer, orchestration = new operating system.

What is Kubernetes (K8s)?What does it mean that Kubernetes “reconciles declared state with actual state”? Main Capabilities:

Open-source container orchestration platform (originally from Google).
Manages containerized applications in distributed systems.

You declare what you want (e.g., 3 containers running).
Kubernetes constantly checks and makes reality match that goal.

• Service naming & discovery

• Load balancing

• Storage orchestration

• Optimized placement

• Auto recovery

• Config & secrets management

• Rollouts & rollbacks

How does Kubernetes handle service naming?

• Users access a service name, not individual containers.

• Kubernetes assigns:

  • Domain name or

  • IP address

• Kubernetes routes requests to one container.

👉 Users don’t see internal container details.

How does Kubernetes scale and balance load? How does Kubernetes manage storage?

• Starts multiple container instances

• Can limit max number

• Uses load balancer to distribute requests

(Load balancer = software that spreads traffic across instances.)

When starting a container, it connects to external storage automatically.

Types:

• Host file system

• File storage (NFS, EFS)

• Block storage (EBS)

(Block storage = raw disk-like storage.)

How does Kubernetes place containers?

• Does NOT request new cloud resources automatically.

• You define available nodes.

• Kubernetes assigns containers efficiently (bin packing).

Important terms:

• Node = server (physical or virtual)

• Cluster = group of nodes

How does Kubernetes handle container startup?

• Starts container → container initialization

• Waits until it’s ready → Delayed availability

• Then sends traffic → failure detection

  • probes each container with a user-defined health check

  • If container fails → it terminates and replaces it

(This prevents sending traffic to unready services.)

Why does Kubernetes separate configuration from container images?

• Allows changing configs without rebuilding images.

Configuration includes:

• Network connections

• Storage connections

• Passwords

• Tokens

• Encryption keys

👉 More flexible & secure.

How does Kubernetes update applications safely? (Rollout & Rollback)

• Gradually replace old containers with new version

• Control update speed

• Keep old containers running

• Rollback if problems occur

👉 Safe updates with minimal downtime.

What are Kubernetes limitations?

• No app-specific optimizations

• Does NOT build containers (Docker does)

• No event-passing middleware

• No built-in logging/monitoring system

👉 It manages containers, not everything around them.

Why is Kubernetes difficult for beginners?

• Very complex system

• Many components

• Constant evolution

• Overlapping features

• Poor documentation

• Inconsistent terminology

👉 Hard to fully master.

Define Cluster, Node, and Pod in Kubernetes.

• Cluster = A group of machines running containerized applications.

• Node = A single machine (physical or virtual) inside the cluster.

• Pod = Smallest deployable unit in Kubernetes.

👉 Think:
Cluster = whole system
Node = one server
Pod = smallest running unit

What is special about a Pod? How do containers inside a pod communicate?

• Often contains one container, but can have multiple.

• All containers in a pod run on the same node.

• Users communicate with the pod, not individual containers.

• Share the same IP address

• Communicate via localhost (like processes on same OS)

• Must not use the same port number

👉 Important: Pod = shared network identity.

What are the main components of a Kubernetes Pod YAML file?

A simplified structure:

• apiVersion → API version used

• kind: Pod → defines resource type

• metadata → name + labels

• spec → container configuration

• image → container image used

• containerPort → port exposed

👉 YAML = configuration file format (human-readable).

Labels are used for organizing and selecting pods.

What is an Init Container? Why are init containers useful?

A special container that runs before main containers start. → in initialization pod

All init containers must finish successfully before the main containers run.

They can:

• Check if required storage exists

• Test access to repositories

• Verify environment

• Stop pod if requirements are missing

👉 Guarantees: Either all main containers start correctly, or none start.

What types of nodes exist in Kubernetes?

1. Control Plane Node (Master Node)

   • Used by cluster owner

   • Runs control software

   • Creates & manages pods

2. Worker Node

   • Runs actual pods

   • Provides service to users

👉 Control plane = brain
Worker node = muscle

What are the main Kubernetes control plane components and their roles?

• API Server (kube-apiserver)
  Entry point of cluster. All communication goes through it.

• Scheduler (kube-scheduler)
  Assigns pods to nodes.

• Cluster State Store (etcd)
  Stores configuration + current state.

• Controller Manager
  Ensures system matches desired state
  (like a thermostat: compares actual vs target)

• Cloud Controller Manager
  Handles cloud provider interactions.

• kubectl
  Command line tool to manage cluster.

👉 API server acts like a switchboard.

How do control plane components communicate?

• All components communicate via the API server.

• API server acts as central hub.

• Stores state in etcd.

• Scheduler & controllers interact through API server.

👉 No direct chaos — everything goes through API server.

What software runs on a worker node?

1. kubelet

   • Runs and monitors pods

   • Communicates with control plane

2. kube-proxy

   • Configures networking (iptables)

   • Handles traffic routing

3. Container Runtime

   • Runs containers (e.g., Docker)

👉 kubelet = local manager
kube-proxy = network manager
container runtime = actual executor

What additional Kubernetes features support scaling and management?

• Replicas

  • Define number of pod copies

• Deployments

  • Automate scaling using intent-based approach
    (You define desired state, system matches it)

• StatefulSets

  • For stateful apps (with memory/data)

  • Each pod gets permanent unique ID

• DaemonSet

  • Runs one pod per specified node

  • Used for background services

• Garbage Collection

  • Removes terminated objects

  • Uses dependency rules

1) What is TTL Controller? 2) What is a Job facility? 3) What is a CronJob facility? 3) What is a Services facility?

1) Deletes finished resources after a defined time.

TTL = Time To Live.

2) Creates specific number of pods and monitors them.
Can stop after certain completion condition.

Example: stop when one pod finishes.

3) Runs jobs periodically (like Unix cron).

4) Group multiple pods under one access policy.
Pods can restart without affecting service.

👉 Important for microservices.

What are advantages and disadvantages of orchestration?

dvantages:

• Higher efficiency

• Better resource coordination

• Fewer human errors → cost savings

Disadvantages:
− Cascading failures (system-wide problems)
− Run-away resource usage
− Larger security attack surface
− High complexity

What orchestration options does AWS provide?

1. Amazon EKS (Elastic Kubernetes Service)

   • Managed Kubernetes

   • Worker nodes on EC2 or Fargate

   • Costs $0.10 per hour

2. Amazon ECS (Elastic Container Service)

   • AWS proprietary orchestration

   • Runs on EC2 or Fargate

   • No extra charge

👉 EKS = Kubernetes
ECS = AWS-native solution