DevOps Classroom notes 16/Feb/2025

Label Selectors

• Refer Here for official docs
• Equality based selector
• Set based selector

Controllers in k8s

• Controllers in k8s indirectly create pods
• Controllers:
  • Replicaset: Scaling Pods
    • Desired State: Number of pods
  • ReplicationController:
    • Desired State: Number of pods
  • Deployment
    • Desired State: Number of pods via Replicasets to deploy apps with zero downtime
  • Job
    • Desired State: Run the Pod to Completion
  • CronJobs
    • Desired State: Run the Pod to Completion on Schedule
  • Statefulset: TBD
  • Daemonset: TBD
• When we are dealing with Controllers we will have some extra elements to consider
  • template: This generally refers to pod specification
  • matching labels and selectors
• Overview of Deployment
  

Replicaset

• official docs
• Lets write a replicaset which creates 3 nginx Pods

---
apiVersion: apps/v1
kind: ReplicaSet
metadata:
  name: nginx-rs
  labels:
    app: nginx
    env: dev
spec:
  minReadySeconds: 5
  replicas: 3
  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
        version: v1
        env: dev
    spec:
      containers:
        - name: web
          image: nginx

• Refer Here for changes
• Kuberenetes replicaset will ensure there are desired number of pods with matching labels.
• Even if the exising pods have matching labels it doesnot try to create more that replicas
• Watch classroom video for further commands
• Note: When using Replication Controller while writing label selectors you can do only equality based selection
• Find out how to write a set based expression for matching labels in k8s manifests for replicaset and write a replicaset yaml for httpd with 5 replicas. Refer Here for changes

Accessing applications in k8s

• Each Pod gets an IP Address, Pods get scaled an replaced, directly accesing pods by ip might not be a good idea
• K8s has a component running for service discovery with DNS functionality mostly coredns
• In linux machines we have /etc/resolve.conf whats the purpose of that?

The /etc/resolv.conf file is a plain-text configuration file that specifies how the Domain Name System (DNS) resolver operates on a system[1]. It contains information that allows applications to translate human-friendly domain names into the numerical IP addresses needed to access resources on a network or the internet[1]. This translation process is known as address resolution[1]. The file is typically located in the /etc directory[1].

The resolv.conf file is used to configure hostname resolution and is commonly used to manage DNS requests on Linux systems[2]. It can be manually configured by a system administrator or automatically populated by network configuration and management tools[2].

Key purposes and functions:
* Nameserver specification The nameserver directive specifies the IP addresses of domain name servers that the resolver can query[1][5]. Multiple nameserver lines can be specified, but each line should only have one IP address, up to a maximum of three[2][5]. The resolver queries the DNS servers in the order listed[2]. If no nameserver entries are present, it defaults to 127.0.0.1[2].
* Search domains The search directive specifies default search domains that are used to complete a given query name to a fully qualified domain name when no domain suffix is supplied[1]. For example, search example.com local.test configures the resolver to try somehost.example.com and somehost.local.test[1].
* Lookup order The /etc/nsswitch.conf file specifies the lookup order for hostnames, determining whether the system consults DNS or the /etc/hosts file first[3].

In systemd-based Linux distributions, /etc/resolv.conf is often a symbolic link to /run/systemd/resolve/stub-resolv.conf[1]. The resolvconf program on FreeBSD and other Unix systems manages the resolv.conf file[1]. On Debian systems, both openresolv and resolvconf can be used to control the alteration of /etc/resolv.conf by networking programs[4].

Citations:
[1] https://en.wikipedia.org/wiki/Resolv.conf
[2] https://wiki.gentoo.org/wiki/Resolv.conf
[3] https://serverfault.com/questions/118923/difference-between-etc-hosts-and-etc-resolv-conf
[4] https://wiki.debian.org/resolv.conf
[5] https://www.baeldung.com/linux/etc-resolv-conf-file
[6] https://superuser.com/questions/570082/in-etc-resolv-conf-what-exactly-does-the-search-configuration-option-do
[7] https://www.youtube.com/watch?v=hflrZcYzcbE
[8] https://docs.oracle.com/en/operating-systems/oracle-linux/6/admin/about-etc-resolve.html

---

Answer from Perplexity: pplx.ai/share

• sample /etc/resolve.conf in a container running in a Pod
  
• Name resolving can be enabled with services and service acts a layer 4 load balancer
• Each Service gets an ip address, But this is an virtual ip which means it never changes till you kill the service
• Service Official docs
  
• Service in k8s can be of following types
  • Cluster IP: This is an internal ip address which works within k8s cluster
  • Node Port: The service is exposed on every node in k8s cluster on a specific port
  • Load Balancer: This generally works with cloud services, this creates a cloud native load balancer (rule)
  • External Name: This is used for DNS records to be used externally
• Port definitions
• Services support only equality based selection for labels

Lets expose httpd containers using service

• Lets run the replicaset to create 5 pods of httpd
• lets create an internal service i.e. type is ClusterIP Refer Here
• Now lets create one more service with type Node Port