Kubernetes as a Service
- In k8s cluster we have
- master nodes/control plane
- nodes
- Making master nodes highly available is our responsibilty and adding authorizations, addons etc is our responsibility
- K8s as a service is where Cloud provider manages Master nodes/cluster and HA. They also provide features for integrating k8s with other cloud services. The upgrades to the k8s cluster are easier to handle
- For nodes they charge usual virtual machine costs.
- Some clouds charge hourly for control plane
Azure Kubernetes Services
- Refer Here for launching the aks cluster
- For doing this as requirements
- azure cli to be installed
- azure cli to be logged in
- From Azure CLI instructions
az group create --name myResourceGroup --location eastus
az aks create -g myResourceGroup -n myAKSCluster --enable-managed-identity --node-count 2 --enable-addons monitoring --enable-msi-auth-for-monitoring --generate-ssh-keys
az aks install-cli
- Refer Here for the cluster ip service
- Lets create nginx-rs and nginx-cluster-ip
- Since we are in Azure and Azure has load balancer. Refer Here for the spec.
Resource Limits
- Refer Here
- Lets create a new replicaset with jenkins application with following
- lower limit => requests
- Memory => 256 Mi
- upper limit => limits
- Memory => 512 Mi
- lower limit => requests
- Refer Here for the changes and apply the manifest
Probes in K8s
- Refer Here
- k8s has 3 probes
- liveness probe:
- If this check fails, Pod will restart the container
- readiness probe:
- If this fails, this Pod will not be served by k8s service.
- startup probe
- liveness probe:
- k8s allows us to check this by
- sending a http(s) request
- sending a tcp request
- sending a grpc request
- send a linux command
Exercises
- Create a mysql pod and pass the necessary environment variables via k8s manifest
- Create a postgresql pod with environmental variables an try running with minimum memory restriction (128Ki)
- Try to write a manifest to run daemonset with alpine with sleep 1d
