Need for K8’s / Why Docker is insufficient to manage Containers (neglect Dock Swarm :) )

As we scale up and up (Horizontally), the containers will increase or our services might even scale up gradually leading to a complex architecture. In this case Docker →

→ Containers are ephemeral , means that a container can die anytime and revive anytime. The life of containers is very short, if there are lack of resources or any issue with configs. → Is a Single Host based : For instance : Say we have about 50 Containers serving for our app, then all of sudden there is a insurge of huge traffic towards Container - 1,2 . Then Linux will have to kill some of the unused Processes, but yeah it will be based on priority given to the process. But this shouldn’t happen na? Like even this deleted container might be vital later on. ” So, Docker is scoped to a single host, and because of which the nature of a container might impact others’ too and if so then there is no way that this container might come up-and-running. "

→ Isn’t having Auto-Healing Auto-Healing : Is a feature where, without manual intervention our containers come up-running automatically EVEN IF THERE WAS A FAILURE. , it means that if a container fails or crashes, it will automatically be restarted or replaced by the system to ensure continuous operation. And run as if nothing had happened.

→ Isn’t having Auto-Scaling is a feature that auto-adjusts the num. of compute resources (like instances, containers, or services) in response to changing workloads. The goal is to ensure that you have the right amount of resources at any time: scaling up during high traffic or usage and scaling down when demand is lower, optimizing performance and cost-efficiency. For Instance : Say we have a Huge Festive Sale coming upp!! Then over-night our container goes down, this happened due to huge-surge-of-requests which made our server overwhelmed resulting to a crash. This happened because of not increasing the capacity (Vertical Scaling) or increasing the nodes (Horizontal) whenever it’s needed or even underutilized during off-season.

→ Is simple and not Enterprise Level Support Some of features of ‘Enterprise Level Supported Platforms’: ⇒ A Load Balancer ⇒ A Firewall (At least A Proxy Firewall) ⇒ Auto-Healing ⇒ Auto-Scaling

Ok, we know issues/problems which are unsolved by Docker. How does Kubernetes resolve these?

Problem of single-host based :

⇒ Kubernetes by default uses Clusters , which are group of contianers/instances having them

So the single-point of failure or the single node/container affecting others is now eleminated.

Kubernetes uses Master-Slave Architecture and so the containers are now isolated completely so that one faulty node doesn’t affect the other if so.

• Master-Slave Architecture :

Problem of Auto-Healing :

⇒ Kubernetes has Replica sets which are used for fallback mechanism, It auto-heals means that it will automatically shift the traffic of the failed node, towards the another void node.

K8’s also has HPA - Horizontal Pod Scaleup which can increase the nodes horizontally, whenever there is need for ‘heal’.

Whenever the Kubernetes-Master notices that a node is about to go down, then immediately it will roll-out a new node even b4 a old one goes out, even the user can’t figure that this healing has happened. This notifying about a node going down is done by - Kube APi-Server .

Problem of not being an Enterprise-level-supported platform : Folks at Google, have developed K8’s. Have made it as the Enterprise-level-supported platform.