1 00:00:00,000 --> 00:00:10,933 Hi let's have a better understanding on various core components of kubernetes cluster. 2 00:00:10,933 --> 00:00:16,332 Already we had a quick discussion on various components involved, now I am going to explain 3 00:00:16,332 --> 00:00:21,733 with an analogy as well as compare it with the cluster that we have installed. So that we will 4 00:00:21,733 --> 00:00:25,999 have better understanding on how they are related to each other and what functionality 5 00:00:26,000 --> 00:00:30,466 they are going to do and we are going to have a detailed discussion on each and every 6 00:00:30,466 --> 00:00:35,632 component in future. When it comes to the core components there are two different categories 7 00:00:35,633 --> 00:00:40,399 available that is controlled plane and runtime components. As a part of controlled plane 8 00:00:40,400 --> 00:00:44,300 components that's going to make sure the kubernetes cluster is up and running 9 00:00:44,300 --> 00:00:50,400 that's where I am going to have kube api server, kube controller manager, kube scheduler and 10 00:00:50,400 --> 00:00:54,900 etcd database. As a part of run time components, I am going to have 11 00:00:54,900 --> 00:01:01,666 kubelet, kube proxy and container runtime like docker rocketd or containerd. These 12 00:01:01,666 --> 00:01:06,866 three components going to make sure the applications within the kubernetes is running 13 00:01:06,866 --> 00:01:13,799 fine. Now let us have a quick analogy and relate these components. I'm going to explain with 14 00:01:13,800 --> 00:01:19,133 an analogy and I'm going to map each and every component of this analogy to the real 15 00:01:19,133 --> 00:01:25,933 component of kubernetes. I'm going to match the functionalities, so don't try to match as it is 16 00:01:25,933 --> 00:01:33,599 for understanding purpose this analogy is made. Assume I do have a head office and n number 17 00:01:33,600 --> 00:01:40,400 of branches operated in different locations. For understanding purpose I am going to have a 18 00:01:40,400 --> 00:01:49,266 location say Chennai and another location in say New York and within head office I will be having 19 00:01:49,266 --> 00:01:57,132 different departments, say one department going to take care of all the communications. 20 00:01:57,133 --> 00:02:02,933 That is within the departments within head office as well as communications to the other 21 00:02:02,933 --> 00:02:07,933 branches the communication department going to take care. The same way I will be having 22 00:02:07,933 --> 00:02:13,099 another department that's going to make all the planning for the projects, how to execute the 23 00:02:13,100 --> 00:02:16,733 current project and what are all the future projects, where to execute the future projects 24 00:02:16,733 --> 00:02:21,299 basically all the planning will be done by the planning department and I will be having 25 00:02:21,300 --> 00:02:27,733 HR department, they'll be taking care of the people and each branch office they will be 26 00:02:27,733 --> 00:02:35,799 having a head, who is going to take care of that particular department or the branch. Let's call 27 00:02:35,800 --> 00:02:43,000 him as branch manager. So each branch going to have a manager and the entire branch will be 28 00:02:43,000 --> 00:02:47,800 under the control of the branch manager in terms of who are all working in that specific 29 00:02:47,800 --> 00:02:53,700 branch whether everyone is getting the required resources or not and everyone is secure and 30 00:02:53,700 --> 00:02:59,633 safe everyone adhere into the same policy, put in within that specific Branch or not. So the 31 00:02:59,633 --> 00:03:04,066 entire branch will be under the control of the branch manager and the branch manager 32 00:03:04,066 --> 00:03:10,399 will be having a secretary to coordinate the communication. So the entire communication 33 00:03:10,400 --> 00:03:16,266 coming in or going out of the branch will be going through the secretary and all the details 34 00:03:16,266 --> 00:03:21,399 about the head office the people who are working what project being managed the data 35 00:03:21,400 --> 00:03:27,100 needs to be stored somewhere. So I'm going to have a data store and that needs to be very very 36 00:03:27,100 --> 00:03:35,833 highly available so I may have a cluster of databases which is going to store all the data about the people what communication 37 00:03:35,833 --> 00:03:42,333 is happening and what are all the reporting hierarchy every information will be stored within 38 00:03:42,333 --> 00:03:49,999 this specific database and the branch manager will be reporting to the head office through the 39 00:03:50,000 --> 00:03:56,666 communication channel and for external communication I will be having a reception. So 40 00:03:56,666 --> 00:04:02,632 whoever wants to communicate they will be coming through this specific or the protocol 41 00:04:02,633 --> 00:04:09,733 with which they will be communicating and that's the reception or the front desk. Now 42 00:04:09,733 --> 00:04:16,333 I am going to replace each and every component related to the kubernetes. Branch 43 00:04:16,333 --> 00:04:23,733 manager that's the kubelet and that's going to run in each and every worker machine and the 44 00:04:23,733 --> 00:04:28,399 communication entire communication will be handled by the kube api server and 45 00:04:28,400 --> 00:04:33,900 the front desk that is like kubectl tool and that will be used to make the communication and 46 00:04:33,900 --> 00:04:40,000 that's going to leverage the kube api server to do all communication and project planning 47 00:04:40,000 --> 00:04:48,266 that's called a scheduler kube scheduler and the HR that's called controller and the database 48 00:04:48,266 --> 00:04:56,599 that is etcd database and that's going to work us a cluster or a quorum. Manager secretary is 49 00:04:56,600 --> 00:05:02,833 the kube-proxy. Now let's assume I do have a cluster where there is no job happening and 50 00:05:02,833 --> 00:05:08,333 how the reporting hierarchy and what is going to report to what. The kubelet that's going to be a 51 00:05:08,333 --> 00:05:14,133 continuously running daemon and that will be made to get installed in each and every machine, 52 00:05:14,133 --> 00:05:19,933 along with kubelet I will be having kube-proxy the reason because I may start n number of 53 00:05:19,933 --> 00:05:25,133 pods and all the communication within the pods within the nodes as well as across the 54 00:05:25,133 --> 00:05:30,799 nodes and going across the cluster needs to be taken care and that will be coordinated by the 55 00:05:30,800 --> 00:05:35,466 kube proxy. The kube proxy will support n number of network provider to support the 56 00:05:35,466 --> 00:05:42,399 network communications and kube scheduler that's going to find out which node can take 57 00:05:42,400 --> 00:05:49,233 what type of work. Say for example I may have a different infrastructure, say for example this 58 00:05:49,233 --> 00:05:56,266 particular machine will have SSD and graph processor, the similar way I may have another 59 00:05:56,266 --> 00:06:03,766 node which is going to have SSD and graph processor. So what that means is any process 60 00:06:03,766 --> 00:06:08,166 related to machine learning if it is executed using the graph processor the efficiency is 61 00:06:08,166 --> 00:06:15,332 going to be much higher. So I need to have a planner which is going to place the container 62 00:06:15,333 --> 00:06:20,399 so that the container is placed within this specific node, where the graph processor is 63 00:06:20,400 --> 00:06:26,033 available. So in the similar way I may have different type of requirements like all 64 00:06:26,033 --> 00:06:33,266 the front end containers needs to have the cache running along with the node. Say 65 00:06:33,266 --> 00:06:38,899 I do have a front end container and it's going to use some cache for better performance. 66 00:06:38,900 --> 00:06:43,200 So the cache should be co-located along with a front end so these type of scheduling 67 00:06:43,200 --> 00:06:49,033 requirement will be required for the containers and the kube scheduler going to take care of 68 00:06:49,033 --> 00:06:54,533 what container should be placed in which node. I may have some constraints that particular 69 00:06:54,533 --> 00:06:57,933 node will not have enough space to accommodate the container 70 00:06:57,933 --> 00:07:04,066 So the kube scheduler going to find out which node going to have the space and it's going to 71 00:07:04,066 --> 00:07:10,332 prioritize which node can do that particular job in a better way, accordingly it will be scheduling 72 00:07:10,333 --> 00:07:17,433 the requirement to put in the container and as per the schedule the kubelet will take care of 73 00:07:17,433 --> 00:07:22,933 creating the container and make it up and running. Once the container is up and running 74 00:07:22,933 --> 00:07:29,066 that will be controlled by the controller, in terms of which are all the group of containers going to 75 00:07:29,066 --> 00:07:35,132 work together and serve the front end. Front end I'm just considering it as an application 76 00:07:35,133 --> 00:07:41,299 a web application just for example. Now within the application I may have a requirement 77 00:07:41,300 --> 00:07:47,066 I need to have two replica of front end and that should be getting executed where I do have the 78 00:07:47,066 --> 00:07:53,466 SSD processor. So accordingly the scheduler will decide which node can do that job in a 79 00:07:53,466 --> 00:07:58,866 better way once that is decided the container will get started. Once it is started that will be 80 00:07:58,866 --> 00:08:04,132 controlled by the controller, in case if that particular node goes offline for some reason 81 00:08:04,133 --> 00:08:13,233 and it couldn't serve the request, it will decide one particular instance or one particular replica 82 00:08:13,233 --> 00:08:18,766 is not available and that needs to be scheduled in a different node. With the help of the 83 00:08:18,766 --> 00:08:23,799 scheduler it will identify that is the controller will identify which node can satisfy 84 00:08:23,800 --> 00:08:29,000 the requirement and accordingly the container will be placed in those nodes. So within 85 00:08:29,000 --> 00:08:34,033 controller I am going to have multiple controllers like node controller in terms of 86 00:08:34,033 --> 00:08:39,933 controlling the nodes replication controller in terms of controlling the replications and every 87 00:08:39,933 --> 00:08:48,699 machine will be having the run time that is docker or rocket or containerd. Accordingly the 88 00:08:48,700 --> 00:08:55,066 container would get created. So I do have a cluster up and running where I do have a master 89 00:08:55,066 --> 00:09:00,799 and node 2 is the master and I do have n number of workers that is node 3 and node 4 90 00:09:00,800 --> 00:09:07,966 I can get all the components as a part of the all namespaces, here I do have kubeapi server 91 00:09:07,966 --> 00:09:13,999 kube controller, kube scheduler and the kube proxy that is running in each and every machine 92 00:09:14,000 --> 00:09:19,633 to coordinate the communication and I am using calico network provider 93 00:09:19,633 --> 00:09:25,599 so the calico pod is also installed. Now you might be wondering where the kubeadm tool 94 00:09:25,600 --> 00:09:32,066 kubectl tool and kubelet. All these three components needs to be installed as a daemon 95 00:09:32,066 --> 00:09:36,932 that's why while doing the installation we will be installing all these components as a daemon. 96 00:09:36,933 --> 00:09:42,599 In this particular case we used kubeadm tool, so that's going to install the network provider 97 00:09:42,600 --> 00:09:49,366 controller api server all as a pod. As a next session as a introduction I am going to revise 98 00:09:49,366 --> 00:09:53,866 the different type of installations and what installation we did. So in a quick summary 99 00:09:53,866 --> 00:09:57,699 we related various components with an analogy for better understanding 100 00:09:57,700 --> 00:10:02,400 and to remember their functionalities.