1 00:00:00,120 --> 00:00:07,410 .In this lecture we will discuss about the part to note relationship and how you can restrict what parts 2 00:00:07,500 --> 00:00:15,820 are placed on what notes the concept of teens and toleration can be a bit confusing for beginners. 3 00:00:15,870 --> 00:00:22,710 So we will try to understand what they are using an analogy of a bug approaching a person. 4 00:00:22,710 --> 00:00:29,480 Now my apologies in advance but this is the best I could come up with to prevent the bug from landing 5 00:00:29,570 --> 00:00:30,800 on the person. 6 00:00:31,010 --> 00:00:36,820 We spray the person with a repellent spray or attained as we will call it. 7 00:00:36,860 --> 00:00:41,750 In this lecture the book is intolerant to the smell. 8 00:00:41,750 --> 00:00:50,350 So on approaching the person detained applied on the person throws the bug off however there could be 9 00:00:50,470 --> 00:00:56,920 other books that are tolerant to the smell and so the taint doesn't really affect them. 10 00:00:57,150 --> 00:01:00,370 And so they end up landing on the person. 11 00:01:00,370 --> 00:01:05,330 So there are two things that decide if a bug can land on a person. 12 00:01:05,530 --> 00:01:08,080 First the taint on the person. 13 00:01:08,290 --> 00:01:15,310 And second the bugs toleration level to that particular taint getting back to Cuba. 14 00:01:15,310 --> 00:01:20,420 NET is the person is a node and the bugs are pods. 15 00:01:20,430 --> 00:01:27,230 Now teens and toleration have nothing to do with security or intrusion on the Cluster. 16 00:01:27,560 --> 00:01:33,800 Teens and toleration are used to set restrictions on what parts can be scheduled. 17 00:01:33,860 --> 00:01:41,510 On a note let us start with a simple cluster with three worker notes the notes are named one two and 18 00:01:41,510 --> 00:01:42,390 three. 19 00:01:42,440 --> 00:01:46,420 We also have a set of paths that are to be deployed on these notes. 20 00:01:46,490 --> 00:01:48,710 Let's call them A B C and D. 21 00:01:49,010 --> 00:01:55,410 When the parts are created common it is scheduler tries to place these parts on the available worker 22 00:01:55,430 --> 00:01:56,560 nodes. 23 00:01:56,690 --> 00:02:04,130 As of now there are no restrictions are limitations and as such the scheduler places the parts across 24 00:02:04,220 --> 00:02:06,910 all of the nodes to balance them out equally. 25 00:02:08,120 --> 00:02:15,510 Now let us assume that we have dedicated resources on Node 1 for a particular use case or application 26 00:02:16,520 --> 00:02:23,230 so we would like only those parts that belong to this application to be placed on Node 1. 27 00:02:23,310 --> 00:02:30,300 First we prevent all parts from being placed on the Node by placing a taint on the Node. 28 00:02:30,900 --> 00:02:33,910 Let's call it blue by default. 29 00:02:33,980 --> 00:02:43,510 Paths have no toleration which means unless specified otherwise none of the parts can tolerate any taint. 30 00:02:43,510 --> 00:02:50,260 So in this case none of the parts can be placed on Node 1 as none of them can tolerate the team. 31 00:02:50,270 --> 00:02:59,170 Blue this solves half of our requirement no unwanted parts are going to be placed on this note. 32 00:02:59,360 --> 00:03:04,100 The other half is to enable certain parts to be placed on this note. 33 00:03:05,930 --> 00:03:11,860 For this we must specify which parts are tolerant to this particular attempt. 34 00:03:12,050 --> 00:03:20,840 In our case we would like to allow only Part D to be placed on this note so we add a toleration to Part 35 00:03:20,960 --> 00:03:21,290 D. 36 00:03:22,280 --> 00:03:30,260 Part D is now tolerant to blue so when the scheduler tries to place this part on Node 1 it goes through 37 00:03:30,650 --> 00:03:35,960 Node 1 can now only accept parts that can tolerate the taint. 38 00:03:35,990 --> 00:03:40,740 Blue so with all the teens and toleration is in place. 39 00:03:40,850 --> 00:03:44,100 This is how the parts would be scheduled. 40 00:03:44,180 --> 00:03:50,450 The scheduler tries to place part a on not one but due to the team it is thrown off and it goes to note 41 00:03:50,450 --> 00:03:57,440 to the scheduler then tries to place part B on not one but again due to the taint it is thrown off and 42 00:03:57,440 --> 00:04:01,490 is placed on Node 3 which happens to be the next free node. 43 00:04:01,490 --> 00:04:07,420 The scheduler then tries to place part c to the Node 1 it is thrown off again and ends up on Node 2. 44 00:04:07,460 --> 00:04:14,520 Finally the scheduler tries to place part D on Node 1 since the part is tolerant to Node 1. 45 00:04:14,540 --> 00:04:22,540 It goes through so remember trains are set on nodes and toleration are set on pods. 46 00:04:22,690 --> 00:04:24,330 So how do you do this. 47 00:04:24,430 --> 00:04:31,750 Use the cube control taint nodes command to taint in node specify the name of the node to train followed 48 00:04:31,750 --> 00:04:35,260 by the train itself which is a key value pair. 49 00:04:35,320 --> 00:04:42,550 For example if you would like to dedicate the Node 2 parts in application blue then the key value pair 50 00:04:42,550 --> 00:04:49,820 would be ab equals blue the taint effect defines what would happen to the pods. 51 00:04:49,840 --> 00:04:58,120 If they do not tolerate the taint there are three taint effects no schedule which means the parts will 52 00:04:58,120 --> 00:05:02,030 not be scheduled on the node which is what we have been discussing. 53 00:05:02,220 --> 00:05:08,950 Prefer No schedule which means the system will try to avoid placing a part on the node but that is not 54 00:05:08,950 --> 00:05:10,520 guaranteed. 55 00:05:10,570 --> 00:05:12,790 And third is no execute. 56 00:05:13,000 --> 00:05:21,010 Which means that new parts will not be scheduled on the node and existing parts on the node if any will 57 00:05:21,010 --> 00:05:24,700 be evicted if they do not tolerate the taint. 58 00:05:24,700 --> 00:05:31,000 These parts may have been scheduled on Node before the 10th was applied to the node. 59 00:05:31,030 --> 00:05:38,200 An example command would be detained Node node one with the key value pair app equals blue and an effect 60 00:05:38,470 --> 00:05:46,440 of no schedule toleration is are added to parts to add a toleration to a pod. 61 00:05:46,450 --> 00:05:52,980 First pull up the pod definition file in the spec section of the pod definition file had a section called 62 00:05:53,010 --> 00:05:59,530 toleration move the same values used to while creating the taint under this section. 63 00:05:59,580 --> 00:06:07,830 The key is app operator is equal value is blue and the effect is no should you. 64 00:06:07,890 --> 00:06:13,810 And remember all of these values need to be encoded in double codes. 65 00:06:14,100 --> 00:06:21,690 When the pods are now created or updated with the new toleration they are either not scheduled on nodes 66 00:06:21,900 --> 00:06:29,780 or evicted from the existing nodes depending on the effect set let us try to understand the No. 67 00:06:29,790 --> 00:06:33,600 Execute takes effect in a bit more depth. 68 00:06:33,960 --> 00:06:37,380 In this example we have three nodes running some workload. 69 00:06:37,380 --> 00:06:43,440 We do not have any teens or toleration at this point so they are scheduled this way. 70 00:06:43,440 --> 00:06:51,420 We then decided to dedicate Node 1 for a special application and as such we obtained the node with the 71 00:06:51,420 --> 00:06:58,380 application name and add a toleration to the part that belongs to the application which happens to be 72 00:06:58,380 --> 00:06:59,280 part D. 73 00:06:59,280 --> 00:07:06,560 In this case while tending the node we set the taint effect to no execute and as such. 74 00:07:06,750 --> 00:07:14,190 Once detained on the Node takes effect it evokes Part C from the node which simply means that the part 75 00:07:14,220 --> 00:07:15,170 is killed. 76 00:07:15,360 --> 00:07:20,930 The Part D continues to run on the node as it has a toleration to the change. 77 00:07:22,030 --> 00:07:26,980 Now going back to our original scenario where we have teens and toleration is configured. 78 00:07:27,070 --> 00:07:35,500 Remember teens and toleration are only meant to restrict nodes from accepting certain parts. 79 00:07:35,500 --> 00:07:45,160 In this case no one can only accept Part D but it does not guarantee that Part D will always be placed 80 00:07:45,220 --> 00:07:46,400 on Node 1. 81 00:07:46,480 --> 00:07:53,800 Since there are no teens or restrictions applied on the other two nodes Part D may very well be placed 82 00:07:53,830 --> 00:08:02,350 on any of the other two nodes so remember teens and toleration does not tell the part to go to a particular 83 00:08:02,440 --> 00:08:03,140 node. 84 00:08:03,160 --> 00:08:09,460 Instead it tells the node to only accept parts with certain toleration. 85 00:08:10,150 --> 00:08:17,950 If your requirement is to restrict a part to certain nodes it is achieved through another concept called 86 00:08:17,980 --> 00:08:23,410 Ask node affinity which we will discuss in the next lecture finally. 87 00:08:23,510 --> 00:08:28,200 While we're on this topic let us also take a look at an interesting fact. 88 00:08:28,460 --> 00:08:32,650 So far we have only been referring to the worker nodes. 89 00:08:32,670 --> 00:08:39,290 But we also have master nodes in the cluster which is technically just another note that has all the 90 00:08:39,290 --> 00:08:41,780 capabilities of hosting a part. 91 00:08:41,790 --> 00:08:45,350 Plus it runs all the management software. 92 00:08:45,470 --> 00:08:52,430 Now I'm not sure if you noticed the scheduler does not schedule any part on the master node. 93 00:08:52,430 --> 00:08:56,590 Why is that when the covenant is cluster is first setup. 94 00:08:56,630 --> 00:09:03,530 It teams is set on the master node automatically that prevents any parts from being schedule on this 95 00:09:03,530 --> 00:09:04,250 note. 96 00:09:04,580 --> 00:09:08,670 You can see this as well as modified his behavior if required. 97 00:09:08,990 --> 00:09:17,330 However a best practice is to not deploy application workloads on a master server to see this taint 98 00:09:17,540 --> 00:09:18,340 run a cube. 99 00:09:18,340 --> 00:09:26,420 Control described node command with Cube master as the node name and look for the tail section. 100 00:09:26,540 --> 00:09:32,200 You will see a taint set to not schedule any parts on the master node. 101 00:09:32,270 --> 00:09:34,310 Well that's it for this lecture. 102 00:09:34,310 --> 00:09:39,320 Head over to the coding exercises section and practice working with teens and toleration is.