The Ultimate Kubernetes Challenge: An Interactive Trivia Game - Aurélie Vache, OVHcloud

This talk covers the management of Kubernetes clusters, focusing on version compatibility and deployment strategies. The speaker discusses how to handle secrets within a cluster, including the types of secrets like passwords and certificates, and the methods to retrieve their values. He explores the capabilities of the Horizontal Pod Autoscaler (HPA) in scaling applications based on CPU usage metrics and demonstrates the deployment and resource configuration processes. Additionally, traffic distribution among pods based on zone and node preferences is highlighted, along with troubleshooting tips for common issues encountered in K8s environments. The session emphasizes the importance of effective resource allocation and cluster management.

ultimate challenge with mix of s and developer advocate cloud and special infrastructure opportunity please scan QR code swag okusters per ok tips aleras So, welcome. Oh yes! Dead, dead, dead. ! Yes! Oh ! Oh no. Ok. I little bit but I will start because time. So question number one I have a cluster running version 144 which version of should I install use4 only4 or less and or

it's up to you easy easy what do you think not easy. Ok. Ok. So yes you true true congratulation null and chisl support version of IP server older and newer so if the IPO server is equals to4 you Inst345 responsible SK or scheduler. yeah. IP server scheduler that will do best to find suitable node and then is scheduled on new easy and now it will be less

easier yes I hope so question Yes. Secret name secret data value password or certificate what do that again is it data or value ex generic type of secret content in demo in the first demo so I use cluster with ss switch it So we will create the three types of secrets. literal login password and URL of my registry and and secret with my certificate with a key

and certificate. are creating ok it's ok and if I dig into one secret as you can see the password in the data fielded and the same thing for the registry secret configuration is in dat field and the same thing for the secret andouf but is it easy to get the content yes but no you have to get the secret with the name of secret with the name

of the space and you have to view secret plugin and it is easy to get the value of secret thanks view secret it is automatically get fine can we create space zer work true or false can we createace with zer tic tac voilà yes we can moving a lot so yes in de souster with zer node. yes zero node in cluster and create a namace created and

yes since seconds if try to delete my name do you OK. I wait I can wait more and the space will be never deleted. Why? controller matrix server is in pending status so we don't have matrix server soace controller can't delete Ah là là go horizontal pod autaler configure CPU CPU request HP request can work normally never scale or can scale in I like it. zpa will

work normally thanks to horizontal autement controller matrix CPU complicated without calculate CPU con define deployment with only CPU limit the HPA will work thanks to the API server super power automatically change on the of deep payment and it now so let's switch in my other commune and now let's deploy an application with only the resource limit so I will create deployment name my app CPU limit OK

service accent in the cluster and I have one that is running that ok now let's check resources of thanks to the coup get command and G gq GQ command as you can see even if define only CPU also CPU CPU request. HP. So it's my app and it will adjust the number of pods to have an average CPU consumption equ 40%. Let's apply it to my cluster.

It's ok. access in order to simulate some CPU conjunction so we will wait a little time ok yeah let's wait some second and I think we can stop it and it and it should Ok display our HPA for the moment calculate the usage of the CPU but oh I didn't OK. Yes. If I more the HPA works. OK. Okay, let's go. Question number six. Are you ready?

Ah, okay. Thank you. service can distribute traffic to any ready pod that match across the cluster pods running in same zone as the client or two running the same as client only created by same deployment you have to check good In fact, it's the good. It's the sweet first. Yeah. to set of IP name but also the node and the zone iPod and thanks to that information

you can do rotating to a specific node or to a specific zone. Not that traffic distribution field allows you to express preferences. If you are a user you have to do things. Fure activated zone and preferred node are not implemented I created issue days request been created. So in several days you will be able to use preferone and prefer same zone same node if you are silium

user. OK. in my cluster per zone. OK. PR close prefer close is allias for preferably my deployment on my cluster and let's apply my service on my cluster to check my service as usual it's ok a service and now let's read my slice object so as you can see it's the list of all the Ok so we will create a pod and then we will execute curl

command in order to access to a pod and we will see if the traffic will be root in the good zone. let's verify where ours with coup gets command and as sender is in so traffic been in same zone. Final and last question. Quick, be fast and accurate. Be quick. configuring state in crushed loop back off state state and is deleted Congratulations Victor and Squad. Take screen

capture phone and go minutes. scheduler try to find suitable node and if not possible zip zipod will be scheduled in node whatever but with cried during rule let's first check the labels of our notes. Ok during rule that match only inck statusinity status. I hope that this particular cook the final word I have for you is thank positive feedback No.