I have setup k8s cluster with nodes having ips from 2 differenet subnets that can't talk to each other. What additional settings I need to do for the cluster to work smoothly? Moreover the ip assigned to ingress controller service is also in a different subnet configured using metallb. Please help..
You will have to allow traffic between control plane subnet and worker node subnets. Also ingress controller subnet should be allowed to talk to worker node subnet
Hello can you help me on this usecase. I am running my application with 3 replica sets. My https requests are not evenly distributed across the pods. What configuration we do need to check?
You need to use a load balancer that can route requests to the service in a weighted manner. Create a service, then in ingress endpoint and then add the LB to send traffic using weighted parameter. Another way is explore your ingress if it has the feature. But you def need to create a k8s service before that
@@shek8s I have load balancer and service and each pod is mapped with different node. what are the configurations I need to cross check Wheather all pods are equally distributed the traffic. And under resources only memory value is set for requests and limits. cpu is not yet set. Will it be impacted for this traffic distribution.
Mam one question if pod networking is doing everything within cluster then why do we need cluster IP service? As far as I know cluster IP service enable communication between two pods at different node then why do we need this IP we can directly do it through flannel right? Waiting for you response. Thanks in advance :)
Cluster IP provides the service endpoints for reaching the service within the cluster. Flannel, calico are some of the common networks that can be setup (mostly provided by the k8s provider so that admins don;t have to set it up). So yes, pod networking via calico/flannel is all that's needed.
I have setup k8s cluster with nodes having ips from 2 differenet subnets that can't talk to each other. What additional settings I need to do for the cluster to work smoothly? Moreover the ip assigned to ingress controller service is also in a different subnet configured using metallb. Please help..
You will have to allow traffic between control plane subnet and worker node subnets. Also ingress controller subnet should be allowed to talk to worker node subnet
As far as I know, AWS CNI is same CIDR between Node IP Range and Pod IP Range.
Yes you are rite if we use AWS CNI addon but in my example i added custom network plugin.
Hi, How can I ping from one pod to another pod using its hostname?
You can use the pod spec.hostname and spec.subdomain for that. I will post a video on that soon.
you are teaching too nice.. Are you from India ? By any chance.. !! 🙂
Nope
Hello can you help me on this usecase. I am running my application with 3 replica sets. My https requests are not evenly distributed across the pods. What configuration we do need to check?
You need to use a load balancer that can route requests to the service in a weighted manner. Create a service, then in ingress endpoint and then add the LB to send traffic using weighted parameter. Another way is explore your ingress if it has the feature. But you def need to create a k8s service before that
@@shek8s I have load balancer and service and each pod is mapped with different node. what are the configurations I need to cross check Wheather all pods are equally distributed the traffic. And under resources only memory value is set for requests and limits. cpu is not yet set. Will it be impacted for this traffic distribution.
@@pradeepbavirisetty9582 it is a good idea to always set cpu and mem limits. also which LB are you using? Is it on AWS?
@@shek8s I'm using oracle cloud, and I'm using spec.type is clusterIP will it be impacted for http request distribution?
Mam one question if pod networking is doing everything within cluster then why do we need cluster IP service? As far as I know cluster IP service enable communication between two pods at different node then why do we need this IP we can directly do it through flannel right? Waiting for you response. Thanks in advance :)
Cluster IP provides the service endpoints for reaching the service within the cluster. Flannel, calico are some of the common networks that can be setup (mostly provided by the k8s provider so that admins don;t have to set it up). So yes, pod networking via calico/flannel is all that's needed.
@@shek8s thanks a lot for ur prompt; got my answer :)
I am get permission denied while executing traceroute
try running as root