aws-efs-csi-driver VS aws-load-balancer-controller

In the past I was have problems with GID allocator, something related to this problem.

However, the docs (https://github.com/kubernetes-sigs/aws-efs-csi-driver/blob/master/examples/kubernetes/dynamic_provisioning/README.md) are telling me to create EFS Mount Targets in the EKS subnets. Thats fine.

Fargate storage A Pod running on Fargate automatically mounts an Amazon EFS file system. You can't use dynamic persistent volume provisioning with Fargate nodes, but you can use static provisioning. For more information, see Amazon EFS CSI Driver on GitHub.

I guess something like this might go a long way to solve the problem https://github.com/kubernetes-sigs/aws-efs-csi-driver/pull/640 ? Though I see it isn't merged yet

I am currently trying to create an EFS for use within an EKS cluster. I've followed all the instructions, and everything seems to be working for the most part. However, when trying to apply the multiple_pods example deployment from here, the pods cannot succesfully mount the file system. The PV and PVC are both bound and look good, however the pods do not start and yield the following error message:

In each namespace, create a PV/PVC using the same fixed volume path. See "Volume Path in EKS CSI Driver" To make this work however, you MUST pre-create this volume path in your EFS (I usually just have an EC2 instance with it mounted to work on). From the docs above "Note: this feature requires the sub directory to mount precreated on EFS before consuming the volume from"

resource "aws_iam_policy" "eks_efs_csi_driver_policy" { # https://github.com/kubernetes-sigs/aws-efs-csi-driver/blob/master/docs/iam-policy-example.json policy = file("./6.AWSEFSpolicy.json") name = "aws-efs-csi-policy" }

efs - IAM Policy for AWS EFS CSI Driver.

I thought at the beginning that such certificate would then expire, but I have seen cert-manager is within ALB code https://github.com/kubernetes-sigs/aws-load-balancer-controller/blob/main/config/certmanager/certificate.yaml so that makes me hesitate about it.

Detailed behavior changes can be found in the release notes for version 2.6.0 of the AWS Load Balancer Controller. https://github.com/kubernetes-sigs/aws-load-balancer-controller/releases The key points include:

If you are using Kubernetes, you can enable security groups on your NLB by using AWS Load Balancer controller version 2.6.0 or later. Enabling NLB security groups using the controller enhances the nodes' security, as inbound rules can be simplified by referencing the NLB security groups. It also provides scaling improvements, as the controller keeps a constant number of security group rules per cluster.

I use the AWS Load Balancer Controller: https://github.com/kubernetes-sigs/aws-load-balancer-controller.

You can install the AWS Load Balancer Controller, and then create an "Ingress" when you install your Nginx Ingress controller, possible with some necessary annotations.

aws-load-balancer-controller

Yes. Google's ingress controller talks to the GCP API and spins up an HTTP LB for you. The AWS LB Controller handles provisioning ALB's for you. And ... I guess other clouds exist too but these are the ones I am familiar with.

NLB don't support weighted target groups and for some reason are extremely slow when adding instances o removing them from target groups, which makes deployments a bit meh (old instances and new instances can be receiving traffic at the same time for up to 2 mins)