aws-cloud-map-mcs-controller-for-k8s
aws-cloud-map-mcs-controller-for-k8s copied to clipboard
aws
K8s controller implementing Multi-Cluster Services API based on AWS Cloud Map.
Instead of hitting server side throttles and subsequent failures, implement client-side traffic shaping to slow down request submission. Example from [app-mesh-controller](https://github.com/aws/aws-app-mesh-controller-for-k8s/blob/db3f146f984b0cab7522d5ceb0322f492a4b5720/pkg/aws/throttle/defaults.go#L11).
Proper error handling during the startup for the non-recoverable states. For example: If we fail to sign request because of missing credentials, terminate the controller instead of retrying ``` failed...
As per the [Multi-Cluster Services API](https://github.com/kubernetes/enhancements/tree/master/keps/sig-multicluster/1645-multi-cluster-services-api#endpoint-ttl), we need to define the strategy for below > Endpoint TTL > > To prevent stale endpoints from persisting in the event that the...
In our Organization we have a huge amount of Teams with lots of EKS Clusters distributed throughout the Organization, making the usage of VPC Peering compared to Transit Gateway a...
We export controller version as an instance attribute. We should check this variable to determine if and how we will import the endpoint.
Current State: Instances correctly deregistered, but dont see the default namespace and services cleaned up outside of that Desired state: Want the ability to allow the controller to properly clean...
Hi, I encountered an error after upgrading from 0.2.3 to 0.3.0. I don't know if the error appears on a new installation: ``` {"level":"error","ts":1664464680.674545,"logger":"controllers.Cloudmap","msg":"error when syncing service","namespace":"bidder-prod","name":"redis-master","error":"endpointslices.discovery.k8s.io is forbidden: User...
We use cloudmap controller in multiple clusters and have the same problem on them. Clusters are linked by vpc peering. Half of the time, an imported service cannot be reached...
Set aws-region via kustomize edit command instead of the environment variable substitution. Environmental substitution was a bug documented as a feature, and is going to be fixed in the future...
1. Configure 2 kind clusters, and ensure the controller features works across them 2. Inject AWS creds into the controlller docker as env variables 3. Use separate namespace for each...