Managed Cluster Usage Instructions
Note
The availability of a CCE cluster depends heavily on the number of managed pods, ConfigMaps, CRDs, and other resources, as well as the frequency of Get/List read operations and QPS, Patch/Delete/Create/Update write operations. Avoid initiating List-like operations on clusters with a large number of resources, using clusters as databases, or writing excessive ConfigMaps/CRDs/EndPoints, as these actions may affect cluster availability.
Common list-like operations are as follows (using pods as an example):
- Query with tags
kubectl get pod -l app=nginx
- Namespace-specific query
kubectl get pod -n default
- Query all pods in the cluster
kubectl get pod --all-namespaces
- List requests initiated via client-go
k8sClient.CoreV1().Pods("").List(metav1.ListOptions{})
To query all cluster resources, it's suggested to use the K8s informer mechanism with a local cache for optimal results. For simpler cases, you can query the kube-apiserver cache by including the ResourceVersion parameter in the List request, like this: k8sClient.CoreV1().Pods("").List(metav1.ListOptions{ResourceVersion: "0\
Recommended cluster scale configuration
When setting up a managed cluster, follow the recommended configuration below and choose an appropriate cluster size based on your actual service requirements to prevent unavailability caused by excessive load on control plane components.
Description:
Currently, it only supports creating L50-scale managed clusters by default. If you wish to create a larger-scale cluster or upgrade the existing cluster to the desired scale, you may submit a ticket or contact us.
For instance, if you intend to deploy 50 nodes in a cluster but require 2,000 pods, select a cluster size with a recommended maximum managed node count of 200 (rather than 50).
| Cluster scale | Maximum managed node count | Recommended maximum pod count | Recommended maximum service count | Recommended maximum ConfigMap count | Recommended maximum Secret count | Recommended maximum PV count | Recommended maximum PVC count | Recommended maximum CRD count |
|---|---|---|---|---|---|---|---|---|
| L50 | 50 | 1500 | 100 | 300 | 300 | 250 | 250 | 1000 |
| L200 | 200 | 6000 | 400 | 1200 | 1200 | 1000 | 1000 | 4000 |
| L500 | 500 | 15000 | 1000 | 3000 | 3000 | 2500 | 2500 | 10000 |
| L1000 | 1000 | 30000 | 2000 | 6000 | 6000 | 5000 | 5000 | 20000 |
| L3000 | 3000 | 90000 | 6000 | 18000 | 18000 | 15000 | 15000 | 60000 |
| L5000 | 5000 | 150000 | 10000 | 30000 | 30000 | 25000 | 25000 | 100000 |
Note:
- The managed pod count does not include pods created by system components in the cluster (e.g., CoreDNS).
- The managed CRD count excludes CRDs generated by system components in the cluster (e.g., network-related system components).
- It is advised that the total volume of all objects within each resource type should not surpass 800 MiB, and the size of each individual resource object should remain under 100 KB.