Updated on 2026-08-13 GMT+08:00

Resource Data Volume of Different Cluster Sizes

As the service scale expands, the maximum number of nodes in a Kubernetes cluster directly affects service scalability and flexibility. When more services or pods are deployed in an existing cluster, cluster resources may become insufficient. As a result, service deployment may fail or performance may deteriorate. The resource data volume of different cluster sizes may be limited. You can select cluster sizes that match the service scale based on Table 1 to meet the requirements of large-scale service deployment and improve resource utilization and service capacity of clusters.

**Table 1** Resource data volume of different cluster sizes
Resource Type (Theoretical Upper Limit)	50 Nodes	200 Nodes	1,000 Nodes	2,000 Nodes
Total etcd storage capacity	200 MiB	500 MiB	2 GiB	4 GiB
Total size of etcd objects of each resource type	20 MiB	50 MiB	200 MiB	400 MiB
Number of nodes	50	200	1,000	2,000
Pod (The average pod size cannot exceed 15 KB, and the number of credential-handling containers in a single pod cannot exceed 2.)	1,000 No more than 2,000 secrets or ConfigMaps can be mounted to pods.	4,000 No more than 8,000 secrets or ConfigMaps can be mounted to pods.	20,000 No more than 40,000 secrets or ConfigMaps can be mounted to pods.	40,000 No more than 80,000 secrets or ConfigMaps can be mounted to pods.
Number of pending pods in a cluster	100	400	2,000	4,000
Number of pods that can be concurrently created or deleted (The average pod size cannot exceed 15 KB, and the number of credential-handling containers in a single pod cannot exceed 2.)	100	400	2,000	4,000
Total HPA/CronHPA policies	20	40	200	400
Namespace	20	80	400	800
ConfigMap (The average resource size cannot exceed 15 KB.)	250	1,000	4,000	10,000
Secret (The average resource size cannot exceed 15 KB.)	250	1,000	4,000	10,000
PVC	500	2,000	10,000	20,000
PV	500	2,000	10,000	20,000
Service	The number of Services cannot exceed 100. The total number of backend endpoints of all Services cannot exceed 1,000. The number of LoadBalancer Services or the number of Services associated with ingresses cannot exceed 10.	The number of Services cannot exceed 400. The total number of backend endpoints of all Services cannot exceed 4,000. The number of LoadBalancer Services or the number of Services associated with ingresses cannot exceed 20.	The number of Services cannot exceed 2,000. The total number of backend endpoints of all Services cannot exceed 20,000. NOTE: When the total number of backend endpoints exceeds 10,000, kube-proxy on the node occupies a large amount of CPU resources. The number of LoadBalancer Services or the number of Services associated with ingresses cannot exceed 40.	The number of Services cannot exceed 4,000. The total number of backend endpoints of all Services cannot exceed 40,000. NOTE: When the total number of backend endpoints exceeds 10,000, kube-proxy on the node occupies a large amount of CPU resources. The number of LoadBalancer Services or the number of Services associated with ingresses cannot exceed 80.
Role	250	1,000	4,000	20,000
RoleBinding	250	1,000	4,000	20,000
CRD	40	80	200	400
Total number of CRs of all CRD types (The average CR size cannot exceed 15 KB.)	2,000	4,000	20,000	40,000
Maximum QPS of all update operations (POST, PUT, PATCH, and DELETE) in a cluster (The average size of to-be-updated objects is estimated to be no more than 15 KB.)	50 QPS	100 QPS	200 QPS	400 QPS
Maximum QPS of all read-only requests (GET, List, and Watch) in a cluster and the total amount of data read per second (Clusters v1.31 and later must support streaming output.)	100 QPS 50 MiB/s	200 QPS 100 MiB/s	400 QPS 200 MiB/s	800 QPS 300 MiB/s
Maximum QPS of GET requests for all custom CRs (15 KB per CR)	5 QPS	5 QPS	20 QPS	20 QPS
Number of persistent watch stream connections for user components to access kube-apiserver	50	200	1,000	2,000
Number of concurrent List requests to large data volume (Clusters v1.31 or later must support streaming output.)	Average data volume returned by List operations at a time: 50 MiB	Average data volume returned by List operations at a time: 100 MiB	Average data volume returned by List operations at a time: 200 MiB	Average data volume returned by List operations at a time: 300 MiB

Except for the number of nodes, other resources are recommended limits. If the cluster usage exceeds the specifications, the cluster may be overloaded. You need to select a cluster of the corresponding size based on your resource usage.
The resource data volume refers to the size of resource data in JSON format queried through the API, including the managedFields field. (You can run kubectl get {resourceType} {resourceName} -ojson -n {namespace} --show-managed-fields > tmp.json to obtain the size of a single resource. The size of the tmp.json file is the resource size.) For example, if the size of a ConfigMap is 10 KB and there are 10,000 ConfigMaps in a cluster, the total size of ConfigMap data is about 100 MiB.
The quantity of each resource is calculated based on the average size of 15 KB. If the average size of resources is greater than 15 KB, the supported data volume decreases. For example, a large-size cluster (1,000 nodes) supports 20,000 pods (with average size less than 15 KB) by default. If the average pod size is 30 KB, the maximum number of pods is reduced to 10,000.
The same calculation method applies to the request QPS. For example, if the QPS of update operations on resources with the average size less than 15 KB is 200 for a large-size cluster (1,000 nodes), and the average size of resources to be updated is 30 KB, the actual QPS decreases to 100.
The resource size in an update request refers to the resource size, not the size of the request body.

Parent Topic: Buying a Cluster

Resource Data Volume of Different Cluster Sizes

What is your overall rating for this page?