Introduction to CCE Cloud Bursting Engine for CCI

Overview

This section describes the functions, required resources, and custom annotations of CCE Cloud Bursting Engine for CCI (also called the bursting add-on).

Functions

The figure shows how workloads running in a CCE cluster with the bursting add-on installed are scheduled to CCI. In this process, you only need to take care of workload creation and scheduling.

Scheduling workloads to CCI is closely related to your workload settings.

Pod Configuration Item	Description	Specification	Reference
Scheduling	You can manage pods in a CCE cluster in multiple ways to control their scheduling to CCI. You can also configure scheduling policies to improve cluster resource utilization.	There are four scheduling policies. There are two methods for managing scheduling policies. Multiple virtual nodes can be scheduled.	Scheduling Pods to CCI
Resource quotas	You can configure fields such as cpu and memory to specify resource specifications and limits for containers in a pod. The add-on selects appropriate resource specifications to an integer for better resource utilization.	Appropriate pod resource quotas are selected.	Resource Specifications
Images	You can configure service images and run your service containers in Huawei Cloud CCE clusters and CCI.	The image configuration mode can be changed. Images can be upgraded in-place.	Images
Storage	You can mount storage volumes to workloads for persistent data storage.	There are multiple storage volume types. The hostPath volume of a workload can be replaced.	Storage
Networking	You can plan the network topology between CCE clusters and CCI.	Pods in CCI can be exposed by the Service. Pods in a CCE cluster can communicate with pods in CCI through the Service. cluster-dns can be specified. Global EIPs can be bound to pods.	Networking
Logging	You can install the Cloud Native Logging and bursting add-ons at the same time to collect pod logs in CCI.	The cloud native logging add-on can report logs of pods scheduled to CCI. Logs can be automatically dumped.	Logging
Monitoring	You can install add-ons to connect to Monitoring Center for better O&M of pods scheduled to CCI.	Monitoring Center can connect to AOM. Monitoring Center can connect to Prometheus.	Monitoring

Resource Usage Description

The following table describes the cloud services involved when the bursting add-on is installed in a CCE cluster.

Involved Cloud Service	Resource Description	Remarks
CCI	A namespace called cce-burst-{CCE cluster ID} will be created for the bursting add-on in CCI.	Do not use this namespace in CCI. If you need to use CCI, create a namespace. CCI namespaces are free for use.
CCE	Workloads, secrets, ConfigMaps, PVs, and PVCs in CCE are synchronized to CCI and occupy CCE node resources.	If there are 1,000 pod and 1,000 ConfigMaps in your CCE cluster, you can apply for 2 cores and 4 GiB of memory. If there are 2,000 pod and 2,000 ConfigMaps in your CCE cluster, you can apply for 4 cores and 8 GiB of memory. If there are 4,000 pod and 4,000 ConfigMaps in your CCE cluster, you can apply for 8 cores and 16 GiB of memory.
ELB	If pods in a CCE cluster can communicate with the pods CCI through the Service, the bursting add-on automatically creates a shared load balancer.	The shared load balancer is named cce-lb-xxx. When the add-on is uninstalled, the load balancer is automatically deleted.
VPC	Pods scheduled to CCI use the same VPC as the CCE cluster.	The CIDR block of the Service in the CCI namespace is 10.247.0.0/16. Do not configure the same CIDR block for the VPC subnet of the CCE cluster.
SWR	When you create a workload in CCE, the image can be pulled from SWR.	Ensure that your image has been pushed to SWR.

Pod Annotations

If the bursting add-on is installed in a CCE cluster, custom annotations are required. The following table describes the annotations.

Annotation Key	Description	Reference
scheduling.cci.io/managed-by-profile	Profile that manages the pods scheduled to CCI	Scheduling Pods to CCI
virtual-kubelet.cci.io/burst-pod-cpu	vCPUs of pods after they are scheduled to CCI	Resource Specifications
virtual-kubelet.cci.io/burst-pod-memory	Memory of pods they are scheduled to CCI	Resource Specifications
coordinator.cci.io/inject-volumes	Annotation injected by the cloud native logging add-on for collecting CCI pod logs	Logging
logconf.k8s.io/fluent-bit-configmap-reference	Annotation injected by the cloud native logging add-on for collecting CCI pod logs	Logging
logconfigs.logging.openvessel.io	Annotation injected by the cloud native logging add-on for collecting CCI pod logs	Logging
sandbox-volume.openvessel.io/volume-names	Annotation injected by the cloud native logging add-on for collecting CCI pod logs	Logging
coordinator.cci.io/image-replacement	Annotation for replacing image path prefixes	Images