Pod deployment method and apparatus

Patent metadata

Field	Value
Publication number	US-12423126-B2
Application number	US-202217752102-A
Country	US
Kind code	B2
Filing date	May 24, 2022
Priority date	Nov 26, 2019
Publication date	Sep 23, 2025
Grant date	Sep 23, 2025

Abstract

A master node receives an instruction for deploying one or more scheduling domains, where the instruction includes a quantity of each type of resources occupied by each of the scheduling domains. The master nodeselects a worker node for deploying the one or more scheduling domains, and sends, to the worker node, the instruction for deploying the one or more scheduling domains. The master node further sends, to the worker node based on association information of a scheduling domain, an instruction for deploying one or more pods, where the instruction for deploying the one or more pods includes a quantity of pods, a quantity of containers included in each pod, and an identifier of the scheduling domain to which resources used by the one or more pods belong.

First claim

What is claimed is:

1. A pod deployment method, wherein the method comprises:
receiving, by a master node, a first instruction for deploying one or more scheduling domains, wherein the first instruction comprises a quantity of each type of resources occupied by each of the one or more scheduling domains, and the resources occupied by the one or more scheduling domains are used by at least one pod of one or more pods;
selecting, by the master node, a worker node for deploying the one or more scheduling domains, and sending, to the worker node, the first instruction for deploying the one or more scheduling domains;
receiving, by the master node, a second instruction for deploying the one or more pods, wherein the second instruction for deploying the one or more pods comprises a quantity of pods, a quantity of containers comprised in each pod, and association information of a scheduling domain of the one or more scheduling domains; and
sending, by the master node to the worker node based on the association information of the scheduling domain, the second instruction for deploying the one or more pods, the second instruction for deploying the one or more pods further comprises an identifier of the scheduling domain to which resources used by the one or more pods belong, wherein
a quantity resources occupied by the scheduling domain comprises a quantity of central processing units (CPUs) occupied by the scheduling domain,
the first instruction for deploying the one or more scheduling domains further comprises a quantity of CPUs for isolated use of a container in a pod of the one or more pods and a quantity of CPUs for shared use of containers in the pod, and
the second instruction for deploying the one or more pods further comprises an identifier of a first container that uses a CPU in isolation from another container.

2. The method according to claim 1 , wherein for any type of resources that are occupied by the scheduling domain and that are used by a pod of the one or more pods, a quantity of the type of resources is less than a sum of a maximum quantity of the type of resources needed by each container when working, and is not less than a maximum quantity of the type of resources needed by any container when working.

3. The method according to claim 1 , wherein the association information of the scheduling domain is the identifier of the scheduling domain, a type of the scheduling domain, or an identifier of the worker node.

4. The method according to claim 1 , wherein after the sending, by the master node to the worker node, the first instruction for deploying the one or more scheduling domains, the method further comprises:
updating, by the master node, a remaining quantity of each type of resources in the worker node in a resource list, and updating a remaining quantity of each type of resources in a resource pool, wherein the resource pool comprises a plurality of worker nodes.

5. The method according to claim 1 , wherein the selecting, by the master node, a worker node for the deploying one or more scheduling domains comprises:
selecting, by the master node from the plurality of worker nodes according to an optimal resource policy, the worker node for deploying the one or more scheduling domains; or
when the first instruction for deploying the one or more scheduling domains received by the master node further comprises an identifier of the worker node, selecting the worker node corresponding to the identifier as the worker node for deploying the one or more scheduling domains.

6. The method according to claim 1 , further comprising:
receiving, by the master node, a third instruction for deleting the scheduling domain, wherein the third instruction comprises the identifier of the scheduling domain;
determining, by the master node, whether there is a pod of the one or more pods that uses a resource in the scheduling domain with the identifier; and
in response to determining that the pod uses the resource in the scheduling domain with the identifier, sending prompt information indicating that the scheduling domain cannot be deleted; and
in response to determining that the pod does not use the resource in the scheduling domain with the identifier, sending the third instruction for deleting the scheduling domain to a worker node on which the scheduling domain is deployed.

7. A pod deployment method, wherein the method comprises:
receiving, by a worker node, a first instruction for deploying one or more scheduling domains, wherein the first instruction comprises a quantity of each type of resources occupied by each of the one or more scheduling domains;
deploying the one or more scheduling domains, wherein the one or more scheduling domains comprise a corresponding quantity of a corresponding type of resources;
receiving, by the worker node, a second instruction for deploying one or more pods, wherein the second instruction for deploying one or more pods comprises a quantity of pods, a quantity of containers comprised in each pod, and an identifier of a scheduling domain of the one or more scheduling domains to which resources used by the one or more pods belong; and
deploying, by the worker node, the one or more pods, wherein each pod in the one or more pods comprises a quantity of containers; and controlling, by the worker node based on a container working, the container to use a resource in the scheduling domain, wherein
a quantity resources occupied by the scheduling domain comprises a quantity of central processing units (CPUs) occupied by the scheduling domain,
the first instruction for deploying the one or more scheduling domains further comprises a quantity of CPUs for isolated use of a container in a pod of the one or more pods and a quantity of CPUs for shared use of containers in the pod, and
the second instruction for deploying the one or more pods further comprises an identifier of a first container that uses a CPU in isolation from another container.

8. The method according to claim 7 , wherein for any type of resources that are occupied by the scheduling domain and that are used by a pod of the one or more pods, a quantity of the type of resources is less than a sum of a maximum quantity of the type of resources needed by each container when working, and is not less than a maximum quantity of the type of resources needed by any container when working.

9. The method according to claim 7 , wherein
the controlling, by the worker node based on the container working, the container to use a resource in the scheduling domain comprises:
controlling, by the worker node when the first container is working, the first container to use the CPU for isolated use in the scheduling domain, and controlling, when another container different from the first container is working, the other container to use another CPU that is in the scheduling domain and that is not for isolated use.

10. A pod deployment apparatus, wherein the apparatus comprises a transceiver; at least one processor; and
one or more memories coupled to the at least one processor and storing programming instructions for execution by the at least one processor to cause the pod deployment apparatus to:
receive a first instruction for deploying one or more scheduling domains, wherein the first instruction comprises a quantity of each type of resources occupied by each of the one or more scheduling domains, and the resources occupied by the one or more scheduling domains are used by at least one pod of one or more pods;
select a worker node for deploying the one or more scheduling domains,
send, to the worker node, the first instruction for deploying the one or more scheduling domains;
receive a second instruction for deploying the one o

Assignees

• Huawei Tech Co Ltd

Inventors

• Zhang Wenbo
• Xia Hai

Classifications

• G06F2209/5011

  Pool · CPC title

• G06F2009/4557

  Distribution of virtual machine instances; Migration and load balancing · CPC title

• G06F2009/45562

  Creating, deleting, cloning virtual machine instances · CPC title

• G06F9/5055

  considering software capabilities, i.e. software resources associated or available to the machine · CPC title

• G06F9/4881

  Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues · CPC title