Scheduling method and system of single-architecture image in CPU heterogeneous k8s cluster

Abstract

The present invention proposes a scheduling method and system for a single-architecture image in a CPU heterogeneous k8s cluster, and relates to the field of computer technology. The method includes: images of different architectures are constructed using independent build scripts; the k8s scheduler is extended in the PreBind phase of the scheduler extension plug-in, and the container image of the Pod is modified according to the Node architecture assigned to the Pod, so that the container image matches the Node architecture; the extended scheduler is deployed to k8s, and a custom scheduler is used for scheduling; when scheduling the Pod, the scheduler modifies the container image of the Pod according to the assigned Node architecture in the PreBind phase. The solution proposed by the present invention realizes a solution in which the Pod of a single-architecture image service can run on any working node in a heterogeneous k8s cluster.

Description

Technical Field

[0001] The present invention belongs to the field of computer technology, and in particular to a scheduling method and system for a single-architecture image in a CPU heterogeneous k8s cluster. Background Art

[0002] In a k8s cluster with heterogeneous CPUs, services are usually required to use multi-architecture images to ensure that Pods of the same service can run on nodes with different architectures. However, there are some problems with building multi-architecture images:

[0003] 1. The base image (Base Image) that is relied on must itself be a multi-architecture image;

[0004] 2. The content of the build script (Dockerfile) cannot be bound to a specific architecture, which limits its flexibility;

[0005] 3. Docker buildx relies on QEMU virtualization when building multi-architecture images, which significantly prolongs the build time;

[0006] 4. When the image built using the QEMU simulator is run on an actual machine, there may still be compatibility issues;

[0007] The above problems make it difficult to implement some build functions in the same build script when building some complex images, and the build time is particularly long. Summary of the invention

[0008] In order to solve the above technical problems, the present invention proposes a technical solution of a scheduling method and system for a single-architecture image in a CPU heterogeneous k8s cluster to solve the above technical problems.

[0009] The first aspect of the present invention discloses a method for scheduling a single-architecture image in a CPU heterogeneous k8s cluster, the method comprising:

[0010] Step S1: Build images of different architectures using their own independent build scripts;

[0011] Step S2: Extend the k8s scheduler in the PreBind stage of the scheduler extension plug-in, modify the container image of the Pod according to the Node architecture assigned to the Pod, so that the container image matches the Node architecture, and complete the extension of the k8s scheduler;

[0012] Step S3, deploy the extended k8s scheduler to k8s, and use the customized extended k8s scheduler for scheduling; when scheduling the Pod, the extended k8s scheduler modifies the container image of the Pod according to the assigned Node architecture in the PreBind stage;

[0013] Step S4: After the Pod scheduling is completed, the kubelet starts the Pod container on the assigned node machine and pulls the image that matches the Node architecture.

[0014] According to the method of the first aspect of the present invention, in step S1, the method further includes: building different images on machines with matching architectures.

[0015] According to the method of the first aspect of the present invention, in step S1, the script naming specification is Dockerfile-CPU architecture name;

[0016] The built image name uses "-CPU architecture name" as the suffix.

[0017] According to the method of the first aspect of the present invention, in step S2, the k8s scheduler is extended in the PreBind stage of the scheduler extension plug-in, the container image of the Pod is modified according to the Node architecture assigned to the Pod, so that the container image matches the Node architecture, and the extension of the k8s scheduler is completed, including:

[0018] According to the development documentation of the Scheduling Framework, create a plug-in named arch-image-selector. In the PreBind method provided by the Scheduling Framework, the plug-in obtains the Node name assigned to the Pod, then queries the kubernetes.io / arch label value of the Node from the k8s apiserver, and stores the label value in the variable nodeArch.

[0019] After obtaining the image paths of all containers under the Pod, modify each path and split the image path into image name and version number to adapt to the assigned Node architecture;

[0020] Update the images of all containers of the modified pod to etcd through k8sapiserver.

[0021] According to the method of the first aspect of the present invention, in step S2, modifying each path includes:

[0022] Change the format to "image name + '-' + nodeArch + ':' + version number".

[0023] According to the method of the first aspect of the present invention, in step S3, deploying the extended k8s scheduler to k8s and using the customized extended k8s scheduler for scheduling includes:

[0024] When creating a service through k8s resources, set the spec.schedulerName field of the Pod to hybrid-arch-scheduler;

[0025] Among them, hybrid-arch-scheduler represents a customized extended k8s scheduler.

[0026] According to the method of the first aspect of the present invention, in step S3, modifying the format of the container image of the Pod includes:

[0027] Image name + '-' + NodeCPU architecture name + ':' + version number.

[0028] The second aspect of the present invention discloses a scheduling system for a single-architecture image in a CPU heterogeneous k8s cluster, the system comprising:

[0029] The first processing module is configured to build images of different architectures using their own independent build scripts;

[0030] The second processing module is configured to extend the k8s scheduler in the PreBind phase of the scheduler extension plug-in, modify the container image of the Pod according to the Node architecture assigned to the Pod, so that the container image matches the Node architecture, and complete the extension of the k8s scheduler;

[0031] The third processing module is configured to deploy the extended k8s scheduler into k8s and use the customized extended k8s scheduler for scheduling; when scheduling the Pod, the extended k8s scheduler modifies the container image of the Pod according to the assigned Node architecture in the PreBind stage;

[0032] The fourth processing module is configured such that after the Pod scheduling is completed, the kubelet starts the Pod container on the assigned node machine and pulls the image that matches the Node architecture.

[0033] The third aspect of the present invention discloses an electronic device. The electronic device includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the steps of any one of the scheduling methods of a single-architecture image in a CPU heterogeneous k8s cluster in the first aspect of the present disclosure are implemented.

[0034] The fourth aspect of the present invention discloses a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of any one of the scheduling methods of a single-architecture image in a CPU heterogeneous k8s cluster in the first aspect of the present disclosure are implemented.

[0035] In summary, the solution proposed in the present invention solves the slow image building and architecture compatibility problems of multi-architecture images in heterogeneous k8s cluster scenarios. Through a novel approach in the industry, a solution is implemented in which the Pod of a single-architecture image service can run on any working node in a heterogeneous k8s cluster. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0037] Figure 1 It is a flowchart of a scheduling method of a single-architecture image in a CPU heterogeneous k8s cluster according to an embodiment of the present invention;

[0038] Figure 2 A comparison diagram of a multi-architecture image construction scheme and a single-architecture image construction scheme according to an embodiment of the present invention;

[0039] Figure 3 A schematic diagram of processing a container image from the creation to the operation of a Pod according to an embodiment of the present invention;

[0040] Figure 4 It is a structural diagram of a scheduling system of a single-architecture image in a CPU heterogeneous k8s cluster according to an embodiment of the present invention;

[0041] Figure 5 The figure is a structural diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0043] The first aspect of the present invention discloses a scheduling method for a single-architecture image in a CPU heterogeneous k8s cluster. Figure 1 The flowchart of a method for scheduling a single-architecture image in a CPU heterogeneous k8s cluster according to an embodiment of the present invention is as follows: Figure 1 As shown, the method includes:

[0044] Step S1: Build images of different architectures using their own independent build scripts;

[0045] Step S2: Extend the k8s scheduler in the PreBind stage of the scheduler extension plug-in, modify the container image of the Pod according to the Node architecture assigned to the Pod, so that the container image matches the Node architecture, and complete the extension of the k8s scheduler;

[0046] Step S3, deploy the extended k8s scheduler to k8s, and use the customized extended k8s scheduler for scheduling; when scheduling the Pod, the extended k8s scheduler modifies the container image of the Pod according to the assigned Node architecture in the PreBind stage;

[0047] Step S4: After the Pod scheduling is completed, the kubelet starts the Pod container on the assigned node machine and pulls the image that matches the Node architecture.

[0048] In step S1, images of different architectures are built using their own independent build scripts.

[0049] In some embodiments, in step S1, different images are built on machines that match their architectures.

[0050] The script naming convention is Dockerfile-CPU architecture name. Taking x86 and arm64 architectures as examples, the build scripts are Dockerfile-x86 and Dockerfile-arm64 respectively.

[0051] The built image name uses "-CPU architecture name" as the suffix. 3. For example, the x86 architecture image of the service is called myservice-x86:v1.0.0, and the arm64 architecture image of the service is called myservice-arm64:v1.0.0.

[0052] Specifically, Figure 2 As shown, the construction method of a single-architecture image is different from that of a multi-architecture image. Multi-architecture images are usually built on a single machine and use the same image name. In this embodiment, the single-architecture image uses different build scripts and is independently built on machines of different architectures, and the built images need to be stored in the image repository using different names. The image name must contain "-CPU architecture name" as a suffix, such as myservice-x86:v1.0 or myservice-arm64:v1.0.

[0053] In step S2, the k8s scheduler is extended in the PreBind stage of the scheduler extension plug-in, and the container image of the Pod is modified according to the Node architecture assigned to the Pod so that the container image matches the Node architecture, thereby completing the extension of the k8s scheduler.

[0054] In some embodiments, in step S2, Figure 3 As shown, the k8s scheduler is extended in the PreBind phase of the scheduler extension plug-in, and the container image of the Pod is modified according to the Node architecture assigned to the Pod so that the container image matches the Node architecture. The extension of the k8s scheduler includes:

[0055] There are many ways to implement k8s scheduler extension. Here we choose the scheduling framework (SchedulingFramework) recommended by the community to implement it;

[0056] According to the development documentation of the Scheduling Framework, create a plug-in named arch-image-selector. In the PreBind method provided by the Scheduling Framework, the plug-in obtains the Node name assigned to the Pod, then queries the kubernetes.io / arch label value of the Node from the k8s apiserver, and stores the label value in the variable nodeArch.

[0057] After obtaining the image paths of all containers under the Pod, modify each path and split the image path into image name and version number to adapt to the assigned Node architecture;

[0058] Update the images of all containers of the modified pod to etcd through k8s apiserver.

[0059] The modification of each path includes:

[0060] Change the format to "image name + '-' + nodeArch + ':' + version number".

[0061] Specifically, in order to correctly distribute and run a single-architecture image in a k8s cluster with heterogeneous CPUs, it is necessary to enable the service Pod to select the corresponding container image according to the CPU architecture of the Node where it is located. This function can be achieved by using the scheduler extension plug-in provided by k8s.

[0062] When choosing the right container image for a Pod, you first need to determine the Node to which the Pod has been scheduled and the CPU architecture of the Node. Based on this requirement, you can extend the scheduler extension plugin in the PreBind phase. The extension code will dynamically modify the Pod's container image based on the Node architecture assigned to the Pod to match the Node architecture, thereby enabling a single architecture image to run on any Node in a heterogeneous k8s cluster.

[0063] In step S3, the extended k8s scheduler is deployed to k8s, and the customized extended k8s scheduler is used for scheduling; when scheduling the Pod, the extended k8s scheduler modifies the container image of the Pod according to the assigned Node architecture in the PreBind stage.

[0064] In some embodiments, in step S3, deploying the extended k8s scheduler to k8s and using the customized extended k8s scheduler for scheduling includes:

[0065] When creating a service through k8s resources, set the spec.schedulerName field of the Pod to hybrid-arch-scheduler;

[0066] Among them, hybrid-arch-scheduler represents a customized extended k8s scheduler.

[0067] Modify the format of the Pod's container image:

[0068] Image name + '-' + NodeCPU architecture name + ':' + version number to ensure that the image is consistent with the Node architecture, and finally save the changed Pod.

[0069] In summary, the solution proposed in the present invention can solve the slow image building and architecture compatibility problems of multi-architecture images in heterogeneous k8s cluster scenarios. In a way that has never been seen in the industry, a solution is implemented that allows the Pod of a single-architecture image service to run on any working node in a heterogeneous k8s cluster.

[0070] The second aspect of the present invention discloses a scheduling system for a single-architecture image in a CPU heterogeneous k8s cluster. Figure 4 It is a structural diagram of a scheduling system of a single-architecture image in a CPU heterogeneous k8s cluster according to an embodiment of the present invention; Figure 4 As shown, the system 100 includes:

[0071] The first processing module 101 is configured to build images of different architectures using their own independent build scripts;

[0072] The second processing module 102 is configured to extend the k8s scheduler in the PreBind phase of the scheduler extension plug-in, modify the container image of the Pod according to the Node architecture assigned to the Pod, so that the container image matches the Node architecture, and complete the extension of the k8s scheduler;

[0073] The third processing module 103 is configured to deploy the extended k8s scheduler into k8s and use the customized extended k8s scheduler for scheduling; when scheduling the Pod, the extended k8s scheduler modifies the container image of the Pod according to the assigned Node architecture in the PreBind stage;

[0074] The fourth processing module 104 is configured such that, after the Pod scheduling is completed, the kubelet starts the container of the Pod on the allocated node machine and pulls the image that matches the Node architecture.

[0075] According to the system of the second aspect of the present invention, the first processing module 101 is specifically configured to build different images on a machine with a matching architecture.

[0076] The script naming convention is Dockerfile-CPU architecture name. Taking x86 and arm64 architectures as examples, the build scripts are Dockerfile-x86 and Dockerfile-arm64 respectively.

[0077] The built image name uses "-CPU architecture name" as the suffix. 3. For example, the x86 architecture image of the service is called myservice-x86:v1.0.0, and the arm64 architecture image of the service is called myservice-arm64:v1.0.0.

[0078] Specifically, Figure 2 As shown, the construction method of a single-architecture image is different from that of a multi-architecture image. Multi-architecture images are usually built on a single machine and use the same image name. In this embodiment, the single-architecture image uses different build scripts and is independently built on machines of different architectures, and the built images need to be stored in the image repository using different names. The image name must contain "-CPU architecture name" as a suffix, such as myservice-x86:v1.0 or myservice-arm64:v1.0.

[0079] According to the system of the second aspect of the present invention, the second processing module 102 is specifically configured as follows: Figure 3As shown, the k8s scheduler is extended in the PreBind phase of the scheduler extension plug-in, and the container image of the Pod is modified according to the Node architecture assigned to the Pod so that the container image matches the Node architecture. The extension of the k8s scheduler includes:

[0080] There are many ways to implement k8s scheduler extension. Here we choose the scheduling framework (SchedulingFramework) recommended by the community to implement it;

[0081] According to the development documentation of the Scheduling Framework, create a plug-in named arch-image-selector. In the PreBind method provided by the Scheduling Framework, the plug-in obtains the Node name assigned to the Pod, then queries the kubernetes.io / arch label value of the Node from the k8s apiserver, and stores the label value in the variable nodeArch.

[0082] After obtaining the image paths of all containers under the Pod, modify each path and split the image path into image name and version number to adapt to the assigned Node architecture;

[0083] Update the images of all containers of the modified pod to etcd through k8s apiserver.

[0084] The modification of each path includes:

[0085] Change the format to "image name + '-' + nodeArch + ':' + version number".

[0086] Specifically, in order to correctly distribute and run a single-architecture image in a k8s cluster with heterogeneous CPUs, it is necessary to enable the service Pod to select the corresponding container image according to the CPU architecture of the Node where it is located. This function can be achieved by using the scheduler extension plug-in provided by k8s.

[0087] When choosing the right container image for a Pod, you first need to determine the Node to which the Pod has been scheduled and the CPU architecture of the Node. Based on this requirement, you can extend the scheduler extension plugin in the PreBind phase. The extension code will dynamically modify the Pod's container image based on the Node architecture assigned to the Pod to match the Node architecture, thereby enabling a single architecture image to run on any Node in a heterogeneous k8s cluster.

[0088] According to the system of the second aspect of the present invention, the third processing module 103 is specifically configured to deploy the extended k8s scheduler to k8s, and use the customized extended k8s scheduler for scheduling, including:

[0089] When creating a service through k8s resources, set the spec.schedulerName field of the Pod to hybrid-arch-scheduler;

[0090] Among them, hybrid-arch-scheduler represents a customized extended k8s scheduler.

[0091] Modify the format of the Pod's container image:

[0092] Image name + '-' + NodeCPU architecture name + ':' + version number to ensure that the image is consistent with the Node architecture, and finally save the changed Pod.

[0093] The third aspect of the present invention discloses an electronic device. The electronic device includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the steps of any one of the first aspects of the present invention in the scheduling method of a single-architecture image in a CPU heterogeneous k8s cluster are implemented.

[0094] Figure 5 is a structural diagram of an electronic device according to an embodiment of the present invention, such as Figure 5 As shown, the electronic device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the electronic device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, near field communication (NFC) or other technologies. The display screen of the electronic device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the electronic device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the housing of the electronic device, or an external keyboard, touchpad or mouse, etc.

[0095] Those skilled in the art will understand that Figure 5The structure shown in the figure is only a structural diagram of the part related to the technical solution of the present disclosure, and does not constitute a limitation on the electronic device to which the technical solution of the present application is applied. The specific electronic device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0096] The fourth aspect of the present invention discloses a computer-readable storage medium. A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, the steps in the method for scheduling a single-architecture image in a CPU heterogeneous k8s cluster according to any one of the first aspects of the present invention are implemented.

[0097] Please note that the technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification. The above embodiments only express several implementation methods of the present application, and their descriptions are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that for ordinary technicians in this field, without departing from the concept of the present application, several variations and improvements can be made, which all belong to the scope of protection of the present application. Therefore, the scope of protection of the patent in this application shall be based on the attached claims.

Claims

1. A scheduling method for a single-architecture image in a CPU heterogeneous k8s cluster, characterized in that: The method comprises: Step S1: Build images of different architectures using their own independent build scripts; Step S2: Extend the k8s scheduler in the PreBind stage of the scheduler extension plug-in, modify the container image of the Pod according to the Node architecture assigned to the Pod, so that the container image matches the Node architecture, and complete the extension of the k8s scheduler; In step S2, the k8s scheduler is extended in the PreBind phase of the scheduler extension plug-in, and the container image of the Pod is modified according to the Node architecture assigned to the Pod so that the container image matches the Node architecture. The extension of the k8s scheduler is completed, including: According to the development documentation of the Scheduling Framework, create a plug-in named arch-image-selector. In the PreBind method provided by the Scheduling Framework, the plug-in obtains the Node name assigned to the Pod, then queries the kubernetes.io / arch label value of the Node from the k8s apiserver, and stores the label value in the variable nodeArch. After obtaining the image paths of all containers under the Pod, modify each path and split the image path into image name and version number to adapt to the assigned Node architecture; Update the images of all containers of the modified pod to etcd through k8sapiserver; Step S3, deploy the extended k8s scheduler to k8s, and use the customized extended k8s scheduler for scheduling; when scheduling the Pod, the extended k8s scheduler modifies the container image of the Pod according to the assigned Node architecture in the PreBind stage; Step S4: After the Pod scheduling is completed, the kubelet starts the Pod container on the assigned node machine and pulls the image that matches the Node architecture.

2. According to claim 1, a scheduling method for a single-architecture image in a CPU heterogeneous k8s cluster is characterized in that: In the step S1, it also includes: different images are built on machines matching the architecture.

3. According to claim 1, a scheduling method for a single-architecture image in a CPU heterogeneous k8s cluster is characterized in that: In step S1, the script naming convention is Dockerfile-CPU architecture name; The built image name uses "-CPU architecture name" as the suffix.

4. According to claim 1, a scheduling method for a single-architecture image in a CPU heterogeneous k8s cluster is characterized in that: In step S2, modifying each path includes: Change the format to "image name + '-' + nodeArch + ':' + version number".

5. According to claim 1, a scheduling method for a single-architecture image in a CPU heterogeneous k8s cluster is characterized in that: In step S3, deploying the extended k8s scheduler to k8s and using the customized extended k8s scheduler for scheduling includes: When creating a service through k8s resources, set the spec.schedulerName field of the Pod to hybrid-arch-scheduler; Among them, hybrid-arch-scheduler represents a customized extended k8s scheduler.

6. According to claim 1, a scheduling method for a single-architecture image in a CPU heterogeneous k8s cluster is characterized in that: In step S3, modifying the format of the container image of the Pod includes: Image name + '-' + NodeCPU architecture name + ':' + version number.

7. A scheduling system for a single-architecture image in a CPU heterogeneous k8s cluster, characterized in that: The system comprises: The first processing module is configured to build images of different architectures using their own independent build scripts; The second processing module is configured to extend the k8s scheduler in the PreBind phase of the scheduler extension plug-in, modify the container image of the Pod according to the Node architecture assigned to the Pod, so that the container image matches the Node architecture, and complete the extension of the k8s scheduler; The k8s scheduler is extended in the PreBind phase of the scheduler extension plug-in, and the container image of the Pod is modified according to the Node architecture assigned to the Pod so that the container image matches the Node architecture. The k8s scheduler extension includes: According to the development documentation of the Scheduling Framework, create a plug-in named arch-image-selector. In the PreBind method provided by the Scheduling Framework, the plug-in obtains the Node name assigned to the Pod, then queries the kubernetes.io / arch label value of the Node from the k8s apiserver, and stores the label value in the variable nodeArch. After obtaining the image paths of all containers under the Pod, modify each path and split the image path into image name and version number to adapt to the assigned Node architecture; Update the images of all containers of the modified pod to etcd through k8sapiserver; The third processing module is configured to deploy the extended k8s scheduler into k8s and use the customized extended k8s scheduler for scheduling; when scheduling the Pod, the extended k8s scheduler modifies the container image of the Pod according to the assigned Node architecture in the PreBind stage; The fourth processing module is configured such that after the Pod scheduling is completed, the kubelet starts the Pod container on the assigned node machine and pulls the image that matches the Node architecture.

Images