Artificial intelligence training platform enabling method, computer device, and storage medium

Abstract

The present application relates to the technical field of artificial intelligence, and discloses an artificial intelligence training platform enabling method, a computer device, and a storage medium. The method comprises: acquiring and storing configuration information of an artificial intelligence training platform to be migrated, wherein said artificial intelligence training platform is deployed on a server cluster; when the server cluster migrates from a source environment to a target environment, updating network addresses of server nodes in the server cluster to Internet protocol addresses in the target environment, and updating source Internet protocol addresses in underlying files and database data of said artificial intelligence training platform to the Internet protocol addresses in the target environment; and recovering the service of said artificial intelligence training platform on the basis of the Internet protocol addresses in the target environment and the stored configuration information, to complete enabling of said artificial intelligence training platform in the target environment. The present application implements rapid enabling of an artificial intelligence training platform in a larger range and a more complex environment.

Description

Artificial intelligence training platform activation method, computer equipment and storage media

[0001] Cross-references to related applications

[0002] This application claims priority to Chinese Patent Application No. 202510053902.X, filed on January 14, 2025, entitled “Method for Activating an Artificial Intelligence Training Platform, Computer Equipment and Storage Medium”, the entire contents of which are incorporated herein by reference. Technical Field

[0003] This application relates to the field of artificial intelligence technology, specifically to methods for enabling artificial intelligence training platforms, computer equipment, and storage media. Background Technology

[0004] With the rise of large models and deep learning, their images and corresponding applications often occupy a huge amount of storage space, which leads to a long installation time for these applications.

[0005] In related technologies, large models and deep learning images and applications are typically deployed on AI training platforms to perform deep learning and large model training operations. When an already deployed AI training platform server cluster faces machine relocation or requires factory pre-installation, in order to quickly enable the AI ​​training platform in the new environment, the IP addresses used in the AI ​​training platform need to be changed to Internet Protocol (IP) addresses that can be used in the new environment.

[0006] However, the relevant technologies can only adjust IP addresses within the same data center, and cannot adjust IP addresses when the server cluster is moved to a new environment or when the server cluster is pre-installed at the factory. This limits the rapid deployment of the artificial intelligence training platform in a wider range and more complex environments. Summary of the Invention

[0007] In a first aspect, this application provides a method for enabling an artificial intelligence training platform, comprising:

[0008] The AI ​​training platform to be migrated has been identified and deployed on a server cluster.

[0009] Obtain and save the configuration information of the AI ​​training platform to be migrated;

[0010] After migrating the server cluster from the source environment to the target environment, the Internet Protocol (IP) address of the target environment is obtained. The network addresses of the server nodes in the server cluster are then updated to the IP addresses of the target environment. The source IP addresses in the underlying files and database data of the AI ​​training platform to be migrated are also updated to the IP addresses of the target environment. Note that the source IP addresses of the source environment and the target environment belong to different network segments.

[0011] Start the service of the AI ​​training platform to be migrated. Based on the Internet Protocol address of the target environment and the saved configuration information, restore the service of the AI ​​training platform to be migrated in the target environment and complete the activation of the AI ​​training platform to be migrated in the target environment.

[0012] In one optional implementation, before acquiring and saving the configuration information of the AI ​​training platform to be migrated, the following steps are also included:

[0013] Detect whether there are any running tasks in the AI ​​training platform to be migrated; and

[0014] In response to the presence of running tasks in the AI ​​training platform to be migrated, stop and delete the running tasks.

[0015] In one optional implementation, the server cluster includes multiple server nodes, and before acquiring and saving the configuration information of the AI ​​training platform to be migrated, it also includes:

[0016] Obtain the status information and firewall status information of each server node;

[0017] The server cluster is confirmed to be in a normal state if all server nodes are in a ready state and their firewalls are enabled.

[0018] In response to the server cluster being in a normal state, the configuration information of the AI ​​training platform to be migrated is obtained and saved.

[0019] In one alternative implementation, an artificial intelligence training platform enabling method further includes:

[0020] If at least one server node's state information is not in a ready state, or if at least one server node's firewall is not enabled, the server cluster is determined to be in an abnormal state; and

[0021] In response to an abnormal state in the server cluster, a state recovery process is performed on the server cluster to restore it to a normal state.

[0022] In one alternative implementation, an artificial intelligence training platform enabling method further includes:

[0023] In response to the failure to obtain configuration information for the AI ​​training platform to be migrated, it is determined that the AI ​​training platform to be migrated has encountered an anomaly, and anomaly information is obtained so that the user can perform anomaly recovery on the AI ​​training platform to be migrated based on the anomaly information; and

[0024] After the AI ​​training platform to be migrated returns to normal, return to the steps of obtaining and saving the configuration information of the AI ​​training platform to be migrated.

[0025] In one optional implementation, the configuration information of the AI ​​training platform to be migrated is acquired and saved, including:

[0026] Obtain the status file of the container orchestration engine of the management node of the server cluster corresponding to the AI ​​training platform to be migrated;

[0027] Save the container orchestration engine's state file according to the preset data serialization format;

[0028] A snapshot of the data in the distributed key-value storage system of the server cluster corresponding to the AI ​​training platform to be migrated is performed to obtain snapshot data of the distributed key-value storage system; and

[0029] Save snapshot data of the distributed key-value store system;

[0030] The configuration information includes the container orchestration engine's state file and snapshot data from the distributed key-value store system.

[0031] In one alternative implementation, before migrating the server cluster from the source environment to the target environment, the following steps are also included:

[0032] Modify the service switching configuration file for any server node of the AI ​​training platform to be migrated to avoid abnormal access to the server node.

[0033] In one optional implementation, the service switching configuration file of any server node of the AI ​​training platform to be migrated is modified, including:

[0034] For any given server node, determine the directory access protocol configuration information from the server node's service switching configuration file; and

[0035] Remove the directory access protocol configuration item from the service switching configuration file.

[0036] The AI ​​training platform activation method provided in this embodiment avoids abnormal access to server nodes by deleting directory access protocol configuration items in the service switching configuration file, thus ensuring the smooth migration of the AI ​​training platform to be migrated.

[0037] In one alternative implementation, before updating the source Internet Protocol (IP) addresses in the underlying files and database data of the AI ​​training platform to be migrated to the IP addresses of the target environment, the method further includes:

[0038] Update the source Internet Protocol address in the firewall whitelist of the server nodes in the server cluster to the Internet Protocol address of the target environment.

[0039] In one optional implementation, updating the source Internet Protocol (IP) addresses in the underlying files and database data of the AI ​​training platform to be migrated to the IP addresses of the target environment includes:

[0040] For any server node in the server cluster, assign the corresponding Internet Protocol address for the target environment to the server node;

[0041] Based on the mapping between the source Internet Protocol addresses (IPAs) of the server nodes and the IPAs of the target environment, a template file is generated; and

[0042] Based on the template file, update the source Internet Protocol address in the underlying file of the server node of the server cluster corresponding to the AI ​​training platform to be migrated to the corresponding Internet Protocol address of the target environment, and update the source Internet Protocol address in the database data of the management node corresponding to the AI ​​training platform to be migrated to the corresponding Internet Protocol address of the target environment.

[0043] The server nodes include management nodes.

[0044] In one optional implementation, the service of the AI ​​training platform to be migrated is restored based on the Internet Protocol address of the target environment and the stored configuration information, including:

[0045] Update the source Internet Protocol address in the distributed key-value storage system certificate of the management node corresponding to the AI ​​training platform to be migrated to the Internet Protocol address of the target environment, and generate a new distributed key-value storage system certificate;

[0046] Stop the service of the distributed key-value store system;

[0047] Deleting data from a distributed key-value store system; and

[0048] Based on the saved snapshot data of the distributed key-value store system and the new distributed key-value store system certificate, the service of the distributed key-value store system in the AI ​​training platform to be migrated is restored.

[0049] The configuration information includes snapshot data from the distributed key-value storage system.

[0050] In one optional implementation, the service of the AI ​​training platform to be migrated is restored based on the Internet Protocol address of the target environment and the stored configuration information, including:

[0051] Update the source Internet Protocol address in the container orchestration engine certificate of the management node corresponding to the AI ​​training platform to be migrated to the corresponding Internet Protocol address of the target environment, and generate a new container orchestration engine certificate; and

[0052] Based on the saved container orchestration engine state file and the new container orchestration engine certificate, restore the container orchestration engine service in the AI ​​training platform to be migrated;

[0053] The configuration information includes the container orchestration engine's status file.

[0054] In one alternative implementation, migrating the server cluster from the source environment to the target environment includes:

[0055] Migrate the server cluster to the target environment by physically moving it; or,

[0056] Migrate the server cluster to the target environment according to the change in network segment usage;

[0057] When a server cluster is migrated to a target environment via physical relocation, the server cluster updates its IP addresses across network segments.

[0058] In one alternative implementation, activating the service of the AI ​​training platform to be migrated includes:

[0059] Start the Docker service for the AI ​​training platform to be migrated. Docker is an application container engine or container orchestration engine.

[0060] In one optional implementation, the preset data serialization format is a format that expresses data serialization.

[0061] In one optional implementation, updating the source Internet Protocol address in the firewall whitelist of the server nodes in the server cluster to the Internet Protocol address of the target environment includes:

[0062] Modify firewall rules using the firewall-cmd command to remove source Internet Protocol addresses from the firewall whitelist and add the target environment's Internet Protocol addresses to the firewall whitelist.

[0063] In one optional implementation, a template file is generated based on the mapping between the source Internet Protocol (IP) address of the server node and the IP address of the target environment, including:

[0064] Replace empty placeholders in the template file with the corresponding source and target Internet Protocol addresses.

[0065] Secondly, this application provides a computer device, including: a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to perform the artificial intelligence training platform activation method of the first aspect or any corresponding embodiment described above.

[0066] Thirdly, this application provides a computer-readable storage medium storing computer instructions for causing a computer to execute the artificial intelligence training platform activation method described in the first aspect or any corresponding embodiment thereof.

[0067] Fourthly, this application provides a computer program product, including computer instructions for causing a computer to execute the artificial intelligence training platform activation method described in the first aspect or any corresponding embodiment thereof. Attached Figure Description

[0068] To more clearly illustrate the technical solutions in the specific embodiments or related technologies of this application, the drawings used in the description of the specific embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0069] Figure 1 is a schematic diagram of the structure of an artificial intelligence training platform in related technologies;

[0070] Figure 2 is a flowchart illustrating an artificial intelligence training platform activation method according to one or more embodiments of this application;

[0071] Figure 3 is a flowchart illustrating another method for enabling an artificial intelligence training platform according to one or more embodiments of this application;

[0072] Figure 4 is a flowchart of the execution of a task script according to one or more embodiments of this application;

[0073] Figure 5 is a flowchart of pre-migration preparation work according to one or more embodiments of this application;

[0074] Figure 6 is a flowchart illustrating another method for enabling an artificial intelligence training platform according to one or more embodiments of this application;

[0075] Figure 7 is a flowchart illustrating the modification of the physical machine IP according to one or more embodiments of this application;

[0076] Figure 8 is a schematic diagram of the structure of an artificial intelligence training platform enabling system according to one or more embodiments of this application;

[0077] Figure 9 is a schematic diagram of the hardware structure of a computer device according to one or more embodiments of this application. Detailed Implementation

[0078] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0079] With the rise of large-scale models and deep learning, the deployment of images and applications used for these technologies often consumes enormous amounts of storage space, leading to lengthy installation times. Large-scale models, specifically Large Language Models (LLMs), are deep learning models trained on large amounts of text data that can generate natural language text or understand the meaning of language text. Large language models can handle various natural language processing tasks, such as text classification, question answering, and dialogue.

[0080] In related technologies, large models and images and applications used for deep learning are deployed on artificial intelligence training platforms, enabling users to perform deep learning and large model training operations on these platforms. The artificial intelligence training platform is a server management platform for artificial intelligence training. The entire platform manages dozens to hundreds of servers containing graphics processing units (GPUs), and distributes artificial intelligence training tasks to the servers by calling algorithms.

[0081] Figure 1 is a schematic diagram of the structure of an artificial intelligence training platform in related technologies. As shown in Figure 1, the artificial intelligence training platform includes a server cluster, meaning the platform is deployed on a server cluster, which comprises multiple server nodes (also referred to as physical machines). These server nodes can be management nodes or computing nodes. Generally, a server cluster includes one or an odd number of management nodes, with the remainder being computing nodes. Figure 1 illustrates an example of a server cluster consisting of one management node and four computing nodes.

[0082] The management node is used to deploy and manage resources, databases, cluster monitoring, user management, platform services, and other services. The compute nodes are used to deploy node services, etc. Compute nodes typically run development environments, which are the infrastructure provided to users for training artificial intelligence.

[0083] The main function of platform services and node services is to ensure the stable operation of the development environment and provide services to users.

[0084] When a server cluster of an AI training platform that has already deployed various large models and deep learning images and applications faces machine relocation or wants to achieve factory pre-installation, in order to quickly enable the AI ​​training platform in the new environment, the IP address used by the AI ​​training platform needs to be changed to an IP address that can be used in the new environment when the server cluster is migrated to the new environment, so as to achieve rapid activation of the AI ​​training platform in the new environment.

[0085] However, the relevant technologies can only adjust the IP addresses of server clusters within the same data center, which does not involve the physical migration of the server cluster. Furthermore, the adjustment of IP addresses of server clusters within the same data center is often limited to the same network segment or the network segment planned for the data center. It cannot be used to adjust IP addresses when servers are moved to a new environment or when server clusters are pre-installed at the factory. In other words, it cannot adjust IP addresses across network segments, which limits the rapid deployment of artificial intelligence training platforms in a wider range and more complex environments.

[0086] This application provides a method for enabling an artificial intelligence training platform. Before migrating a server cluster from a source environment to a target environment, the method acquires and saves the configuration information of the AI ​​training platform to be migrated. After migrating, it acquires the Internet Protocol address (IPA) of the target environment, updates the network addresses of the server nodes in the server cluster to the IPA of the target environment, updates the source IPA addresses in the underlying files and database data of the AI ​​training platform to the target environment, starts the service of the AI ​​training platform, and restores the service of the AI ​​training platform based on the IPA of the target environment and the saved configuration information. This completes the activation of the AI ​​training platform in the target environment. The source IPA of the source environment and the IPA of the target environment belong to different network segments. This allows for rapid restoration of the AI ​​training platform in a new environment when the deployed server cluster faces machine relocation or requires factory pre-installation, reducing the time required to redeploy the AI ​​training platform in a new environment and enabling the AI ​​training platform to better adapt to the needs of a wider range and more complex environments.

[0087] According to an embodiment of this application, an embodiment of an artificial intelligence training platform enabling method is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0088] This embodiment provides a method for enabling an artificial intelligence training platform, which can be used on mobile terminals such as servers and central processing units. Figure 2 is a flowchart of the method for enabling an artificial intelligence training platform according to an embodiment of this application. As shown in Figure 2, the process includes the following steps:

[0089] Step S201: Identify the AI ​​training platform to be migrated. The AI ​​training platform to be migrated is deployed on the server cluster.

[0090] Among them, the AI ​​training platform to be migrated is an AI training platform that has already been deployed on the server cluster and is in normal working order.

[0091] Step S202: Obtain and save the configuration information of the AI ​​training platform to be migrated.

[0092] After identifying the AI ​​training platform to be migrated, the configuration information of the AI ​​training platform to be migrated is obtained and saved.

[0093] Step S203: After migrating the server cluster from the source environment to the target environment, obtain the Internet Protocol address of the target environment, update the network address of the server node in the server cluster to the Internet Protocol address of the target environment, and update the source Internet Protocol address in the underlying files and database data of the AI ​​training platform to be migrated to the Internet Protocol address of the target environment. The source Internet Protocol address of the source environment and the Internet Protocol address of the target environment belong to different network segments.

[0094] After acquiring and saving the configuration information of the AI ​​training platform to be migrated, the server cluster can be migrated, which means migrating the AI ​​training platform to be migrated.

[0095] It is understandable that migrating a server cluster from the source environment to the target environment means migrating the AI ​​training platform to be migrated, which is deployed on the server cluster, from the source environment to the target environment.

[0096] It's important to note that migrating a server cluster from a source environment to a target environment can mean either physically moving the server cluster, or changing the network segment used by the server cluster. In the latter case, the migration refers to the server cluster not physically moving, but only the network segment it uses changing.

[0097] The physical movement of a server cluster can also be understood as the updating of IP addresses across network segments. This physical movement can refer to relocating the server cluster from one physical location to another, i.e., machine relocation or factory replacement of the server cluster.

[0098] Understandably, after migrating the server cluster from the source environment to the target environment, it is necessary to quickly enable the AI ​​training platform to be migrated in the target environment. First, the network addresses of the server nodes in the server cluster need to be updated to the Internet Protocol addresses (IPAs) of the target environment, and the source IPAs in the underlying files and database data of the AI ​​training platform to be migrated need to be updated to the IPAs of the target environment.

[0099] Step S204: Start the service of the AI ​​training platform to be migrated. Based on the Internet Protocol address of the target environment and the saved configuration information, restore the service of the AI ​​training platform to be migrated in the target environment, and complete the activation of the AI ​​training platform to be migrated in the target environment.

[0100] After completing step S203 above, the service of the AI ​​training platform to be migrated is started. Based on the Internet Protocol address of the target environment and the saved configuration information, the service of the AI ​​training platform to be migrated is restored, and the activation of the AI ​​training platform to be migrated in the target environment is completed.

[0101] Understandably, at this point, the basic services of the AI ​​training platform to be migrated are started, such as executing the `systemctl start docker` command to start the Docker service.

[0102] Docker is an application container engine that can serve as the underlying container engine for Kubernetes.

[0103] The AI ​​training platform activation method provided in this embodiment obtains and saves the configuration information of the AI ​​training platform to be migrated before migrating the server cluster from the source environment to the target environment. After migrating the server cluster, it obtains the Internet Protocol address (IPA) of the target environment, updates the network addresses of the server nodes in the server cluster to the IPA of the target environment, updates the source IPA addresses in the underlying files and database data of the AI ​​training platform to the source IPA addresses of the target environment, starts the service of the AI ​​training platform to be migrated, and restores the service of the AI ​​training platform to be migrated based on the IPA address of the target environment and the saved configuration information, thus completing the activation of the AI ​​training platform to be migrated in the target environment. The source IPA address of the source environment and the IPA address of the target environment belong to different network segments. This method enables the rapid restoration of normal operation of the AI ​​training platform in a new environment when the server cluster of the already deployed AI training platform faces machine relocation or requires factory pre-installation, reducing the time required to redeploy the AI ​​training platform in a new environment and allowing the AI ​​training platform to better adapt to the needs of a wider range and more complex environments.

[0104] This embodiment provides a method for enabling an artificial intelligence training platform, which can be used on mobile terminals such as servers and central processing units. Figure 3 is a flowchart of the method for enabling an artificial intelligence training platform according to an embodiment of this application. As shown in Figure 3, the process includes the following steps:

[0105] Step S301: Determine the AI ​​training platform to be migrated. The AI ​​training platform to be migrated is deployed on a server cluster. For details, please refer to step S201 of the embodiment shown in Figure 2, which will not be repeated here.

[0106] Step S302: Obtain and save the configuration information of the AI ​​training platform to be migrated.

[0107] Specifically, step S302 includes:

[0108] Step S3021: Obtain the status file of the container orchestration engine of the management node of the server cluster corresponding to the AI ​​training platform to be migrated.

[0109] Figure 4 is a flowchart illustrating the execution of the task script according to an embodiment of this application. As shown in Figure 4, by executing the task script, the state file of the container orchestration engine is first acquired and saved. The task script is implemented using Ansible, an automated operations and maintenance tool used for deploying, managing, and orchestrating computer systems. It can achieve batch processing of remotely controlled nodes based on resource lists and scripted tasks.

[0110] Specifically, the kubectl command is used to obtain the status file of the container orchestration engine of the management node of the server cluster corresponding to the AI ​​training platform to be migrated.

[0111] It should be noted that the kubectl command is a Kubernetes control command. Kubernetes (k8s) is a container orchestration engine used to manage containerized applications across multiple hosts on a cloud platform. Its goal is to make deploying containerized applications simple and efficient, providing mechanisms for application deployment, planning, updating, and maintenance. The AI ​​training platform to be migrated adopts a container orchestration engine as its underlying foundational service component.

[0112] Step S3022: Save the container orchestration engine's status file according to the preset data serialization format.

[0113] The default data serialization format is YAML. YAML is a highly readable format used for expressing data serialization.

[0114] Step S3023: Perform snapshot processing on the data in the distributed key-value storage system of the management node of the server cluster corresponding to the AI ​​training platform to be migrated, and obtain snapshot data of the distributed key-value storage system.

[0115] Specifically, by executing task scripts, snapshot data from the distributed key-value storage system is obtained and saved.

[0116] The etcdctl snapshot save command is used to create a snapshot of the data in the distributed key-value storage system of the server cluster corresponding to the AI ​​training platform to be migrated, and obtain the snapshot data of the distributed key-value storage system.

[0117] Step S3024: Save the snapshot data of the distributed key-value storage system.

[0118] Use the etcdctl snapshot save command to save snapshot data of a distributed key-value store system.

[0119] The etcdctl snapshot save command saves snapshot data from a distributed key-value store system as a data file.

[0120] It should be noted that the distributed key-value store system (etcd) is used for shared configuration, service discovery and service coordination in distributed systems or computer clusters. It helps to facilitate more secure automatic updates, coordinate work scheduled to hosts, and help set up the overlay network for containers.

[0121] The configuration information includes the container orchestration engine's state file and snapshot data from the distributed key-value store system.

[0122] Step S303: After migrating the server cluster from the source environment to the target environment, obtain the Internet Protocol address (IPA) of the target environment, update the network addresses of the server nodes in the server cluster to the IPA of the target environment, and update the source IPA addresses in the underlying files and database data of the AI ​​training platform to be migrated to the IPA addresses of the target environment. Note that the source IPA addresses of the source environment and the target environment belong to different network segments. For details, please refer to step S203 of the embodiment shown in Figure 2, which will not be repeated here.

[0123] It should be noted that after obtaining and saving the configuration information of the AI ​​training platform to be migrated, the server cluster migration can proceed. If this is a physical migration of the server cluster, the server cluster services must first be stopped—that is, all server nodes in the cluster must be shut down—and the server cluster must be powered off before proceeding with the migration. If this is not a physical migration of the server cluster, then it is not necessary to stop the server cluster services or power off the server cluster.

[0124] Step S304: Start the service of the AI ​​training platform to be migrated. Based on the target environment's Internet Protocol address and saved configuration information, restore the service of the AI ​​training platform to be migrated in the target environment, thus completing the activation of the AI ​​training platform to be migrated in the target environment. For details, please refer to step S204 of the embodiment shown in Figure 2, which will not be repeated here.

[0125] The AI ​​training platform activation method provided in this embodiment saves the state file of the container orchestration engine and the snapshot data of the distributed key-value storage system, ensuring that the AI ​​training platform to be migrated can be restored to the target environment based on the state file of the container orchestration engine and the snapshot data of the distributed key-value storage system, thereby enabling the AI ​​training platform to be migrated to be quickly activated in the target environment.

[0126] In some optional implementations, before acquiring and saving the configuration information of the AI ​​training platform to be migrated, the above-mentioned AI training platform activation method further includes:

[0127] Step a1: Detect whether there are any running tasks in the AI ​​training platform to be migrated.

[0128] Figure 5 is a flowchart of the pre-migration preparation work according to an embodiment of this application. As shown in Figure 5, the administrator logs into the management page of the AI ​​training platform to be migrated. In response to the administrator's login operation on the management page, the system checks whether there are any running tasks in the AI ​​training platform to be migrated, that is, whether there are any running tasks.

[0129] Step a2: If there are running tasks in the AI ​​training platform to be migrated, stop and delete the running tasks.

[0130] If there are running tasks in the AI ​​training platform to be migrated, i.e., if running tasks are detected, then all running tasks will be stopped and deleted.

[0131] The AI ​​training platform activation method provided in this embodiment improves the reliability of the AI ​​training platform migration by stopping and deleting running tasks in the AI ​​training platform to be migrated, thus preventing data inconsistency or corruption due to running tasks during the migration process.

[0132] In some optional implementations, the server cluster includes multiple server nodes, and the above-mentioned method for enabling the artificial intelligence training platform before acquiring and saving the configuration information of the AI ​​training platform to be migrated further includes:

[0133] Step b1: Obtain the status information and firewall enabling information for each server node.

[0134] As shown in Figure 5, when no tasks are running in the AI ​​training platform to be migrated (i.e., no running tasks are detected), a low-level check script is executed to check the status of the server cluster and determine whether the server cluster is in a normal state. This low-level check script is implemented using Ansible.

[0135] Specifically, the first step is to execute the `kubectl get node` command to obtain the status information of each server node. Here, `get node` indicates the process of retrieving server node status information.

[0136] The firewall status command `systemctl status firewalld` can be executed to obtain firewall information for each server node. `systemctl` is the system control command on the server.

[0137] Step b2: If all server nodes are in a ready state and all server nodes have firewalls enabled, then the server cluster is in a normal state.

[0138] After obtaining the status information and firewall status information of each server node, the system determines whether there are any abnormal nodes, i.e., whether the server cluster is in a normal state.

[0139] Specifically, for any given server node, determine whether the server node's status information is in a ready state and whether the server node's firewall is enabled.

[0140] If the server node's status information is "ready" and its firewall is "enabled", then the server node is considered to be functioning normally.

[0141] If the server node's status information is not in a ready state, or the server node's firewall is not enabled, then the server node is determined to be abnormal.

[0142] If all server nodes are in a ready state and all server nodes have their firewalls enabled, then there are no abnormal nodes and the server cluster is in a normal state.

[0143] Step b3: If the server cluster is in a normal state, obtain and save the configuration information of the AI ​​training platform to be migrated.

[0144] Understandably, when the server cluster is in a normal state, the step of obtaining and saving the configuration information of the AI ​​training platform to be migrated continues, that is, executing the task script to obtain and save the configuration information of the AI ​​training platform to be migrated.

[0145] The AI ​​training platform activation method provided in this embodiment ensures that all server nodes are in a ready state, thus preventing migration failure due to partial server node failure or unpreparedness. By ensuring that the firewalls on all server nodes are enabled, the security of the server cluster is guaranteed, preventing external attacks or unauthorized access and ensuring a smooth migration process for the AI ​​training platform.

[0146] In some alternative implementations, the above-mentioned method for enabling the artificial intelligence training platform further includes:

[0147] Step c1: If the status information of at least one server node is not in the ready state, or the firewall information of at least one server node is not in the enabled state, determine that the server cluster is in an abnormal state.

[0148] If at least one server node in the server cluster is not in a ready state, or if at least one server node's firewall is not enabled, i.e., there is an abnormal node, then the server cluster is determined to be in an abnormal state.

[0149] Step c2: If the server cluster is in an abnormal state, perform state recovery processing on the server cluster to restore the server cluster to a normal state.

[0150] This includes performing state recovery processing on the server cluster, which involves handling abnormal nodes, restoring their state information to a ready state, restoring firewall information to an enabled state, and thus restoring the server cluster to a normal state.

[0151] The artificial intelligence training platform activation method provided in this embodiment improves the reliability and stability of the server cluster by performing state recovery processing on the server cluster when it is in an abnormal state, thus ensuring the reliability of the migration of the artificial intelligence training platform to be migrated.

[0152] In some alternative implementations, the above-mentioned method for enabling the artificial intelligence training platform further includes:

[0153] Step d1: If the configuration information of the AI ​​training platform to be migrated is not obtained, it is determined that the AI ​​training platform to be migrated is abnormal. Abnormal information is obtained so that the user can perform abnormal recovery of the AI ​​training platform to be migrated based on the abnormal information.

[0154] As shown in Figure 5, the configuration information of the AI ​​training platform to be migrated is obtained by executing a task script. During script execution, it is determined whether the script is executing normally. If the configuration information of the AI ​​training platform to be migrated is not obtained, it indicates that the script has not executed normally, and an anomaly has been identified in the AI ​​training platform to be migrated. The anomaly information of the AI ​​training platform to be migrated is then obtained so that the user can perform anomaly recovery based on this information, i.e., anomaly handling.

[0155] Step d2: After the AI ​​training platform to be migrated returns to normal, return to the step of obtaining and saving the configuration information of the AI ​​training platform to be migrated.

[0156] The step of returning to the execution of the task script is to retrieve and save the configuration information of the AI ​​training platform to be migrated.

[0157] It should be noted that if the step of retrieving and saving the configuration information of the AI ​​training platform to be migrated is returned, but the configuration information of the AI ​​training platform to be migrated is still not obtained, then steps d1 and d2 should be re-executed until the task script is executed normally, that is, until the configuration information of the AI ​​training platform to be migrated is obtained.

[0158] The AI ​​training platform activation method provided in this embodiment determines that the AI ​​training platform to be migrated is abnormal when the configuration information of the AI ​​training platform to be migrated is not obtained, obtains the abnormal information, performs abnormal recovery on the AI ​​training platform to be migrated based on the abnormal information, and returns to the step of obtaining and saving the configuration information of the AI ​​training platform to be migrated, thus ensuring the rapid and accurate activation of the AI ​​training platform to be migrated in the target environment.

[0159] In some alternative implementations, the method for enabling the AI ​​training platform before migrating the server cluster from the source environment to the target environment also includes:

[0160] Step e1 involves modifying the service switching configuration file of any server node of the AI ​​training platform to be migrated, in order to avoid abnormal access to the server node.

[0161] As shown in Figure 4, before migrating the server cluster from the source environment to the target environment, it is necessary to modify the service switching configuration file nsswitch.conf of any server node of the AI ​​training platform to be migrated, so that the system does not access the Lightweight Directory Access Protocol (LDAP) and prevents abnormal access to the server node.

[0162] The AI ​​training platform activation method provided in this embodiment modifies the service switching configuration files of the server nodes in the server cluster before migrating the server cluster from the source environment to the target environment, thereby avoiding abnormal access to the server nodes and ensuring the smooth migration of the AI ​​training platform to be migrated.

[0163] In some optional implementations, step e1 above includes:

[0164] Step e11: For any server node, determine the directory access protocol configuration item information from the server node's service switching configuration file.

[0165] The directory access protocol configuration information is the same as the LDAP configuration information. LDAP provides access control and maintains distributed directory information via the IP protocol.

[0166] Step e12: Delete the directory access protocol configuration item information in the service switching configuration file.

[0167] After determining the directory access protocol configuration information, the directory access protocol configuration information in the service switching configuration file of each server node is deleted.

[0168] It should be noted that the sed tool can be used to delete the directory access protocol configuration item in the service switching configuration file. The sed tool is a tool for modifying server files.

[0169] The AI ​​training platform activation method provided in this embodiment avoids abnormal access to server nodes by deleting directory access protocol configuration items in the service switching configuration file, thus ensuring the smooth migration of the AI ​​training platform to be migrated.

[0170] In some alternative implementations, the AI ​​training platform activation method further includes updating the source Internet Protocol (IP) addresses in the underlying files and database data of the AI ​​training platform to be migrated to the IP addresses of the target environment before:

[0171] Step f1 updates the source Internet Protocol address in the firewall whitelist of the server nodes in the server cluster to the Internet Protocol address of the target environment.

[0172] Specifically, this involves updating the source Internet Protocol (IP) addresses in the firewall whitelists of server nodes within the server cluster to match the IP addresses of the target environment. This is achieved by modifying firewall rules using the `firewall-cmd` command to remove the source IP addresses from the firewall whitelist and add the target environment's IP addresses to the whitelist, allowing the firewall to access these whitelisted IP addresses.

[0173] The AI ​​training platform activation method provided in this embodiment enables the rapid and accurate activation of the AI ​​training platform to be migrated in the target environment by updating the source Internet Protocol address in the firewall whitelist of the server nodes in the server cluster to the Internet Protocol address of the target environment.

[0174] In some optional implementations, step S303 above includes:

[0175] Step g1: For any server node in the server cluster, assign the corresponding Internet Protocol address for the target environment to the server node.

[0176] After migrating the server cluster to the target environment, each server node is assigned an Internet Protocol address available in the target environment.

[0177] Step g2 generates a template file based on the correspondence between the source Internet Protocol address of the server node and the Internet Protocol address of the target environment.

[0178] The source Internet Protocol address is the Internet Protocol address of the server node in the server cluster in the source environment.

[0179] Obtain the mapping between the source Internet Protocol address of the server node and the Internet Protocol address of the target environment assigned to the server node.

[0180] The configuration file is updated based on the mapping between the source Internet Protocol (IP) address of the server node and the IP address of the target environment assigned to that server node. In other words, the configuration file records the mapping between the source IP address of the server node and the IP address of the target environment assigned to that server node.

[0181] Based on the configuration file, a template file is generated. Specifically, the template file contains many placeholders with empty values. When generating the template file based on the configuration file, these empty placeholders are replaced with the corresponding source Internet Protocol addresses and target environment Internet Protocol addresses, so that the database and underlying files of the AI ​​training platform to be migrated can be modified based on the template file.

[0182] Step g3: Based on the template file, update the source Internet Protocol address in the underlying files of the server nodes of the server cluster corresponding to the AI ​​training platform to be migrated to the corresponding Internet Protocol address of the target environment, and update the source Internet Protocol address in the database data of the management node corresponding to the AI ​​training platform to be migrated to the corresponding Internet Protocol address of the target environment.

[0183] The server nodes include management nodes.

[0184] It should be noted that modifications to the database apply to the management node of the AI ​​training platform. Modifications to the underlying files apply to all server nodes of the AI ​​training platform.

[0185] The template file determines the source Internet Protocol address in the underlying file of the server node, and updates the source Internet Protocol address in the underlying file to the corresponding Internet Protocol address of the target environment.

[0186] It should be further noted that after starting the service of the AI ​​training platform to be migrated, the step of updating the source Internet Protocol address in the database data of the management node corresponding to the AI ​​training platform to be migrated to the corresponding Internet Protocol address of the target environment can be performed to restore the database service.

[0187] The AI ​​training platform activation method provided in this embodiment achieves rapid and accurate activation of the AI ​​training platform in the target environment by updating the source Internet Protocol address in the underlying files of the server nodes of the server cluster corresponding to the AI ​​training platform to be migrated to the corresponding Internet Protocol address of the target environment, and updating the source Internet Protocol address in the database data of the management node corresponding to the AI ​​training platform to be migrated to the corresponding Internet Protocol address of the target environment.

[0188] In some optional implementations, step S304 above includes:

[0189] Step h1: Update the source Internet Protocol address in the distributed key-value storage system certificate of the management node corresponding to the AI ​​training platform to be migrated to the Internet Protocol address of the target environment, and generate a new distributed key-value storage system certificate.

[0190] Specifically, after starting the service of the AI ​​training platform to be migrated, the source Internet Protocol address in the distributed key-value storage system certificate of the management node corresponding to the AI ​​training platform to be migrated is updated to the Internet Protocol address of the target environment, and a new distributed key-value storage system certificate is generated using the openssl command.

[0191] Step h2: Stop the service of the distributed key-value storage system.

[0192] Step h3: Delete the data in the distributed key-value store system.

[0193] It is understandable that data generated by the distributed key-value storage system after the AI ​​training platform to be migrated is moved to the target environment should be deleted.

[0194] Step h4: Based on the saved snapshot data of the distributed key-value store system and the new distributed key-value store system certificate, restore the services of the distributed key-value store system in the AI ​​training platform to be migrated.

[0195] The configuration information includes snapshot data from the distributed key-value storage system.

[0196] The data recovery command etcdctl snapstore restore copies the saved snapshot data of the distributed key-value store system to the distributed key-value store system, and restores the service of the distributed key-value store system in the AI ​​training platform to be migrated based on the new distributed key-value store system certificate.

[0197] The AI ​​training platform activation method provided in this embodiment generates a new distributed key-value storage system certificate. Based on the saved snapshot data of the distributed key-value storage system and the new distributed key-value storage system certificate, it restores the services of the distributed key-value storage system in the AI ​​training platform to be migrated, thereby enabling the AI ​​training platform to be migrated to be quickly and accurately activated in the target environment.

[0198] In some optional implementations, step S304 above includes:

[0199] Step i1: Update the source Internet Protocol address in the container orchestration engine certificate of the management node corresponding to the AI ​​training platform to be migrated to the Internet Protocol address of the target environment, and generate a new container orchestration engine certificate.

[0200] After starting the service of the AI ​​training platform to be migrated, the source Internet Protocol address in the container orchestration engine certificate of the management node corresponding to the AI ​​training platform to be migrated is updated to the Internet Protocol address of the target environment, and a new container orchestration engine certificate is generated using the openssl command.

[0201] Step i2: Based on the saved container orchestration engine state file and the new container orchestration engine certificate, restore the container orchestration engine service in the AI ​​training platform to be migrated.

[0202] The configuration information includes the container orchestration engine's status file.

[0203] The saved state file of the container orchestration engine is copied into the container orchestration engine, and the service of the container orchestration engine in the AI ​​training platform to be migrated is restored based on the new container orchestration engine certificate.

[0204] The AI ​​training platform activation method provided in this embodiment generates a new container orchestration engine certificate, and restores the container orchestration engine service in the AI ​​training platform to be migrated based on the saved container orchestration engine status file and the new container orchestration engine certificate, thereby enabling the AI ​​training platform to be migrated to be quickly and accurately activated in the target environment.

[0205] In some alternative implementations, the above-mentioned method for enabling the artificial intelligence training platform further includes:

[0206] Step j1: After restoring the service of the AI ​​training platform to be migrated based on the Internet Protocol address of the target environment and the saved configuration information, if the AI ​​training platform to be migrated fails to be successfully enabled in the target environment, then return to the steps of obtaining the Internet Protocol address of the target environment, updating the network address of the server node in the server cluster to the Internet Protocol address of the target environment, updating the source Internet Protocol address in the underlying files and database data of the AI ​​training platform to be migrated to the Internet Protocol address of the target environment, and the subsequent steps, until the AI ​​training platform to be migrated is successfully enabled in the target environment.

[0207] If the number of times the steps of obtaining the Internet Protocol address of the target environment, updating the network address of the server node in the server cluster to the Internet Protocol address of the target environment, and updating the source Internet Protocol address in the underlying files and database data of the AI ​​training platform to be migrated to the Internet Protocol address of the target environment reaches a preset number, and the AI ​​training platform to be migrated still fails to be successfully enabled in the target environment, an alarm will be issued so that technicians can handle the problem based on the alarm.

[0208] The preset number of attempts is set by technical personnel and no specific limit is set here.

[0209] The AI ​​training platform activation method provided in this embodiment reduces activation failures caused by operational errors and improves the activation efficiency of the AI ​​training platform to be migrated. By returning to the corresponding steps until the AI ​​training platform to be migrated is successfully activated in the target environment when it fails to be activated in the target environment, the method improves the activation efficiency of the AI ​​training platform to be migrated.

[0210] This embodiment provides a method for enabling an artificial intelligence training platform, which can be used on mobile terminals such as servers and central processing units. Figure 6 is a flowchart of the method for enabling an artificial intelligence training platform according to an embodiment of this application. As shown in Figure 6, the process includes the following steps:

[0211] The first step is preparatory work before modification. The detailed process for preparatory work before modification is shown in Figure 5, and the process description is provided in the corresponding section above; it will not be repeated here.

[0212] The second step is to modify the physical machine IPs. This involves updating the network addresses of the server nodes in the server cluster to the Internet Protocol (IP) addresses of the target environment.

[0213] Specifically, Figure 7 is a flowchart illustrating the process of modifying the physical machine IP according to an embodiment of this application. As shown in Figure 7, after migrating the server cluster from the source environment to the target environment, if the migration of the server cluster is a physical migration, then the server cluster needs to be powered on; if the migration of the server cluster is not a physical migration, then the server cluster does not need to be powered on.

[0214] Next, configure the network. This involves updating the network addresses of the server nodes in the server cluster to the Internet Protocol (IP) addresses of the target environment.

[0215] It should be noted that after updating the network addresses of the server nodes in the server cluster to the Internet Protocol addresses of the target environment, it is also necessary to check the network, that is, to check whether each server node in the server cluster can connect to the network normally.

[0216] The third step is to modify the cluster IP and restore the cluster to usability.

[0217] This step corresponds to step S303 above, which involves updating the source Internet Protocol address in the underlying files and database data of the AI ​​training platform to be migrated to the Internet Protocol address of the target environment, and step S304 above. For details, please refer to the relevant descriptions of steps S303 and S304 above, which will not be repeated here.

[0218] The AI ​​training platform activation method provided in this embodiment solves the problem of long redeployment time required when the server cluster of the already deployed AI training platform faces machine relocation, factory pre-installation, or network segment modification. By modifying the IP address in the server cluster where the AI ​​training platform is deployed, the service of the AI ​​training platform in the target environment can be quickly restored and activated, making the AI ​​training platform ready to use out of the box and enriching the product form.

[0219] This embodiment provides an artificial intelligence training platform activation system. Figure 8 is a schematic diagram of the structure of the artificial intelligence training platform activation system according to an embodiment of this application. As shown in Figure 8, the artificial intelligence training platform activation system includes a pre-modification preparation module, a configuration file module, and a cluster IP modification module.

[0220] Among them, the pre-modification preparation module and the cluster IP modification module operate on the physical cluster, that is, on the server cluster where the AI ​​training platform to be migrated is deployed.

[0221] The pre-modification preparation module is used to execute the pre-modification preparation process.

[0222] The configuration file module is used as input to the module for modifying the cluster IP, and records the mapping between the source Internet Protocol address of the server node and the Internet Protocol address of the target environment assigned to that server node.

[0223] The Modify Cluster IP module is used to perform the process of modifying the cluster IP and restoring the cluster to a usable state, that is, to perform the aforementioned steps S303 and S304.

[0224] The AI ​​training platform activation system provided in this embodiment enables rapid recovery and activation of the AI ​​training platform's services in the target environment by modifying the IP addresses in the server cluster where the AI ​​training platform is deployed. This allows the AI ​​training platform to be used out of the box, enriches the product offerings, and reduces the activation time of the AI ​​training platform in the target environment.

[0225] This application also provides a computer device. Referring to Figure 9, which is a schematic diagram of the structure of a computer device according to an optional embodiment of this application, the computer device includes one or more processors 901, a memory 902, and interfaces for connecting the various components, including high-speed interfaces and low-speed interfaces. The various components communicate with each other using different buses and can be installed on a common motherboard or otherwise as needed. The processor can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on an external input / output device (such as a display device coupled to the interface). In some optional embodiments, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 9 shows an example of a single processor 901.

[0226] Processor 901 may be a central processing unit, a network processor, or a combination thereof. Processor 901 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GDA), or any combination thereof.

[0227] The memory 902 stores instructions executable by at least one processor 901 to cause at least one processor 901 to perform the method shown in the above embodiments.

[0228] The memory 902 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the computer device. Furthermore, the memory 902 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 902 may optionally include memory remotely located relative to the processor 901, and these remote memories can be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0229] The memory 902 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 902 may also include a combination of the above types of memory.

[0230] The computer device also includes a communication interface 903 for communicating with other devices or communication networks.

[0231] This application also provides a computer-readable storage medium. The methods described in this application can be implemented in hardware or firmware, or implemented as recordable on a storage medium, or implemented as computer code downloaded over a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and subsequently stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the methods shown in the above embodiments are implemented.

[0232] A portion of this application can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to this application through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0233] Although embodiments of this application have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of this application, and all such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A method for enabling an artificial intelligence training platform, characterized in that, The method includes: Identify the AI ​​training platform to be migrated, which is deployed in a server cluster; Obtain and save the configuration information of the AI ​​training platform to be migrated; After migrating the server cluster from the source environment to the target environment, the Internet Protocol (IP) address of the target environment is obtained. The network addresses of the server nodes in the server cluster are updated to the IP addresses of the target environment. The source IP addresses in the underlying files and database data of the AI ​​training platform to be migrated are also updated to the IP addresses of the target environment. The source IP addresses of the source environment and the IP addresses of the target environment belong to different network segments. The service of the AI ​​training platform to be migrated is started, and the service of the AI ​​training platform to be migrated is restored based on the Internet Protocol address of the target environment and the saved configuration information, thus completing the activation of the AI ​​training platform to be migrated in the target environment.

2. The method according to claim 1, characterized in that, Before acquiring and saving the configuration information of the AI ​​training platform to be migrated, the method further includes: Detect whether there are any running tasks in the AI ​​training platform to be migrated; and In response to the existence of a running task in the AI ​​training platform to be migrated, the running task is stopped and deleted.

3. The method according to claim 1, characterized in that, The server cluster includes multiple server nodes. Before acquiring and saving the configuration information of the AI ​​training platform to be migrated, the method further includes: Obtain the status information and firewall status information of each server node; If all server nodes are in a ready state and all server nodes have their firewalls enabled, the server cluster is determined to be in a normal state. In response to the server cluster being in a normal state, the configuration information of the AI ​​training platform to be migrated is obtained and saved.

4. The method according to claim 3, characterized in that, The method further includes: In response to the presence of at least one server node whose status information is not in a ready state, or the presence of at least one server node whose firewall is not enabled, the server cluster is determined to be in an abnormal state; and In response to the server cluster being in an abnormal state, a state recovery process is performed on the server cluster to restore it to a normal state.

5. The method according to claim 1, characterized in that, The method further includes: In response to the failure to obtain configuration information of the AI ​​training platform to be migrated, it is determined that the AI ​​training platform to be migrated has encountered an anomaly, and anomaly information is obtained so that the user can perform anomaly recovery on the AI ​​training platform to be migrated based on the anomaly information; and After the AI ​​training platform to be migrated returns to normal, return to the step of obtaining and saving the configuration information of the AI ​​training platform to be migrated.

6. The method according to claim 1, characterized in that, The process of acquiring and saving the configuration information of the AI ​​training platform to be migrated includes: Obtain the status file of the container orchestration engine of the management node of the server cluster corresponding to the AI ​​training platform to be migrated; The state file of the container orchestration engine is saved according to a preset data serialization format; A snapshot of the data in the distributed key-value storage system of the server cluster corresponding to the AI ​​training platform to be migrated is performed to obtain snapshot data of the distributed key-value storage system; and Save snapshot data of the distributed key-value storage system; The configuration information includes the state file of the container orchestration engine and the snapshot data of the distributed key-value storage system.

7. The method according to claim 1, characterized in that, Before migrating the server cluster from the source environment to the target environment, the method further includes: Modify the service switching configuration file of any server node of the AI ​​training platform to be migrated to avoid abnormal access to the server node.

8. The method according to claim 7, characterized in that, Modifying the service switching configuration file of any server node of the AI ​​training platform to be migrated includes: For any given server node, determine the directory access protocol configuration information from the server node's service switching configuration file; and Delete the directory access protocol configuration item information in the service switching configuration file.

9. The method according to claim 1, characterized in that, Before updating the source Internet Protocol (IP) addresses in the underlying files and database data of the AI ​​training platform to be migrated to the IP addresses of the target environment, the method further includes: Update the source Internet Protocol address in the firewall whitelist of the server nodes in the server cluster to the Internet Protocol address of the target environment.

10. The method according to claim 1, characterized in that, The step of updating the source Internet Protocol (IP) addresses in the underlying files and database data of the AI ​​training platform to be migrated to the IP addresses of the target environment includes: For any server node in the server cluster, assign the corresponding Internet Protocol address for the target environment to the server node; Based on the correspondence between the source Internet Protocol (IP) addresses of the server nodes and the IP addresses of the target environment, a template file is generated; and Based on the template file, the source Internet Protocol address in the underlying file of the server node of the server cluster corresponding to the AI ​​training platform to be migrated is updated to the Internet Protocol address of the corresponding target environment, and the source Internet Protocol address in the database data of the management node corresponding to the AI ​​training platform to be migrated is updated to the Internet Protocol address of the corresponding target environment. The server node includes a management node.

11. The method according to claim 1, characterized in that, The restoration of services for the AI ​​training platform to be migrated, based on the Internet Protocol address and stored configuration information of the target environment, includes: Update the source Internet Protocol address in the distributed key-value storage system certificate in the management node corresponding to the AI ​​training platform to be migrated to the Internet Protocol address of the target environment, and generate a new distributed key-value storage system certificate. Stop the service of the distributed key-value store system; Delete the data in the distributed key-value storage system; and Based on the saved snapshot data of the distributed key-value storage system and the new distributed key-value storage system certificate, the service of the distributed key-value storage system in the AI ​​training platform to be migrated is restored. The configuration information includes snapshot data of the distributed key-value storage system.

12. The method according to claim 1, characterized in that, The restoration of services for the AI ​​training platform to be migrated, based on the Internet Protocol address and stored configuration information of the target environment, includes: Update the source Internet Protocol address in the container orchestration engine certificate of the management node corresponding to the AI ​​training platform to be migrated to the corresponding Internet Protocol address of the target environment, and generate a new container orchestration engine certificate; and Based on the saved container orchestration engine state file and the new container orchestration engine certificate, the container orchestration engine service in the AI ​​training platform to be migrated is restored. The configuration information includes the status file of the container orchestration engine.

13. The method according to claim 1, characterized in that, The migration of the server cluster from the source environment to the target environment includes: Migrate the server cluster to the target environment by physical relocation; or... The server cluster will be migrated to the target environment according to the method of changing the network segment used; When the server cluster is migrated to the target environment by physical relocation, the server cluster updates its IP addresses across network segments.

14. The method according to claim 1, characterized in that, The service of activating the AI ​​training platform to be migrated includes: Start the Docker service of the AI ​​training platform to be migrated, where Docker is an application container engine or container orchestration engine.

15. The method according to claim 6, characterized in that, The preset data serialization format is a format for expressing data serialization.

16. The method according to claim 9, characterized in that, The step of updating the source Internet Protocol address in the firewall whitelist of the server nodes in the server cluster to the Internet Protocol address of the target environment includes: Modify firewall rules using the firewall-cmd command to remove source Internet Protocol addresses from the firewall whitelist and add the target environment's Internet Protocol addresses to the firewall whitelist.

17. The method according to claim 10, characterized in that, The template file generated based on the mapping relationship between the source Internet Protocol address of the server node and the Internet Protocol address of the target environment includes: Replace empty placeholders in the template file with the corresponding source and target Internet Protocol addresses.

18. A computer device, characterized in that, include: A memory and a processor are communicatively connected, the memory stores computer instructions, and the processor executes the computer instructions to perform the artificial intelligence training platform activation method according to any one of claims 1 to 17.

19. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute the artificial intelligence training platform activation method according to any one of claims 1 to 17.

20. A computer program product, characterized in that, It includes computer instructions for causing a computer to execute the artificial intelligence training platform activation method according to any one of claims 1 to 17.

Images