A full-process automation method and system for a bare metal server

Abstract

The application provides a full-process automatic management method and system for a bare metal server, and the method comprises the following steps: automatically obtaining initial configuration information of a server to be managed; automatically allocating an out-of-band management network address for the server to be managed; controlling the server to be managed to start and load a memory operating system through a network, obtaining state information according to an identity in the memory operating system, and the state information is used to determine a preset processing operation to be executed; determining a target operation to be executed from the preset processing operation by a proxy process and executing the target operation; after the target operation is executed, if the state information indicates that node registration is needed, a node registration request is initiated to a cloud platform bare metal service, and the server to be managed is registered as a bare metal node; and after the server to be managed is successfully registered as the bare metal node, a switch port connected with the server to be managed is automatically configured, and batch installation verification is performed. The application realizes full-process automatic management of the bare metal server.

Description

Technical Field

[0001] This invention relates to the field of cloud computing technology, specifically to a fully automated management method and system for bare metal servers. Background Technology

[0002] With the rapid development of fields such as artificial intelligence and high-performance computing, traditional virtualization technologies, due to issues such as virtualization loss and resource contention, are struggling to meet the performance requirements of scenarios like large-scale model training and high-performance databases. Bare metal servers, as a form of physical server, offer advantages such as no virtualization loss, powerful performance, and good hardware isolation, and are widely used in intelligent computing centers and high-performance computing clusters. The OpenStack cloud platform natively supports bare metal server management through its Ironic component, allowing users to register physical servers as bare metal nodes and manage their lifecycle uniformly through the cloud platform.

[0003] The existing bare metal server management process suffers from several problems: First, it involves numerous manual steps, leading to low efficiency. Before use, bare metal servers need to register with the OpenStack Ironic service. Registration requires a large amount of configuration information, including IPMI IP address, account password, PXE NIC MAC address, service NIC MAC address, switch port information, and server configuration details. Furthermore, for ease of maintenance, server serial number (SN) and rack unit location information also need to be entered. Traditionally, after the server is installed, data center maintenance personnel manually enter the SN and rack unit location information, manually enter the BIOS to configure IPMI and various BIOS settings, network personnel configure the switch ports, and finally, cloud computing maintenance personnel compile all information and register with the Ironic service. When the number of servers reaches hundreds or thousands, the entire process can take days or even weeks, severely hindering the rapid delivery of intelligent computing projects. Second, existing automation solutions are incomplete and lack quality assurance. For example, scripts are used for partial automation, such as batch registering nodes and batch executing tests by importing Excel spreadsheets. However, these solutions still have the following shortcomings: First, they fail to achieve a fully automated closed loop from information acquisition to node registration, requiring manual coordination between each step; second, they lack an automatic verification mechanism for server hardware configuration, making it impossible to detect hardware configuration errors or omissions before management; third, they lack a server stress testing process, failing to verify server stability under high load, leading to frequent problems and high repair costs. Thirdly, there is insufficient capacity for large-scale delivery. As the scale of intelligent computing data centers continues to expand, the number of servers delivered in a single batch has increased from dozens to thousands. Traditional manual or semi-automated management methods can no longer support the demands of large-scale delivery, becoming a bottleneck restricting the implementation of intelligent computing projects.

[0004] Therefore, there is an urgent need for a fully automated management method and system for bare metal servers that can integrate asset identification, hardware verification, stress testing, data cleaning, out-of-band management network configuration, BIOS configuration, node registration, switch configuration, and installation verification into a unified automated process, achieving unattended management from server racking to business readiness. Summary of the Invention

[0005] In view of this, it is necessary to provide a fully automated management method and system for bare metal servers to solve the technical problems of low delivery efficiency, high failure rate and difficulty in large-scale deployment caused by the reliance on manual operation or semi-automated scripts in existing bare metal server management methods, which cannot achieve a fully automated closed loop from information acquisition to node registration, and lack a hardware quality assurance mechanism before management.

[0006] To address the aforementioned technical problems, in a first aspect, the present invention provides a fully automated management method for bare-metal servers, comprising: Automatically obtain the initial configuration information of the server to be managed, including its identity, physical location, and expected hardware configuration information; Automatically assign out-of-band management network addresses to the servers to be managed according to the preset network planning strategy; The server to be managed is controlled to start up and load a memory operating system via the network. The agent process running in the memory operating system obtains the status information of the server to be managed based on the identity identifier. The status information is used to determine the preset processing operation to be executed. The agent process determines the target operation to be executed from the preset processing operations and executes the target operation; the preset processing operations include at least one of hardware configuration verification, hardware stability verification, data cleanup, and node registration pre-configuration, and the node registration pre-configuration includes obtaining the network connection information of the server to be managed; After the target operation is completed, if the status information indicates that node registration is required, the identity identifier, the out-of-band management network address and the network connection information are used as registration parameters to initiate a node registration request to the cloud platform bare metal service, and the server to be managed is registered as a bare metal node. After the server to be managed is successfully registered as a bare metal node, the switch port connected to the server to be managed is automatically configured according to the network connection information. Perform batch installation verification on registered bare metal nodes.

[0007] In one possible implementation, the automatic acquisition of the initial configuration information of the server to be managed includes: The initial configuration information is automatically obtained by reading the radio frequency identification tag or near field communication tag set on the server to be managed; or, By reading the power data of the intelligent power distribution unit and the power consumption characteristics generated by the agent process, the identity and physical location are determined, and the expected hardware configuration information is obtained based on manual input.

[0008] In one possible implementation, the step of automatically assigning out-of-band management network addresses to the server to be managed according to a preset network planning strategy includes: If the network planning strategy is static allocation, then a static out-of-band management network address is allocated to the server to be managed according to the physical location, and the out-of-band management network address is recorded as static. If the network planning strategy is dynamic allocation, then the out-of-band management network address is recorded as a Dynamic Host Configuration Protocol (DHCP) type.

[0009] In one possible implementation, when the target operation is hardware configuration verification, the target operation is executed, including: Collect the actual hardware configuration information of the server to be managed, and compare the actual hardware configuration information with the expected hardware configuration information; Obtain the log data of the server to be managed; When the comparison results are inconsistent, or the log data indicates that there is an abnormal hardware status, the fault handling mechanism is triggered. The fault handling mechanism includes generating email alerts or, by calling a command to illuminate the fault indicator light on the server for physical location.

[0010] In one possible implementation, when the target operation is hardware stability verification, the target operation is executed, including: Perform stress tests on the central processing unit, memory, storage media, and graphics processor of the server to be managed; Collect stress test results of all servers of the same model to be managed and compare them horizontally. If the test data of a certain server to be managed deviates from the average data of other servers of the same model to be managed by more than a threshold, the fault handling mechanism will be triggered. If a hardware error occurs during the stress test, the fault indicator light of the server to be managed will be lit by command call to physically locate the error.

[0011] In one possible implementation, when the target operation is data cleaning, the target operation is performed, including: A light cleanup of the storage medium of the server to be managed is performed to clean up the metadata, or a deep cleanup is performed by overwriting multiple random data entries.

[0012] In one possible implementation, when the target operation is a node registration pre-configuration, the target operation is executed, including: Obtain the network connection information of the server to be managed, including the network card address and the connected switch port information; Configure the out-of-band management system of the server to be managed. The configuration of the out-of-band management system includes setting the out-of-band management account password and setting the out-of-band management network address of the server to be managed according to the type of the out-of-band management network address. Configure the basic input / output system of the server to be managed. The configuration of the basic input / output system includes at least one of the following: virtualization switch, hyper-threading switch, power performance mode, startup mode, serial port redirection, and system time.

[0013] In one possible implementation, the step of automatically configuring the switch port connected to the server to be managed based on the network connection information includes: The switch port to be configured is determined based on the switch port information, and the correspondence between the switch port and the server to be managed is verified based on the network card address. Configure the switch port as an aggregated independent port mode so that each member port of the switch port can be used independently for network startup during the operating system installation phase, and after the operating system installation is completed, the switch port can be used by the server to be managed through link aggregation.

[0014] In one possible implementation, performing batch installation verification on the registered bare metal nodes includes: Call the OpenStack Nova application programming interface to perform batch installation operations on registered bare metal nodes; Perform network connectivity testing on bare metal nodes after installation. If the nodes are connected within the preset timeout period, the verification is confirmed as successful; otherwise, an alarm is generated.

[0015] Secondly, the present invention also provides a fully automated management system for bare metal servers, comprising: The information acquisition module is used to automatically acquire the initial configuration information of the server to be managed, including the identity identifier, physical location and expected hardware configuration information. The network planning module is used to automatically allocate out-of-band management network addresses to the servers to be managed according to a preset network planning strategy. The startup control module is used to control the server to be managed to start up via the network and load the memory operating system; The agent process module runs in the memory operating system and is used to obtain the status information of the server to be managed based on the identity identifier, and to determine and execute the target operation to be executed from the preset processing operations based on the status information; the preset processing operations include at least one of hardware configuration verification, hardware stability verification, data cleaning, and node registration pre-configuration, and the node registration pre-configuration includes obtaining the network connection information of the server to be managed. The node registration module is used to initiate a node registration request to the cloud platform bare metal service by using the identity identifier, the out-of-band management network address and the network connection information as registration parameters after the target operation is completed and the status information indicates that node registration is required, and register the server to be managed as a bare metal node. The switch configuration module is used to automatically configure the switch port connected to the server to be managed according to the network connection information after the server to be managed is successfully registered as a bare metal node. The installation verification module is used to perform batch installation verification on registered bare metal nodes.

[0016] The beneficial effects of this invention are as follows: The fully automated bare-metal server management method provided by this invention automatically executes each step, driven by status information, from the automatic acquisition of initial server configuration information to the automatic allocation of out-of-band management network addresses, and then to hardware verification, stress testing, data cleanup, node registration pre-configuration, node registration, switch port configuration, and batch installation verification, without manual intervention. Compared with existing technologies, the management time for a single server to be managed is reduced from several hours to minutes, and the delivery cycle for thousands of servers is reduced from several weeks to several hours, greatly improving the rapid deployment capability of intelligent computing projects. At the same time, it provides reusable automation capabilities for large-scale, high-frequency bare-metal server delivery, exhibiting good engineering scalability.

[0017] Furthermore, this invention performs hardware configuration verification and hardware stability verification before node registration, intercepting quality problems at the management stage, avoiding problems, and reducing failure rate and repair costs.

[0018] Furthermore, this invention automatically obtains the network connection information of the server to be managed during the pre-registration configuration stage of the node, and automatically configures the connected switch port according to the network connection information after the node is registered. This realizes an automated closed loop from obtaining the information of the server to be managed to the switch configuration, eliminating manual intervention on the network side and reducing the communication cost and configuration error risk of cross-team collaboration.

[0019] Meanwhile, this invention supports flexible combinations of operations such as verification, stress testing, cleanup, and registration for different servers to be managed, through state-driven preset processing operations, which can adapt to the management needs of different customers and different scenarios. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 A schematic flowchart of an embodiment of the fully automated management method for bare metal servers provided by the present invention; Figure 2 For the present invention Figure 1 A schematic flowchart of an embodiment of step S102; Figure 3 A schematic diagram of an embodiment of hardware configuration verification provided by the present invention; Figure 4 A schematic diagram of an embodiment of hardware stability verification provided by the present invention; Figure 5 A schematic diagram of an embodiment of data cleaning provided by the present invention; Figure 6 A schematic diagram illustrating a specific embodiment of the node registration pre-configuration and configuration-based registration provided by the present invention; Figure 7 A schematic diagram of a specific embodiment of the fully automated management method for bare metal servers provided by the present invention; Figure 8 This is a schematic diagram of an embodiment of the fully automated management system for bare metal servers provided by the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0023] It should be understood that the illustrative drawings are not drawn to scale. The flowcharts used in this invention illustrate operations implemented according to some embodiments of the invention. It should be understood that the operations in the flowcharts may be implemented out of order, and steps without logical contextual relationships may be reversed or performed simultaneously. Furthermore, those skilled in the art, guided by the content of this invention, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor systems and / or microcontroller systems.

[0024] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0025] This invention provides a fully automated management method and system for bare metal servers, which will be described below.

[0026] Before describing the specific embodiments, it is important to note that the fully automated management method and system for bare metal servers proposed in this invention is an enhanced automation solution built upon the OpenStack cloud platform and its bare metal management component (Ironic). By interacting with the APIs of components such as OpenStack Ironic and Nova, the previously scattered manual operation steps, including automatic asset information acquisition, automatic out-of-band network planning, hardware configuration verification, stress testing, data cleanup, automatic BIOS / IPMI setup, automatic switch port configuration, and final installation verification, are chained into a state-driven, fully automated pipeline. This ultimately registers physical servers efficiently and reliably as bare metal nodes that can be uniformly managed by the OpenStack cloud platform, thereby solving the problems of low efficiency, poor quality, and difficulty in scaling caused by relying on manual or semi-automated scripts in large-scale intelligent computing center delivery scenarios.

[0027] Specifically, a bare metal server refers to a physical server that has not yet installed an operating system. A bare metal node is a resource abstraction formed within the OpenStack cloud platform after a bare metal server is managed. The core difference between the two is that a bare metal server is a hardware entity, while a bare metal node is the cloud platform's interface for managing and using that hardware. Only after completing the management process can a bare metal server become a bare metal node in the cloud platform, and then be managed by users through the Nova API, performing lifecycle management operations such as creation, startup, and destruction.

[0028] Figure 1 This is a schematic flowchart of an embodiment of the fully automated management method for bare metal servers provided by the present invention, as shown below. Figure 1 As shown, the fully automated management method for bare metal servers includes: S101. Automatically obtain the initial configuration information of the server to be managed. The initial configuration information includes the identity, physical location and expected hardware configuration information.

[0029] It should be understood that the server to be managed is the bare metal server.

[0030] The identification identifier can be the serial number (SN) of the server to be managed.

[0031] The physical location can be the rack U-position information of the server to be managed. The rack U-position is a standard unit used to measure the installation height of equipment in a data center rack. Therefore, the physical location of the server can be determined after obtaining the rack U-position information of the server to be managed.

[0032] The expected hardware configuration information refers to the server hardware specifications. This includes, but is not limited to, the model, number of cores, and clock speed of the central processing unit (CPU), the capacity, quantity, and type of memory, the interface type, capacity, and quantity of storage media, the model, video memory size, and quantity of the graphics processing unit (GPU), and the model, speed, and number of interfaces of the network card.

[0033] S102. Automatically assign out-of-band management network addresses to the servers to be managed according to the preset network planning strategy.

[0034] It should be noted that, in order to enable data management and subsequent querying and tracing, in some embodiments of the present invention, the initial configuration information and the allocation results of out-of-band management network addresses can be written into the server table servers_info.

[0035] S103. Control the server to be managed to start up and load the memory operating system (ramdisk) via the network. The agent process running in the memory operating system obtains the status information of the server to be managed based on the identity identifier. The status information is used to determine the preset processing operation to be executed.

[0036] Specifically, the agent process is IPA (ironic-python-agent), which is a core component of the Ironic project. It is designed to run in the memory operating system of bare metal servers and is responsible for performing operations such as hardware information collection, configuration distribution, and node registration.

[0037] In this context, ramdisk refers to a lightweight operating system where the operating system kernel and necessary runtime environment are fully loaded into the RAM of the server to be managed, rather than being installed on a hard drive. Specifically, after the server to be managed boots via a PXE network, it downloads and loads a pre-configured miniature operating system from the network into memory. This memory-based operating system does not depend on the server's local hard drive, so the system can still run normally even if the server's hard drive is empty or contains residual data.

[0038] S104. The agent process determines the target operation to be executed from the preset processing operations and executes the target operation. The preset processing operations include at least one of hardware configuration verification, hardware stability verification, data cleanup, and node registration pre-configuration. The node registration pre-configuration includes obtaining the network connection information of the server to be managed.

[0039] It should be noted that whether a preset processing operation is a target operation is determined by different flag bits or configuration parameters. Specifically, the configuration parameters include hardware configuration verification parameters, hardware stability verification parameters, data cleanup parameters, and node registration pre-configuration parameters. When the hardware configuration verification parameters, hardware stability verification parameters, data cleanup parameters, and node registration pre-configuration parameters are all true, it indicates that the preset processing operation is a target operation. When the hardware configuration verification parameters, hardware stability verification parameters, data cleanup parameters, and node registration pre-configuration parameters are all false, it indicates that the preset processing operation is not a target operation and does not need to be executed.

[0040] S105. After the target operation is completed, if the status information indicates that node registration is required, the identity identifier, out-of-band management network address and network connection information are used as registration parameters to initiate a node registration request to the bare metal service of the cloud platform, and the server to be managed is registered as a bare metal node.

[0041] Similarly, whether node registration is required can be determined through configuration parameters. When the configuration parameter is true, registration is required; when the configuration parameter is false, registration is not required.

[0042] S106. After the server to be managed is successfully registered as a bare metal node, the switch port connected to the server to be managed is automatically configured according to the network connection information.

[0043] S107. Perform batch installation verification on registered bare metal nodes.

[0044] The fully automated bare-metal server management method proposed in this invention constructs a complete automated closed loop from physical racking to cloud platform readiness through seven core steps: S101 establishes an initial asset ledger and quality benchmark for the management process by automatically acquiring the server's identity, physical location, and expected hardware configuration information, replacing the traditional manual data entry step; S102 provides a predictable and traceable network configuration foundation for the server's remote management channel by automatically allocating out-of-band management network addresses; S103 enables the agent process to actively acquire status information in the absence of a system by controlling the server's network startup and loading the memory operating system, transforming the server itself into a programmable execution mechanism. Unit; S104: Through the agent process, the target operation is determined and executed from the preset processing operations based on the status information. This enables the orchestratable execution of hardware configuration verification, stability verification, data cleanup, and pre-registration configuration for node registration, placing quality assurance and security cleanup before the registration stage. S105: By using the identity identifier, out-of-band address, and network connection information as parameters, a registration request is initiated to the bare metal service of the cloud platform, completing the resource abstraction transformation from physical server to cloud platform manageable node. S106: By automatically configuring the switch port based on the network connection information, the engineering problem of configuration disconnect between the network side and the server side is solved. S107: Successful management is confirmed through batch installation verification closed loop, ensuring delivery quality. Overall, this embodiment of the invention takes state-driven operation as its core, integrating scattered manual operations into a unified automated pipeline, realizing a fully automated delivery process from server racking to bare metal node readiness.

[0045] It should be understood that the fully automated bare-metal server management method in this embodiment of the invention can be implemented in any cloud computing management device based on the fully automated bare-metal server management method, such as: the controller of a cloud computing management platform, a data center automated operation and maintenance server, an intelligent computing center resource scheduling system, etc. Specifically, the fully automated bare-metal server management method is stored in the above-mentioned device as a pre-written program. When the device starts up or receives a management trigger command, the program is called, and the fully automated bare-metal server management method is implemented, thereby realizing fully automated management without human intervention from server racking to registration as a bare-metal node.

[0046] Compared to existing technologies, the fully automated bare-metal server management method provided in this invention, from the automatic acquisition of initial server configuration information to the automatic allocation of out-of-band management network addresses, and then to hardware verification, stress testing, data cleanup, node registration pre-configuration, node registration, switch port configuration, and batch installation verification, is automatically executed by status information without manual intervention. Compared to existing technologies, the management time for a single server to be managed is reduced from several hours to minutes, and the delivery cycle for thousands of servers is reduced from several weeks to several hours, greatly improving the rapid deployment capability of intelligent computing projects. At the same time, it provides reusable automation capabilities for large-scale, high-frequency bare-metal server delivery, exhibiting good engineering scalability.

[0047] Furthermore, in this embodiment of the invention, hardware configuration verification and hardware stability verification are performed before node registration, intercepting quality problems at the management stage, avoiding problems from occurring, and reducing failure rate and repair costs.

[0048] Furthermore, in the pre-registration configuration stage of the present invention, the network connection information of the server to be managed is automatically obtained, and the connected switch port is automatically configured according to the network connection information after the node is registered. This realizes an automated closed loop from obtaining the information of the server to be managed to the switch configuration, eliminating manual intervention on the network side and reducing the communication cost and configuration error risk of cross-team collaboration.

[0049] Meanwhile, the embodiments of the present invention support flexible combinations of operations such as verification, stress testing, cleanup, and registration to be performed on different servers to be managed through state-driven preset processing operations, which can adapt to the management needs of different customers and different scenarios.

[0050] In some embodiments of the present invention, step S101 includes: Initial configuration information is automatically obtained by reading the radio frequency identification (RFID) or near field communication (NFC) tags set on the server to be managed; or, By reading the power data of the intelligent power distribution unit (PDU) and the power consumption characteristics generated by the agent process, the identity and physical location are determined, and the expected hardware configuration information is obtained based on manual input.

[0051] Specifically, the process of automatically acquiring initial configuration information based on RFID tags is as follows: a smart asset management strip is installed on the rack side, with a reader for each U-position. An RFID tag with a unique ID is affixed to the mounting ear of the server to be managed, and the initial configuration and SN information of the server are written into the tag. When the server to be managed is inserted into the rack, the reader of the U-position senses the tag ID. Then, the initial configuration information of all servers to be managed can be obtained through the smart asset management strip.

[0052] The process of automatically obtaining initial configuration information based on NFC tags is as follows: a smart NFC asset strip is installed on the rack side, and each U-position can read the anti-metal NFC tag information through a multi-channel multiplexed antenna array; an anti-metal NFC tag containing the server's initial configuration and SN information is affixed to the mounting ear of the server to be managed; when the server to be managed is inserted into the rack, the reader of the U-position reads the anti-metal NFC tag; then the initial configuration information of all servers to be managed can be obtained through the smart NFC asset strip.

[0053] The process of determining the identity and physical location based on PDU and power consumption characteristics is as follows: First, an instruction is sent to a server to be managed via IPMI, controlling the agent process (IPA) running in the server's memory operating system to create a momentary high CPU load according to a preset timing pattern (e.g., 3 seconds of full load, 1 second of idle, 2 seconds of full load, etc.), thereby generating identifiable power consumption characteristics. Simultaneously, real-time power data from each port monitored by the intelligent PDU is continuously read. When a power fluctuation pattern perfectly matching the preset instruction is detected on a port corresponding to a particular PDU, that PDU is determined as the physical location of the server. Then, based on the server identity identifier (SN) used when the instruction was issued, the server identity identifier corresponding to that physical location can be synchronously determined. Through this process, the identity identifier and physical location can be obtained.

[0054] This invention provides two different methods for obtaining initial configuration information, ensuring that server identity, physical location, and expected hardware configuration information can be obtained efficiently and accurately under different deployment conditions, from fully automatic to semi-automatic, laying a reliable data foundation for subsequent automated management processes.

[0055] It should be noted that, in order to further improve the diversity of methods for obtaining initial configuration information and the reliability under extreme conditions, in some embodiments of the present invention, the initial configuration information can also be manually entered according to an Excel template.

[0056] In some embodiments of the present invention, step S102 includes: If the network planning strategy is static allocation, then a static out-of-band management network address (IPMI IP) is assigned to the server to be managed based on its physical location, and the out-of-band management network address is recorded as static. If the network planning strategy is dynamic allocation, then the out-of-band management network address is recorded as Dynamic Host Configuration Protocol (DHCP) type.

[0057] In a specific embodiment of the present invention, the network planning strategy is determined through a configuration parameter, namely auto_ipmi_ip. Specifically, as follows... Figure 2 As shown, when auto_ipmi_ip is true, IPMI IPs are assigned to the servers to be managed in ascending order of rack U position, and the allocation type of the IPMI IP is recorded as static. When auto_ipmi_ip is false, the allocation type of the IPMI IP is recorded as dhcp, and the IPMI IP and classification type are written to servers_info.

[0058] The embodiments of this invention can flexibly adapt to the planning requirements of different data centers for out-of-band management networks. Specifically, static allocation is used in scenarios requiring strict IP management, while dynamic allocation is used in scenarios with existing DHCP infrastructure, achieving unified support for two mainstream network management modes through a single automated process.

[0059] In some embodiments of the present invention, when the target operation is hardware configuration verification, hardware configuration verification needs to be performed, which specifically includes: Collect the actual hardware configuration information of the server to be managed, and compare the actual hardware configuration information with the expected hardware configuration information; Obtain log data from the server to be managed; When the comparison results are inconsistent, or the log data indicates that there is an abnormal hardware status, the fault handling mechanism is triggered. The fault handling mechanism includes generating email alerts or, by calling a command to illuminate the server's fault indicator light for physical location.

[0060] The process of obtaining actual hardware configuration information is as follows: IPA obtains actual hardware configuration information by using Python psutil and gputil, combined with commands such as lscpu, lsblk, and dmidecode.

[0061] In a specific embodiment of the present invention, the actual hardware configuration information is compared with the model and quantity in the expected hardware configuration information, then... Figure 3As shown, when the hardware configuration verification is true, IPA obtains the actual hardware configuration information and compares it with the expected hardware configuration information obtained from servers_info to determine if the model and quantity are the same. If they are different, an email alarm is generated and the indicator light is activated via ipmotool. If they are the same, the BMC log of the server to be managed is obtained via ipmotool, and it is determined whether there are any errors in the BMC log. If there are no errors, servers_info is updated to update its verification status to verified. If there are errors, an email alarm is generated and the indicator light is activated via ipmotool.

[0062] The BMC log is a history of system events recorded internally by the server's out-of-band management chip (BMC). It includes server startup records, hardware error events (such as memory ECC errors, CPU overheating, power supply failures, and fan stoppages), system restart reasons, and hardware change records. This log is independent of the server's operating system; it remains accessible even if the server is shut down or the operating system crashes.

[0063] It should be noted that, to further reduce the probability of failure, in some embodiments of the present invention, such as... Figure 3 As shown, while acquiring BMC logs, sensor data can also be acquired. Sensor data consists of physical quantity values ​​collected in real time by various sensors inside the server. The sensor data is used to determine whether there are any operational problems such as abnormal temperature or unstable voltage in the server to be managed. If so, it is determined that the server to be managed has potential stability issues, triggering an alarm and illuminating the location.

[0064] In some embodiments of the present invention, when the target operation is hardware stability verification, hardware stability verification needs to be performed, which specifically includes: Perform stress tests on the central processing unit, memory, storage media, and graphics processing unit of the managed server; Collect stress test results of all servers of the same model to be managed and compare them horizontally. If the test data of a certain server to be managed deviates from the average data of other servers of the same model to be managed by more than a threshold, the fault handling mechanism will be triggered. If a hardware error occurs during the stress test, the fault indicator light of the server to be managed will be lit by command call to physically locate the error.

[0065] Specifically, such as Figure 4As shown, when the hardware stability verification parameter is true, the IPA executes the load test command, performing load tests on the CPU and memory via sysbench, on the hard disk image via fio, and on the GPU via DCGM and gpu_burn. If a hardware error occurs during the load test, the bare metal management service generates an email alert and performs LED activation on the server to be managed via ipmitool. After the load test, ipmitool is used again to retrieve the BMC logs and sensor data of the server to be managed. If errors are found in the BMC logs or sensors, an alert is generated. Finally, the load test results are uploaded to the bare metal management service, which collects load test data from all servers of the same model. If a server's load test data differs significantly from other servers, an alert is also generated. Subsequent hardware repairs by operations and maintenance personnel or server vendor after-sales service are required. After a successful load test, the load test status in the servers_info table is updated to "load tested".

[0066] Among them, the bare metal management service is the core software service module deployed on the control end, which serves as the orchestration engine and status management center for the entire bare metal automated management process.

[0067] This invention, by re-acquiring and verifying BMC logs and sensor data after stress testing and combining this with real-time monitoring during the testing process, forms a dual verification mechanism of real-time monitoring during testing and log verification after testing. This mechanism can capture latent problems that, although not directly manifested as hardware errors during testing, have been recorded by the BMC, such as transient temperature exceeding thresholds, brief voltage fluctuations, fan response lag, or accumulated memory ECC errors. This effectively avoids the failure to detect faults due to the neglect of intermittent failures or critical states. Simultaneously, by collecting stress test data from all servers of the same model awaiting management and conducting a horizontal comparison, statistical methods are used to identify abnormal individuals whose stress test indicators (such as CPU performance, memory bandwidth, GPU computing power, or disk IOPS) significantly deviate from the group average. This allows for the precise identification of servers awaiting management that are correctly configured, have no hardware errors, but exhibit abnormalities, and their interception before delivery.

[0068] In other words, the embodiments of the present invention, through the above two fault detection mechanisms, jointly elevate hardware quality acceptance from a binary judgment of whether a fault exists to a multi-dimensional assessment of whether a group consistency level has been reached, significantly enhancing the depth and breadth of hardware stability detection, and truly realizing pre-emptive quality assurance in large-scale bare metal delivery scenarios.

[0069] In some embodiments of the present invention, when the target operation is data cleaning, the target operation is executed, including: Perform light cleaning (metadata cleaning) on ​​the storage media of the managed server, or deep cleaning by overwriting multiple random data entries.

[0070] In specific embodiments of the present invention, such as Figure 5 As shown, when data cleanup is true, IPA performs the cleanup action, clearing data from all hard drives on the servers to be managed. It supports light cleanup and deep cleanup. Light cleanup is faster, only cleaning up the hard drive's metadata information. Deep cleanup is slower, using the dd command to write random data to the hard drive multiple times to completely delete the original data. After successful cleanup, the cleanup status in the servers_info table is updated to "cleaned".

[0071] In some embodiments of the present invention, when the target operation is a node registration pre-configuration, the target operation is executed, including: Obtain the network connection information of the server to be managed, including the network card address and the port information of the switch it is connected to; Configure the out-of-band management system for the server to be managed. The configuration of the out-of-band management system includes setting the out-of-band management account password and setting the out-of-band management network address of the server to be managed according to the type of out-of-band management network address. Configure the Basic Input Output System (BIOS) of the server to be managed. The BIOS configuration includes at least one of the following: virtualization switch, hyper-threading switch, power performance mode, boot mode, serial port redirection, and system time.

[0072] In a specific embodiment of the present invention, the specific processes of node registration pre-configuration and configuration-based registration are as follows: Figure 6As shown, if the configuration parameter before node registration is true, the IPA process first calls the ipmitool command to set the management account and password for the server to be managed. The password can be a unified password or a random password, depending on the bare metal management service's ipmi_password configuration. After setting the password, the server's ipmi account password field in the servers_info table is updated. Then, IPA sets the ipmi IP for the server to be managed according to the configured ipmi IP allocation type attribute: if the ipmi IP allocation type is static, it retrieves the ipmi IP, subnet mask, and gateway information pre-allocated by the bare metal management service from the servers_info table and sets it using the ipmitool command; if the ipmi IP allocation type is dhcp, it retrieves the currently used ipmi IP, subnet mask, and gateway information using the ipmitool command and updates it to the servers_info table. Afterward, the bare metal management service writes the MAC address and IP address information of the ipmi network card into the dhcp configuration file to ensure that the same MAC address is always assigned to the same IP address and increases the dhcp IP lease time. Next, the IPA will configure the server's BIOS via IPMI or Redfish protocols. BIOS configurations can be configured through the bare metal management service to further improve scalability. The default settings are as follows: 1. Enable virtualization; 2. Enable hyper-threading; 3. Enable power performance mode and disable power saving; 4. Set UEFI boot mode and disable Secure Boot; 5. Enable IPMI over LAN; 6. Enable serial on LAN and set the serial port speed to 115200; 7. Set the BIOS time to the current UTC time; 8. Set the NUMA group to the number of CPU physical cores. Then, the IPA uses the lldpctl command to obtain the MAC addresses of all connected network cards on the current server and the corresponding switch port information. The IPA then registers the current server as an OpenStack bare metal node with the OpenStack Ironic bare metal service using the server configuration, IPMI IP, account password, switch port, and network card MAC address as parameters. After successful registration, the IPA updates the registration status in the servers_info table to "registered" and shuts down the server to be managed.

[0073] In some embodiments of the present invention, step S106 specifically includes: Determine the switch port to be configured based on the switch port information, and verify the correspondence between the switch port and the server to be managed based on the network card address. Configure the switch port as an aggregated independent port mode so that each member port of the switch port can be used independently for network startup during the operating system installation phase, and after the operating system installation is completed, the switch port can be used by the server to be managed through link aggregation.

[0074] Specifically, during the bare metal PXE installation process, switch ports cannot be configured for aggregation. However, after installation, the system needs to configure bond aggregation to ensure high network availability. By setting the switch ports to independent port mode, switch configuration can be eliminated. During the bare metal PXE installation process, each port in the aggregation port can be used as an independent switch port, and the PXE network card can obtain a DHCP IP and further execute PXE actions. After installation, cloud-init automatically sets bond aggregation within the bare metal node, and no changes are required on the switch side.

[0075] In some embodiments of the present invention, step S107 includes: Call the OpenStack Nova application programming interface to perform batch installation operations on registered bare metal nodes; Perform network connectivity testing on bare metal nodes after installation. If the nodes are connected within the preset timeout period, the verification is confirmed as successful; otherwise, an alarm is generated.

[0076] It should be understood that the timeout period can be set or adjusted according to the actual application scenario. For example, the timeout period can be set to 1 hour.

[0077] It should be noted that servers to be managed may contain residual operating systems, configuration files, test data, or even sensitive information on their hard drives before being racked. If these are not cleaned up before delivery to users, it will pose a serious risk of data leakage. Furthermore, residual partition tables or boot records may cause subsequent installation failures. To address this technical issue, after successful verification, a cleanup process can be performed on normal bare metal nodes to remove data.

[0078] In summary, such as Figure 7As shown in the embodiments of this invention, the fully automated bare-metal server management method achieves initial configuration information acquisition and input, out-of-band management network address configuration, switch port configuration, and batch installation verification through the bare-metal management service, and verifies through IPA (In-Process Automation). Specifically, the bare-metal management service acts as the control center, responsible for global process orchestration, status management, asset information collection, out-of-band network planning, interaction with OpenStack and switches, and final installation verification; the IPA, as an execution agent running in the server's memory operating system, is responsible for performing specific operations locally on the server, such as hardware information collection, configuration verification, stress testing, data cleanup, out-of-band management and BIOS configuration, network information acquisition, and node registration. The two are decoupled and collaborate through stored status information, jointly constituting a fully automated, unattended management capability from physical server mounting to bare-metal node readiness on the cloud platform.

[0079] In summary, the fully automated bare-metal server management method proposed in this invention integrates all aspects of the bare-metal server process, from physical racking to cloud platform management, including automatic asset information acquisition, automatic out-of-band network planning, hardware configuration verification, stress testing, data cleaning, out-of-band management and automatic BIOS configuration, automatic switch port configuration, and batch installation verification, into a unified, programmable automated closed loop by constructing a state-driven, fully automated pipeline. This completely eliminates the drawbacks of traditional methods, such as reliance on multiple roles for manual intervention, fragmented processes, and lack of quality assurance. It achieves large-scale delivery capabilities, significantly improves delivery efficiency, and effectively reduces hardware failure rates and rework costs through pre-emptive quality interception and horizontal comparison mechanisms. Furthermore, it achieves a closed loop for online anomaly and physical location through fault indicator linkage, providing a systematic technical solution for the rapid, reliable, and secure delivery of large-scale bare-metal servers in intelligent computing centers.

[0080] On the other hand, embodiments of the present invention also provide a fully automated management system for bare metal servers, such as... Figure 8 As shown, the fully automated management system 800 for bare metal servers includes: The information acquisition module 801 is used to automatically acquire the initial configuration information of the server to be managed. The initial configuration information includes the identity identifier, physical location and expected hardware configuration information. Network planning module 802 is used to automatically assign out-of-band management network addresses to servers to be managed according to preset network planning strategies; The startup control module 803 is used to control the server to be managed to start up via the network and load the operating system in memory. The agent process module 804 runs in the memory operating system and is used to obtain the status information of the server to be managed based on the identity identifier, and to determine the target operation to be executed from the preset processing operations based on the status information and execute it; the preset processing operations include at least one of hardware configuration verification, hardware stability verification, data cleaning, and node registration pre-configuration, and the node registration pre-configuration includes obtaining the network connection information of the server to be managed. The node registration module 805 is used to initiate a node registration request to the bare metal service of the cloud platform by using the identity identifier, out-of-band management network address and network connection information as registration parameters after the target operation is completed and the status information indicates that node registration is required, and register the server to be managed as a bare metal node. The switch configuration module 806 is used to automatically configure the switch port connected to the server to be managed based on the network connection information after the server to be managed has been successfully registered as a bare metal node. The installation verification module 807 is used to perform batch installation verification on registered bare metal nodes.

[0081] The fully automated management system 800 for bare metal servers provided in the above embodiments can realize the technical solutions described in the embodiments of the fully automated management method for bare metal servers. The specific implementation principles of each module or unit can be found in the corresponding content in the embodiments of the fully automated management method for bare metal servers, and will not be repeated here.

[0082] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.), and the computer program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0083] The above provides a detailed description of the fully automated management method and system for bare metal servers provided by this invention. Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A fully automated management method for bare metal servers, characterized in that, include: Automatically obtain the initial configuration information of the server to be managed, including its identity, physical location, and expected hardware configuration information; Automatically assign out-of-band management network addresses to the servers to be managed according to the preset network planning strategy; The server to be managed is controlled to start up and load a memory operating system via the network. The agent process running in the memory operating system obtains the status information of the server to be managed based on the identity identifier. The status information is used to determine the preset processing operation to be executed. The agent process determines the target operation to be executed from the preset processing operations and executes the target operation; the preset processing operations include at least one of hardware configuration verification, hardware stability verification, data cleanup, and node registration pre-configuration, and the node registration pre-configuration includes obtaining the network connection information of the server to be managed; After the target operation is completed, if the status information indicates that node registration is required, the identity identifier, the out-of-band management network address and the network connection information are used as registration parameters to initiate a node registration request to the cloud platform bare metal service, and the server to be managed is registered as a bare metal node. After the server to be managed is successfully registered as a bare metal node, the switch port connected to the server to be managed is automatically configured according to the network connection information. Perform batch installation verification on registered bare metal nodes.

2. The fully automated management method for bare metal servers according to claim 1, characterized in that, The automatic acquisition of the initial configuration information of the server to be managed includes: The initial configuration information is automatically obtained by reading the radio frequency identification tag or near field communication tag set on the server to be managed; or, By reading the power data of the intelligent power distribution unit and the power consumption characteristics generated by the agent process, the identity and physical location are determined, and the expected hardware configuration information is obtained based on manual input.

3. The fully automated management method for bare metal servers according to claim 1, characterized in that, The step of automatically assigning out-of-band management network addresses to the servers to be managed according to a preset network planning strategy includes: If the network planning strategy is static allocation, then a static out-of-band management network address is allocated to the server to be managed according to the physical location, and the out-of-band management network address is recorded as static. If the network planning strategy is dynamic allocation, then the out-of-band management network address is recorded as a Dynamic Host Configuration Protocol (DHCP) type.

4. The fully automated management method for bare metal servers according to claim 1, characterized in that, When the target operation is hardware configuration verification, the target operation is executed, including: Collect the actual hardware configuration information of the server to be managed, and compare the actual hardware configuration information with the expected hardware configuration information; Obtain the log data of the server to be managed; When the comparison results are inconsistent, or the log data indicates that there is an abnormal hardware status, the fault handling mechanism is triggered. The fault handling mechanism includes generating email alerts or, by calling a command to illuminate the fault indicator light on the server for physical location.

5. The fully automated management method for bare metal servers according to claim 1, characterized in that, When the target operation is hardware stability verification, the target operation is executed, including: Perform stress tests on the central processing unit, memory, storage media, and graphics processor of the server to be managed; Collect stress test results of all servers of the same model to be managed and compare them horizontally. If the test data of a certain server to be managed deviates from the average data of other servers of the same model to be managed by more than a threshold, the fault handling mechanism will be triggered. If a hardware error occurs during the stress test, the fault indicator light of the server to be managed will be lit by command call to physically locate the error.

6. The fully automated management method for bare metal servers according to claim 1, characterized in that, When the target operation is data cleaning, the target operation is executed, including: A light cleanup of the storage medium of the server to be managed is performed to clean up the metadata, or a deep cleanup is performed by overwriting multiple random data entries.

7. The fully automated management method for bare metal servers according to claim 3, characterized in that, When the target operation is a node registration pre-configuration, the target operation is executed, including: Obtain the network connection information of the server to be managed, including the network card address and the connected switch port information; Configure the out-of-band management system of the server to be managed. The configuration of the out-of-band management system includes setting the out-of-band management account password and setting the out-of-band management network address of the server to be managed according to the type of the out-of-band management network address. Configure the basic input / output system of the server to be managed. The configuration of the basic input / output system includes at least one of the following: virtualization switch, hyper-threading switch, power performance mode, startup mode, serial port redirection, and system time.

8. The fully automated management method for bare metal servers according to claim 7, characterized in that, The automatic configuration of the switch port connected to the server to be managed based on the network connection information includes: The switch port to be configured is determined based on the switch port information, and the correspondence between the switch port and the server to be managed is verified based on the network card address. Configure the switch port as an aggregated independent port mode so that each member port of the switch port can be used independently for network startup during the operating system installation phase, and after the operating system installation is completed, the switch port can be used by the server to be managed through link aggregation.

9. The fully automated management method for bare metal servers according to claim 1, characterized in that, The step of performing batch installation verification on registered bare metal nodes includes: Call the OpenStack Nova application programming interface to perform batch installation operations on registered bare metal nodes; Perform network connectivity testing on bare metal nodes after installation. If the nodes are connected within the preset timeout period, the verification is confirmed as successful; otherwise, an alarm is generated.

10. A fully automated management system for bare metal servers, characterized in that, include: The information acquisition module is used to automatically acquire the initial configuration information of the server to be managed, including the identity identifier, physical location and expected hardware configuration information. The network planning module is used to automatically allocate out-of-band management network addresses to the servers to be managed according to a preset network planning strategy. The startup control module is used to control the server to be managed to start up via the network and load the memory operating system; The agent process module runs in the memory operating system and is used to obtain the status information of the server to be managed based on the identity identifier, and to determine and execute the target operation to be executed from the preset processing operations based on the status information; the preset processing operations include at least one of hardware configuration verification, hardware stability verification, data cleaning, and node registration pre-configuration, and the node registration pre-configuration includes obtaining the network connection information of the server to be managed. The node registration module is used to initiate a node registration request to the cloud platform bare metal service by using the identity identifier, the out-of-band management network address and the network connection information as registration parameters after the target operation is completed and the status information indicates that node registration is required, and register the server to be managed as a bare metal node. The switch configuration module is used to automatically configure the switch port connected to the server to be managed according to the network connection information after the server to be managed is successfully registered as a bare metal node. The installation verification module is used to perform batch installation verification on registered bare metal nodes.

Images