Switching control method, device and equipment for cloud platform management nodes

Abstract

The embodiment of the invention provides a switching control method, device and equipment for a cloud platform management node. The method comprises the following steps: acquiring state data of a cloud platform management main node; when the state data is a fault, selecting a new cloud platform management main node from at least two groups of cloud platform management backup nodes; the virtual IP address of the original cloud platform management main node is bound with the new cloud platform management main node, and routing redirection is completed; controlling the task of the original cloud platform management main node to be switched to the new cloud platform management main node for processing; and controlling the cloud platform management main node and the cloud platform management backup node to keep data synchronization. According to the embodiment of the invention, the new master node is selected from the at least two groups of backup nodes, the selected new master node is ensured to be in a normal state and have enough resources, and new problems after switching are avoided.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of cloud computing technology, and in particular to a switching control method, device, and equipment for a cloud platform management node. Background Art

[0002] To effectively manage cloud platform computing resources, a unified management node is typically deployed within the cloud platform to handle tasks such as task scheduling, resource allocation, virtual machine status monitoring, and configuration management. Current cloud platform master management nodes are typically deployed in a single location or manually switch between master and backup nodes. If the master node fails, the cloud platform cannot promptly complete core functions such as task scheduling and virtual machine control. Furthermore, the status of backup nodes is not monitored, leading to the risk of failures or insufficient resources after a switchover. Summary of the Invention

[0003] The technical problem to be solved by the embodiments of the present invention is to provide a switching control method, device and equipment for a cloud platform management node, select a new master node from at least two groups of backup nodes, ensure that the selected new master node is in a normal state and has sufficient resources, and avoid new problems after switching.

[0004] To solve the above technical problems, the technical solutions of the embodiments of the present invention are as follows: A cloud platform management node switching control method, comprising: Obtain the status data of the cloud platform management master node; When the status data indicates a failure, a new cloud platform management master node is selected from at least two groups of cloud platform management backup nodes; Bind the virtual IP address of the original cloud platform management master node to the new cloud platform management master node and complete the routing redirection; Control the tasks of the original cloud platform management master node to be switched to the new cloud platform management master node for processing; and control the cloud platform management master node and the cloud platform management backup node to maintain data synchronization.

[0005] Optionally, the status data is a response signal of the cloud platform management master node to a detection signal periodically sent by the cloud platform management backup node.

[0006] Optionally, when the status data indicates a failure, selecting a new cloud platform management master node from at least two groups of cloud platform management backup nodes includes: When the status data is unresponsive for multiple consecutive times, it is determined that the cloud platform management master node is faulty; Control all cloud platform management backup nodes to enter the election state and start the master node election process; After the election is successful, the winning cloud platform management backup node becomes the new cloud platform management master node.

[0007] Optionally, all cloud platform management backup nodes are controlled to enter a campaign state, and a master node election process is started, including: All cloud platform management backup nodes are controlled to suspend providing data processing services to the outside, and only communication functions related to the election are reserved; Each cloud platform management backup node is subjected to qualification verification; According to Z = p 1 × A + p 2 × B + p 3 × C + p 4 × D + p 5 × E , the priority score of each cloud platform management backup node is determined; Wherein, Z is the priority score of the cloud platform management backup node, A is a preset basic level, B is a hardware resource health degree, C is a historical takeover success rate, D is a data synchronization delay, E is network connection stability, p 1 is a first weight, p 2 is a second weight, p 3 is a third weight, p 4 is a fourth weight, p 1 is a fifth weight; The cloud platform management backup node with the highest priority score is determined as the new cloud platform management master node.

[0008] Optionally, the determination of the hardware resource health degree, the historical takeover success rate, the data synchronization delay, and the network connection stability includes: According to B = Y- ( p 6 × B 1 +p 7 × B 2+p 8 × B 3 determining the hardware resource health degree of each cloud platform management backup node; wherein, B is the hardware resource health degree of the cloud platform management backup node, Y is the highest value of the hardware resource health degree, B 1 is the CPU usage rate, B 2 is the memory occupancy rate, B 3 is the disk IO load rate, p 6 is the sixth weight, p 7 is the seventh weight, p 8 is the eighth weight; According to C = C 1 / C 2 determining the historical takeover success rate of each cloud platform management backup node; wherein, C is the historical takeover success rate of the cloud platform management backup node, C 1 is the number of successful takeovers, C 2 is the total number of takeover attempts; According to D=1÷ ( 1+D 1 determining the data synchronization delay of each cloud platform management backup node; wherein, D is the data synchronization delay of the cloud platform management backup node, D 1 is the average delay in seconds; According to E= ( p 9 × E 1 +p 10 × E 2 determining the network connection stability of each cloud platform management backup node; wherein, E is the network connection stability of the cloud platform management backup node, E 1 is the network connectivity rate, E 2 is the packet loss rate, p9 is a ninth weight, p 10 is a tenth weight.

[0009] Optionally, the virtual IP address of the original cloud platform management master node is bound to the new cloud platform management master node, and routing redirection is completed, including: unbinding the original cloud platform management master node from the virtual IP address; binding the virtual IP address to a network interface of the new cloud platform management master node; sending an address mapping update notification to all devices in the network through an address resolution protocol (ARP) broadcast to complete routing redirection.

[0010] Optionally, the task of the original cloud platform management master node is switched to the new cloud platform management master node for processing, including: controlling the new cloud platform management master node to obtain the task synchronized from the original cloud platform management master node; controlling the new cloud platform management master node to process the task of the original cloud platform management master node.

[0011] Embodiments of the application also provide a switching control device for a cloud platform management node, including: an obtaining module configured to obtain state data of a cloud platform management master node; a processing module configured to, when the state data is a fault, select a new cloud platform management master node from at least two groups of cloud platform management backup nodes, bind a virtual IP address of an original cloud platform management master node to the new cloud platform management master node, and complete routing redirection, control the task of the original cloud platform management master node to be switched to the new cloud platform management master node for processing, and control the cloud platform management master node and the cloud platform management backup node to keep data synchronization.

[0012] Embodiments of the application also provide a computing device, including: one or more processors; a storage device configured to store one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the method described above.

[0013] Embodiments of the application also provide a computing device readable storage medium, the computing device readable storage medium stores a program, and the program is executed by a processor to implement the method described above.

[0014] The above-mentioned scheme of the embodiments of the application has at least the following beneficial effects: The scheme of the embodiment of the application, by automatically detecting the state of the master node, automatically selects a new master node from the backup nodes when the master node fails, without manual intervention, greatly shortens the fault recovery time, and ensures the continuous operation of the core function of the cloud platform.

[0015] By binding and routing redirection of the virtual IP address, the new master node seamlessly takes over the network identity of the original master node, and external requests can be automatically routed to the new master node without modifying the configuration. At the same time, the new master node directly handles the tasks of the original master node, and in combination with the data synchronization mechanism of the master and backup nodes, ensures that the tasks are not lost and the business is not interrupted.

[0016] Selecting a new master node from at least two groups of backup nodes ensures that the selected new master node is in a normal state and has sufficient resources, avoiding new problems after switching and improving switching reliability. The present application performs qualification verification on each cloud platform management backup node through a determination mode of hardware resource health degree, historical takeover success rate, data synchronization delay and network connection stability, greatly improving the accuracy and normal operation efficiency of hardware operation.

[0017] Through the master and backup node data synchronization mechanism, the data of the backup node is updated in real time, avoiding data differences. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 is a flow chart of the switching control method of the cloud platform management node of the embodiment of the application.

[0019] Figure 2 is a component schematic diagram of the switching control system of the cloud platform management node of the embodiment of the application.

[0020] Figure 3 is a specific structure schematic diagram of the switching control system of the cloud platform management node of the embodiment of the application.

[0021] Figure 4 is a switching control flow chart of the cloud platform management node of the embodiment of the application.

[0022] Figure 5 is a whole structure schematic diagram of the switching control system of the cloud platform management node of the embodiment of the application.

[0023] Figure 6 is a module schematic diagram of the switching control device of the cloud platform management node of the embodiment of the application. DETAILED DESCRIPTION

[0024] Exemplary embodiments of the present application will be described herein below with reference to the accompanying drawings. While exemplary embodiments of the present application are shown in the drawings, it is understood that the present application can be embodied in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the application to those skilled in the art.

[0025] As shown in Figure 1 An embodiment of the present application provides a switching control method of a cloud platform management node, comprising: Step 11, obtaining state data of a cloud platform management master node; Step 12, when the state data is a fault, selecting a new cloud platform management master node from at least two groups of cloud platform management backup nodes; Step 13, binding a virtual IP address of the original cloud platform management master node with the new cloud platform management master node, and completing route redirection; specifically, the virtual IP address is an Internet Protocol address; Step 14, controlling task switching of the original cloud platform management master node to the new cloud platform management master node for processing; and controlling the cloud platform management master node and the cloud platform management backup node to keep data synchronization.

[0026] In this embodiment, by automatically detecting the state of the master node, a new master node is automatically selected from the backup nodes when the master node fails, without manual intervention, greatly shortening the fault recovery time and ensuring the continuous operation of the core function of the cloud platform.

[0027] Through virtual IP address binding and route redirection, the new master node seamlessly takes over the network identity of the original master node, and external requests can be automatically routed to the new master node without modifying the configuration. At the same time, the new master node directly processes the tasks of the original master node, and in combination with the data synchronization mechanism of the master and backup nodes, it ensures that the tasks are not lost and the business is not interrupted.

[0028] The new master node is selected from at least two groups of backup nodes to ensure that the selected new master node is in a normal state and has sufficient resources, avoiding new problems after switching and improving the switching reliability.

[0029] Through the master and backup node data synchronization mechanism, the backup node ensures real-time updating of the master node data, avoiding data differences.

[0030] In an optional embodiment of the present application, in step 11, the state data is a response signal of the cloud platform management master node to a detection signal periodically sent by the cloud platform management backup node.

[0031] Specifically, the cloud platform management backup node generates a standardized detection signal according to a preset period, the detection signal contains a time stamp, a backup node identifier, a random check code and a serial number of this detection, and the detection signal is sent to the cloud platform management master node through a special communication link inside the cluster. After the cloud platform management master node receives the detection signal, the validity of the random check code is verified immediately, and a response signal is generated within a fixed time limit; the response signal contains a master node identifier, confirmation information corresponding to the current detection serial number, the running state of the master node current core process and the key resource occupation rate snapshot, and then the response signal is returned to the cloud platform management backup node that sent the detection signal; the system obtains the response signal.

[0032] In this embodiment, the detection signal contains a random check code, which can effectively prevent the signal from being tampered or forged during transmission, and ensure that the master node receives a real and valid detection request.

[0033] The setting of the fixed response time limit (such as within 100ms) ensures that the response state of the master node can be accurately timed, and provides a clear basis for continuous non-response failure determination.

[0034] In an optional embodiment of the application, when the state data is a fault in step 12, a new cloud platform management master node is selected from at least two groups of cloud platform management backup nodes, comprising: Step 121, when the state data is continuous multiple (such as 3 times) non-response (or continuous multiple key resource occupation rate exceeds a preset threshold), it is determined that the cloud platform management master node is a fault; Step 122, control all cloud platform management backup nodes to enter the election state and start the master node election process; Step 123, after selection, the winning cloud platform management backup node becomes the new cloud platform management master node.

[0035] In this embodiment, continuous multiple non-responses or continuous multiple key resource overruns are used as fault determination conditions, rather than single abnormality triggering switching, which effectively filters non-fatal problems such as network transient jitter and temporary resource peaks, reduces cluster turbulence caused by misjudgment, and ensures the stability of system operation.

[0036] All backup nodes are required to enter the election state and start the standardized process, avoiding conflicts caused by multiple nodes occupying the master node role in disorder.

[0037] Even if some backup nodes are abnormal, the best one can still be selected from the remaining nodes, further improving the risk resistance of the cluster.

[0038] In an optional embodiment of the present application, in step 122, all cloud platform management backup nodes are controlled to enter a candidate state, and a master node election process is started, including: Step 1221, control all cloud platform management backup nodes to suspend providing data processing services to the outside, and only keep the communication function related to the election; Step 1222, perform qualification verification on each cloud platform management backup node; Step 1223, according to Z = p 1 × A + p 2 × B + p 3 × C + p 4 × D + p 5 × E , determine the priority score of each cloud platform management backup node; wherein, Z is the priority score of the cloud platform management backup node, A is a preset basic level, B is a hardware resource health degree, C is a historical takeover success rate, D is a data synchronization delay, E is a network connection stability, p 1 is a first weight, p 2 is a second weight, p 3 is a third weight, p 4 is a fourth weight, p 1 is a fifth weight; The cloud platform management backup node with the highest priority score is determined as the new cloud platform management master node.

[0039] In this embodiment, by suspending the data processing service to the outside and keeping the election communication function, all backup nodes are temporarily separated from the regular business load, avoiding resource contention (such as too high CPU / memory occupation affecting the transmission of election signals) caused by parallel processing tasks, ensuring that the nodes concentrate computing power on information interaction and calculation related to the election, and reducing the risk of process interruption.

[0040] The qualification verification mechanism (such as verifying data synchronization integrity, hardware health, network connectivity, etc.) can eliminate backup nodes that do not meet the operating requirements of the primary node in advance (for example, nodes with excessive data synchronization lags or hardware failure risks), ensuring the quality of the candidate pool from the source.

[0041] Priority calculation formula ( Z = p 1 × A + p 2 × B + p 3 × C + p 4 × D + p 5 × E ) quantifies and integrates key factors influencing masternode competence (pre-set baseline level, hardware health, historical performance, data synchronization quality, and network stability), assigning weights to reflect the importance of each factor. This approach avoids bias caused by reliance on a single metric (such as hardware performance alone) or subjective judgment, making prioritization more objective and interpretable.

[0042] The rule that the highest-scoring node is elected simplifies the final decision-making process and avoids lengthy selection processes caused by complex negotiations. Combined with pre-qualification screening, all participating nodes are qualified candidates, further ensuring the highest-scoring node's comprehensive advantages in hardware, data, and network. This allows it to quickly take over master node functions, shortening the time between the original master node failure and the new master node coming online, and minimizing the impact of business interruptions.

[0043] In an optional embodiment of the present invention, in step 1223, the method for determining the hardware resource health, historical takeover success rate, data synchronization delay, and network connection stability includes: Step 12231, according to B = Y- ( p 6 × B 1 +p 7 × B 2 +p 8 × B 3 ), determine the health of hardware resources of each cloud platform management backup node; in, B Manage the hardware resource health of backup nodes for the cloud platform,Y is the highest value of the hardware resource health, B 1 is the CPU usage rate, B 2 is the memory occupancy rate, B 3 is the disk IO load rate, p 6 is the sixth weight, p 7 is the seventh weight, p 8 is the eighth weight; Step 12232, according to C = C 1 / C 2 , determine the historical takeover success rate of each cloud platform management backup node; the historical takeover success rate of the cloud platform management backup node participating in the election for the first time C is the preset success rate C 3 ; wherein, C is the historical takeover success rate of the cloud platform management backup node, C 1 is the number of successful takeovers, C 2 is the total number of takeover attempts; Step 12233, according to D=1÷ ( 1+D 1 ), determine the data synchronization delay of each cloud platform management backup node; wherein, D is the data synchronization delay of the cloud platform management backup node, D 1 is the average delay in seconds; Step 12234, according to E= ( p 9 × E 1 +p 10 × E 2 ), determine the network connection stability of each cloud platform management backup node; wherein, E is the network connection stability of the cloud platform management backup node, E 1 is the network connectivity rate, E 2 is the packet loss rate, p 9 is the ninth weight, p10 is the tenth weight.

[0044] In the embodiment, the hardware resource health degree is calculated by the maximum value minus the resource occupation weighted sum, which accurately reflects the node redundancy capability, eliminates the influence of hardware difference, and ensures the fairness of evaluation.

[0045] The historical takeover success rate is calculated as a preset value for a new node, which solves the problem of no historical data, balances experience and fairness, and enhances the inclusiveness of the mechanism.

[0046] The data synchronization delay is converted by the reciprocal, amplifies the advantage of low delay node, quickly identifies the data risk node, and ensures the service continuity.

[0047] The network connection stability formula integrates the connectivity rate and the packet loss rate, comprehensively evaluates the communication quality, and reduces the scheduling delay and data loss.

[0048] In an optional embodiment of the application, in the step 13, the virtual IP address of the original cloud platform management master node is bound to the new cloud platform management master node, and routing redirection is completed, comprising: Step 131, the binding relationship between the original cloud platform management master node and the virtual IP address is released; Step 132, the virtual IP address is bound to the network interface of the new cloud platform management master node; Step 133, address mapping update notification is sent to all devices in the network through address resolution protocol (ARP) broadcast, and routing redirection is completed.

[0049] In the embodiment, step 131 actively releases the binding between the original master node and the virtual IP, avoiding address conflict caused by the original node still occupying the virtual IP after failure; and step 132 binds the virtual IP to the new master node, ensuring the unique pointing property of network identification.

[0050] Step 133 forcibly pushes address mapping update through ARP broadcast, directly notifying all devices (such as switches, routers and other nodes) in the network.

[0051] Unbinding, binding and broadcast updating form a complete virtual IP migration closed loop, which fundamentally avoids address conflict, routing confusion and other problems, and ensures the accuracy and stability of network communication after the new master node takes over.

[0052] In an optional embodiment of the application, in the step 14, the task switching of the original cloud platform management master node is controlled to the new cloud platform management master node for processing, comprising: Step 141, controlling the new cloud platform management master node to obtain the task synchronized from the original cloud platform management master node; Step 142, control the new cloud platform management master node to process the tasks of the original cloud platform management master node.

[0053] Specifically, the new cloud platform management master node receives the synchronization data sent by the original master node through data synchronization, which contains task execution records, virtual machine running status and configuration information, resource usage, and other key contents. Data synchronization uses a multicast communication mechanism to ensure that the new master node can efficiently and reliably obtain the above-mentioned task-related data, ensuring data integrity and consistency. The new master node parses and verifies the synchronization data obtained, confirms that the data has not been damaged or lost, and provides a reliable data foundation for subsequent task processing. The new master node judges the task state synchronized from the original master node, determines whether the task is in execution, has not started, or has been interrupted, etc. If there are unfinished tasks, the new master node continues to execute these tasks according to the task information record, according to the priority and scheduling logic of the task; for example, for the virtual machine migration task being performed, the new master node will take over the migration process, complete the subsequent operations such as target virtual machine routing and resource allocation, key data and network information backup, etc., to ensure the smooth completion of the task. If an exception is found in the task, such as task execution failure due to original master node failure, the new master node triggers a retry mechanism according to the preset strategy to re-schedule resources to execute the task; at the same time, it avoids repeated execution of the task to ensure the rational use of system resources. When detecting a key virtual machine related task, the new master node will prioritize processing, and if the key virtual machine is in an abnormal state (such as down, resource depletion), it will immediately trigger a migration task through the task scheduling module and execute the migration operation by the migration module to ensure the continuous operation of the key business. During the processing of the task, the new master node monitors the host resource situation in real time, allocates resources reasonably, and avoids resource conflicts and waste.

[0054] In this embodiment, relying on the data synchronization mechanism, the new master node efficiently obtains complete task data, which is verified to be reliable, avoiding task loss or state confusion, and ensuring business continuity.

[0055] The new master node judges the task state, continues the unfinished task according to the priority, triggers the retry for the abnormal task, avoids repeated execution, and realizes seamless takeover.

[0056] Prioritize processing of key virtual machine tasks, quickly trigger migration in case of exception, ensure the stability of core business, and improve the risk resistance ability of high-priority business.

[0057] Real-time monitoring of host resources, reasonable allocation to avoid conflicts, combined with unified control module to ensure single node activation, optimize resource utilization and system stability.

[0058] In the above embodiments of the application, automatic master-backup switching can be completed when the master node fails, ensuring continuous operation of critical functions. The virtual IP fault switching component diagram is shown in FIG. 1. Figure 2 、 3 The software module structure mainly includes the following functional modules: Virtual IP fault switching manager: This module is responsible for the selection of the host and the automatic binding of the virtual IP. After the host is started, the master node is selected through the master-backup competition mechanism, and the virtual IP is bound to the network interface. Then the unified control module is notified to activate the current host to take over the following functions: data collection, task scheduling, virtual machine drift, etc. Through the binding switching of the virtual IP, the change of the master node is realized without awareness of the business access.

[0059] Unified control module: In a high-availability cluster environment, to avoid multiple nodes simultaneously performing data collection, task execution, etc., resulting in resource waste, repeated consumption or system pressure, this module is responsible for managing the active node state in the cluster. Its core function is to ensure that only one master node is in the active state at the same time, and the rest are hot standby nodes, realizing single scheduling of tasks and data consistency in a high-availability environment.

[0060] Host management module: This module is responsible for managing all host resources in the cloud platform. The main functions include: setting host access parameters (IP, port, authorization code); realizing host registration, query and state monitoring.

[0061] After the host information is configured, the system can collect the virtual machine running state on the cloud server password machine management interface specification, realizing standardized and unified resource access and management.

[0062] Data collection module: This module is responsible for collecting the following virtualization resource information from each host: Virtual machine basic information (such as name, UUID); virtual machine running state (running, shutdown, suspended); resource usage (CPU usage, memory usage, etc.). The collected data will serve as the basis for task scheduling, assisting in judging virtual machine availability and resource health.

[0063] Task scheduling module: Based on the information provided by the data collection module, logical analysis is performed to determine whether the virtual machine is in an abnormal state (such as downtime, resource depletion). Once an exception is found, a migration task is triggered and submitted to the migration module for execution, ensuring uninterrupted business high availability.

[0064] Migration Module: Responsible for executing virtual machine migration operations. Core functions include: migration task logging (recording the execution status of the migration task and the target host), target virtual machine routing and resource allocation, key data and network information backup, resource recovery and reconstruction during virtual machine migration, and factory reset of the source virtual machine to avoid resource conflicts and duplicate operations. This module ensures the rapid and accurate restoration of the virtual machine operating environment on the target host, enabling seamless business migration.

[0065] Data Synchronization Module: This module uses a multicast communication mechanism to synchronize data between the platform's active and standby nodes. The primary synchronization content includes task execution records, virtual machine operating status and configuration information, and resource usage. This module ensures that key data on the active and standby nodes remains consistent, providing reliable data support for task takeover after automatic switchover.

[0066] Through the coordinated operation of the above modules, the present invention realizes a high-availability automatic switching mechanism for the cloud platform management system in the event of a master node failure, ensures task continuity and data consistency at the platform level, and is suitable for key business scenarios in multi-node cloud management systems.

[0067] like Figure 4 、 5 As shown in the figure, the cloud platform's unified control and virtual machine high-availability switching process includes: 1. Heartbeat Monitoring Mechanism: Each node periodically performs heartbeat monitoring to continuously monitor the health of the master node within the cluster. If the master node is detected to be unresponsive for multiple consecutive times, the system automatically determines that it has failed and immediately initiates a master-slave switchover process.

[0068] 2. Master Node Election and Virtual IP Switching: The backup node initiates master node election based on a preemptive election algorithm. Upon successful election, the winning node automatically binds to the virtual IP address of the failed master node and notifies the network layer via an ARP broadcast, enabling rapid routing redirection and ensuring uninterrupted service.

[0069] 3. Activation of master control functions: After taking over, the new master node will immediately activate the unified control module and assume core business management responsibilities, including: data collection module (such as virtual machine operation status and resource usage monitoring), task scheduling module (such as migration scheduling and status determination), virtual machine migration module (such as smooth switching of key businesses), etc.

[0070] 4. Task Recovery Mechanism: The new master node receives and parses task status information transmitted by the data synchronization module. If any unfinished tasks are found, they will automatically resume execution based on their progress, ensuring task continuity. If a task fails, the system also supports policy-based retry triggering, improving task stability and robustness.

[0071] 5. Key virtual machine high availability guarantee: when the data acquisition module detects that the key virtual machine is unavailable, the new master node will immediately start the migration process. The identification of the key virtual machine is based on the pre-set strategy identification and business importance configuration to ensure the stable and continuous operation of the core business system.

[0072] 6. Hot standby node management: the backup node that is not elected as the master node will automatically enter the hot standby state, keep data synchronization with the master node, but not provide services to the outside, and be on standby to participate in the next round of master-standby switching to improve the overall availability and response speed of the system.

[0073] The above embodiments of the present application face the high availability automatic switching mechanism of the cloud platform management system, solve the management complexity and response lag problems caused by manual switching of master-standby nodes in traditional systems, and improve the self-fault-tolerant and continuous operation capability of the system. Specifically, the stability and reliability of the system are improved: by building an automatic master-standby switching mechanism, the system can quickly and stably switch to the standby node when the master control node fails, effectively avoiding system unavailability and business interruption problems; data consistency and task continuity are guaranteed: through the efficient data synchronization mechanism between the master-standby nodes, the key data such as management state, scheduling task, and virtual machine control information are kept consistent between the nodes to prevent state loss and task duplication; automatic and intelligent high availability control is realized: combined with health detection, state perception, and scheduling logic, the master-standby takeover control process without human intervention is realized, greatly improving the intelligence and response speed of system fault handling; the self-recovery capability and management flexibility of the system are enhanced: a unified control module is introduced to continuously monitor the master control service, virtual machine state, and host resource situation, automatically schedule resources and restore cloud platform management capability in abnormal conditions, and realize the continuity and automatic repair of platform services.

[0074] Example 1 For a cloud platform including 1 master and 5 backups, a total of 6 management nodes, the master node is responsible for overall management of 30 virtual machines on 10 host computers, of which 2 are key business virtual machines; example 1 provides a switching control method for cloud platform management nodes, including: Step 21, master node failure detection and judgment: The master node is in a non-responsive state due to a sudden disk IO failure, and the 5 backup nodes send detection signals containing random check codes at a pre-set period (such as every 50ms), and if the master node does not respond for 3 consecutive times (exceeding the fixed time limit of 100ms), the system determines that the master node has failed, triggering the switching process; Step 22, new master node selection process: The 5 backup nodes immediately suspend external data processing services and only retain election communication functions; after qualification verification, one node is excluded due to a data synchronization delay exceeding the threshold (average delay 1.2s), and the remaining 4 nodes enter the election; Priority Formula Z p 1 A p 2 B p 3 C p 4 D p 5 E Calculation: Node C's hardware resource health (CPU usage 15%, memory usage 20%), historical takeover success rate (all 3 attempts were successful), data synchronization delay (0.08s), and network stability (connectivity rate 99.9%, packet loss rate 0.1%) comprehensive performance is the best, with the highest score as the new master node; Step 23, virtual IP switching and route redirection: The system automatically removes the binding of the original master node and the virtual IP (10.0.0.10) to the network interface of node C, and notifies all hosts and network devices to update the address mapping through ARP broadcast, and the access request of external OA system and customer service system is seamlessly routed to node C; Step 24, task takeover and business recovery: Node C obtains the synchronization data of the original master node through multicast communication, including: 20 normal running task records, 5 pending tasks (such as host resource patrol), 3 interrupted tasks (including 1 OA virtual machine migration to new host unfinished task); For the interrupted OA virtual machine migration task, node C determines that it is in the resource allocation phase, and continues to execute: complete the CPU / memory allocation of target host 4 (8 cores 16GB), OA data and network configuration backup, and trigger the OA virtual machine on source host 2 to restore factory settings to avoid conflicts; Detecting that the customer service virtual machine appears a short stall (CPU usage rises to 95%) due to the original master node failure, node C prioritizes task scheduling, triggers migration through the task scheduling module, and migrates it to a host with sufficient resources within 3 minutes. 6, service is restored to normal; For the 5 pending tasks, sort by priority (customer service system log backup is prior to non-critical virtual machine snapshot), execute in turn, and monitor resources through the host management module during execution to avoid overload.

[0075] ​​​​​​​​​​The application can automatically detect the state of the master node, trigger the fault switching, select a new master node from the backup node without manual intervention, greatly shorten the fault recovery time, avoid system paralysis caused by single point failure, and ensure the continuous operation of the core function of the cloud platform.

[0076] The virtual IP address binding and ARP broadcast routing redirection mechanism enable the new master node to seamlessly take over the network identity of the original master node, automatically route external requests to the new node without modifying the configuration, realize the non-perception of service access to master node switching, and reduce the impact of service interruption.

[0077] The master and backup nodes synchronize critical data such as task records and virtual machine states in real time through a multicast communication mechanism, the new master node can accurately obtain complete task information and verify integrity, avoiding task loss or repeated execution; combined with task state judgment and priority scheduling, the seamless continuation of unfinished tasks is realized, and the business process is uninterrupted.

[0078] The new master node monitors the host resources in real time and reasonably allocates them, avoiding conflicts and waste; it prioritizes processing of critical virtual machine tasks and quickly triggers migration in the event of an exception, ensuring the continuous operation of core business and improving the risk resistance of high-priority business.

[0079] By eliminating unqualified nodes through qualification verification, and combining a multi-dimensional priority formula (hardware health, historical performance, data synchronization quality, etc.), the backup nodes are quantitatively evaluated to ensure that the selected new master node has comprehensive advantages, reduce system risks after switching, and improve the risk resistance of the cluster.

[0080] As shown in Figure 6 The embodiment of the application also provides a switching control device 60 of a cloud platform management node, which comprises: An acquisition module 61 is configured to acquire state data of a cloud platform management master node. A processing module 62 is configured to select a new cloud platform management master node from at least two groups of cloud platform management backup nodes when the state data is a fault; bind the virtual IP address of the original cloud platform management master node with the new cloud platform management master node, and complete routing redirection; control the task switching of the original cloud platform management master node to the new cloud platform management master node for processing; and control the cloud platform management master node and the cloud platform management backup node to keep data synchronization.

[0081] Optionally, the state data is a response signal of the cloud platform management master node to a detection signal periodically sent by the cloud platform management backup node.

[0082] Optionally, when the state data is a fault, the new cloud platform management master node is selected from at least two groups of cloud platform management backup nodes, comprising: When the status data is unresponsive for multiple consecutive times, it is determined that the cloud platform management master node is faulty; Control all cloud platform management backup nodes to enter the election state and start the master node election process; After successful selection, the winning cloud platform management backup node becomes the new cloud platform management master node.

[0083] Optionally, control all cloud platform management backup nodes to enter the election state and start the master node election process, including: Control all cloud platform management backup nodes to suspend external data processing services, retaining only campaign-related communication functions; Conduct qualification verification on each cloud platform management backup node; according to Z = p 1 × A + p 2 × B + p 3 × C + p 4 × D + p 5 × E ,determine the priority score of each cloud platform management backup node; in, Z Manage the priority scores of backup nodes for the cloud platform, A For the preset basic level, B For hardware resource health, C is the historical takeover success rate, D For data synchronization delay, E For network connection stability, p 1 is the first weight, p 2 is the second weight, p 3 is the third weight, p 4 is the fourth weight, p 1 is the fifth weight; The cloud platform management backup node with the highest priority score is determined as the new cloud platform management master node.

[0084] Optionally, the hardware resource health, historical takeover success rate, data synchronization delay, and network connection stability are determined in the following manner: according to B = Y- ( p6 × B 1 +p 7 × B 2 +p 8 × B 3 determining the hardware resource health degree of each cloud platform management backup node; wherein, B is the hardware resource health degree of the cloud platform management backup node, Y is the highest value of the hardware resource health degree, B 1 is the CPU usage rate, B 2 is the memory occupancy rate, B 3 is the disk IO load rate, p 6 is the sixth weight, p 7 is the seventh weight, p 8 is the eighth weight; according to C = C 1 / C 2 determining the historical takeover success rate of each cloud platform management backup node; wherein, C is the historical takeover success rate of the cloud platform management backup node, C 1 is the number of successful takeovers, C 2 is the total number of takeover attempts; according to D=1÷ ( 1+D 1 determining the data synchronization delay of each cloud platform management backup node; wherein, D is the data synchronization delay of the cloud platform management backup node, D 1 is the average delay in seconds; according to E= ( p 9 × E 1 +p 10 × E 2 determining the network connection stability of each cloud platform management backup node; wherein, EFor the network connection stability of the cloud platform management backup node, E 1 For the network connectivity rate, E 2 For the packet loss rate, p 9 For the ninth weight, p 10 For the tenth weight.

[0085] Optionally, the virtual IP address of the original cloud platform management master node is bound to the new cloud platform management master node, and routing redirection is completed, including: Unbinding the original cloud platform management master node and the virtual IP address; Binding the virtual IP address to the network interface of the new cloud platform management master node; Sending address mapping update notifications to all devices in the network through address resolution protocol (ARP) broadcast to complete routing redirection.

[0086] Optionally, the task switching of the original cloud platform management master node is controlled to the new cloud platform management master node for processing, including: Controlling the new cloud platform management master node to obtain tasks synchronized from the original cloud platform management master node; Controlling the new cloud platform management master node to process the tasks of the original cloud platform management master node.

[0087] In some embodiments, on the basis of "at least two groups of backup nodes", two independent lease sources (Lease-A / Lease-B) and a monotonically increasing epoch number are introduced. Only when the same backup node holds valid Lease-A and Lease-B at the same time, and successfully elects a new epoch and fencing token, is it allowed: 1. Bind the virtual IP and activate the unified control module; 2. Issue control instructions with epoch and fencing_token to storage / task / network subsystems; 3. Hardly reject or isolate instructions from the old master node after the recovery of the old master node, and realize the prevention of brain split and write isolation.

[0088] Before the master election, the backup nodes respectively initiate lease applications to the first and second lease sources, obtaining LA and LB; record their respective TTL and expiration time. The winning node generates a monotonically increasing epoch (64-bit unsigned integer) and a one-time fencing_token from the security module. Activation conditions: LA and LB are valid at the same time and the remaining time tr = min(TTL(LA), TTL(LB)) ≥ τ (security threshold). Otherwise, only enter the shadow state. Unified verification: all key operations on shared resources (such as task state submission, metadata writing, snapshot marking, and route injection) must carry <epoch, fencing_token>; subsystems locally save the latest confirmed epoch*, and if an epoch < epoch* request is received or the fencing_token does not match, it is directly rejected. Storage level isolation: When the new master is activated, exclusive reservations / blacklist instructions are issued to the shared resource layer to ensure that the old master cannot write to shared volumes or state libraries even if it "resumes". Clock and validity period guarantee: Use a monotonic hardware timer (such as TSC / HPET or a monotonic source after PTP calibration) to calculate the remaining lease time, avoiding false positives caused by system clock rollback; when tr ≤ τ, enter the renewal period, and if renewal fails, downgrade and deactivate and release the virtual IP.

[0089] Before selecting a new master node, the cloud platform management backup node respectively obtains lease tokens LA and LB from independent first and second lease sources; only when LA and LB are both within the valid period, the virtual IP is allowed to be bound and the new master node is activated. After the election is successful, a monotonically increasing generation number epoch and a one-time fencing_token are generated; all key control instructions for shared resources carry <epoch, fencing_token>, and if the target subsystem detects that the epoch is lower than its confirmed latest generation or the token is invalid, the instruction execution is rejected. When the new master node is activated, exclusive access control is issued to the shared resource layer to isolate the old master node from writing, so that the old master node cannot perform write operations on shared resources after recovery. Lease validity is based on a monotonic timer to calculate the remaining time, and when the remaining time is less than the threshold τ, renewal is triggered; if the renewal fails, the virtual IP is automatically released and the node enters an inactive state. The outstanding technical effect is to prevent split-brain: double lease + epoch + fencing forms a triple constraint, eliminating "double master simultaneous writing" after partition recovery. Strongly consistent takeover: use epoch as a global sequence, all old master's behind instructions are hard blocked, avoiding state rollback and task confusion. Engineering verifiable: consistency can be verified by injecting network partition / clock rollback faults, and switching accuracy can be quantitatively evaluated.

[0090] In some embodiments, further comprising introducing an online prediction model (such as: exponential weighted + sequential test, or Kalman filter for joint state estimation of CPU / IO / delay) on top of existing "health B, synchronization delay D, network stability E" indicators, when determining that the probability of failure within future At exceeds threshold θ, pre-selecting a shadow master and pre-warming: pre-synchronization: synchronizing task queues, scheduling contexts, hot state increments of key VMs to the shadow master; pre-drift: mirroring a differential set of four-layer connection / session states (conntrack entries); lossless switching: when the switching condition is reached, only performing VIP fast switching + differential compensation, keeping existing long connections open, achieving zero or near-zero packet loss. Prediction trigger: using the collected B1 CPU, B2 memory, B3 disk IO, D1 synchronization delay, E1 connectivity rate, E2 packet loss rate to form an observation vector, online estimating the master node health state ŝ(t) and its drift trend dŝ / dt; when P(failure | At) ≥ θ or consecutive L times of early warning are triggered, entering the pre-switching stage. Shadow master selection and pre-warming: selecting the shadow master with the highest score according to the existing priority scoring formula (not serving externally, only "on standby"); the shadow master receives the task queue increments of the master node, checkpoint metadata of key virtual machines, and routing / scheduling contexts. Connection state mirroring: establishing flow_id = (5-tuple, epoch) for active connections, periodically pushing differential entries (newly added / changed / soon to expire) to the shadow master, and the shadow master only forwards externally after "activation". Switch execution: when the switching criteria (predicted to actual failure or master node lease expiration) are met, first freeze the outgoing queue of the old master, broadcast ARP updates and replay the differential connection table on the shadow master side, restore forwarding after sequence number compensation based on ΔSEQ / ΔACK, and realize long connection without interruption. Backoff mechanism: if the early warning is lifted (P(failure | At) < θ') and the switch is not activated during this period, discard the differential mirror on the shadow master and restore normality, only retaining low-frequency state baselines.

[0091] In some embodiments, the predictive pre-switching step: based on time series of CPU usage, memory occupancy, disk IO load, data synchronization delay, network connectivity rate and packet loss rate, online estimates the master node health status, when predicting that the probability of failure within a preset time window Delta t is greater than a threshold theta, selects a shadow master node and performs preheating. The preheating includes: incrementally synchronizing the task queue and scheduling context; generating checkpoint metadata for virtual machines marked as critical services and synchronizing to the shadow master. The lossless traffic drift includes: before switching, mirroring the session state of the master node to the shadow master node in a differential manner; when switching, replaying the differential on the shadow master node and performing sequence number compensation, thereby maintaining existing connections open. If the prediction alarm is removed and no switching occurs, discard the differential session mirror of the shadow master node and revert to normal synchronization frequency. The prominent technical effect is that RTO / RPO is advanced: the recovery point and recovery time are moved forward to before the failure occurs, realizing an engineering path for "second-level or even sub-second-level" master control switching. Lossless experience: connection-level differential mirroring + sequence number compensation, maintaining existing TCP / UDP sessions open, and the service end is "unaware" of the switching. Controllable rollback: automatically rollback when false alarm, avoiding unnecessary switching jitter.

[0092] It should be noted that the device corresponds to the above method, and all implementation manners in the above method embodiments are applicable to this embodiment, and the same technical effects can also be achieved.

[0093] Embodiments of the application also provide a computing device, comprising: one or more processors; a storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, so that the one or more processors implement the method as described above.

[0094] Embodiments of the application also provide a computing device readable storage medium, storing instructions, when the instructions are run on a computing device, so that the computing device executes the method as described above. All implementation manners in the above method embodiments are applicable to this embodiment, and the same technical effects can also be achieved.

[0095] Those skilled in the art can appreciate that the units and algorithm steps of the examples described in combination with the embodiments disclosed in the present application can be realized by electronic hardware, or a combination of software and electronic hardware of a computing device. Whether the functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0096] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the system, device and unit described above can refer to the corresponding processes in the foregoing method embodiments, which will not be repeated here.

[0097] In the embodiments provided by the present application, it should be understood that the disclosed device and method can be implemented in other ways. For example, the device embodiments described above are merely schematic, for example, the division of the units is only a logical function division, and there can be another division manner in actual implementation, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interfaces, devices or units, and can be electrical, mechanical or in other forms.

[0098] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, they can be located in one place, or can be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0099] In addition, each functional unit in each embodiment of the present application can be integrated into a processing unit, or each unit can exist physically independently, or two or more units can be integrated into one unit.

[0100] If the functions are realized in the form of software function units and sold or used as independent products, they can be stored in a computer device readable storage medium. Based on this understanding, the technical solutions of the present application or the parts of the present application that make essential contributions to the prior art or the parts of the technical solutions can be embodied in the form of software products. The computer device software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer device, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The foregoing storage medium includes various program code storage media such as U disk, mobile hard disk, ROM, RAM, magnetic disk or optical disk, etc.

[0101] Moreover, it is pointed out that in the device and method of the present application, obviously, the components or steps can be decomposed and / or recombined. These decompositions and / or recombination should be considered as equivalent solutions of the present application. Also, the steps of performing the above series of processes can naturally be executed in time sequence according to the order of description, but do not necessarily have to be executed in time sequence. Some steps can be executed in parallel or independently of each other. It can be understood by those skilled in the art that all or any steps or components of the method and device of the present application can be implemented in hardware, firmware, software or a combination thereof in any computing device (including a processor, a storage medium, etc.) or a network of computing devices, using basic programming skills upon reading the description of the present application.

[0102] Therefore, the object of the present application can also be achieved by running a program or a set of programs on any computing device. The computing device can be a commonly known general-purpose device. Therefore, the object of the present application can also be achieved only by providing a program product containing program code for implementing the method or device. That is, such a program product also constitutes the present application, and a storage medium storing such a program product also constitutes the present application. Obviously, the storage medium can be any commonly known storage medium or any storage medium developed in the future. It is also pointed out that in the device and method of the present application, obviously, the components or steps can be decomposed and / or recombined. These decompositions and / or recombination should be considered as equivalent solutions of the present application. Also, the steps of performing the above series of processes can naturally be executed in time sequence according to the order of description, but do not necessarily have to be executed in time sequence. Some steps can be executed in parallel or independently of each other.

[0103] The above is the preferred embodiment of the present application. It should be pointed out that for those skilled in the art, several improvements and refinements can be made without departing from the principles of the present application, and these improvements and refinements should also be considered as the protection scope of the present application.

Claims

1. A cloud platform management node switching control method, characterized in that: include: Obtain the status data of the cloud platform management master node; When the status data indicates a failure, a new cloud platform management master node is selected from at least two groups of cloud platform management backup nodes; Bind the virtual IP address of the original cloud platform management master node to the new cloud platform management master node and complete the routing redirection; Control the tasks of the original cloud platform management master node to be switched to the new cloud platform management master node for processing; and control the cloud platform management master node and the cloud platform management backup node to maintain data synchronization.

2. The cloud platform management node switching control method according to claim 1, characterized in that: The status data is a response signal of the cloud platform management master node to the detection signal periodically sent by the cloud platform management backup node.

3. The cloud platform management node switching control method according to claim 2, characterized in that: When the status data indicates a failure, a new cloud platform management master node is selected from at least two groups of cloud platform management backup nodes, including: When the status data is unresponsive for multiple consecutive times, it is determined that the cloud platform management master node is faulty; Control all cloud platform management backup nodes to enter the election state and start the master node election process; After successful selection, the winning cloud platform management backup node becomes the new cloud platform management master node.

4. The cloud platform management node switching control method according to claim 3, characterized in that: Control all cloud platform management backup nodes to enter the election state and start the master node election process, including: Control all cloud platform management backup nodes to suspend external data processing services, retaining only campaign-related communication functions; Conduct qualification verification on each cloud platform management backup node; according to Z = p 1 × A + p 2 × B + p 3 × C + p 4 × D + p 5 × E ,determine the priority score of each cloud platform management backup node; in, Z Manage the priority scores of backup nodes for the cloud platform, A For the preset basic level, B For hardware resource health, C is the historical takeover success rate, D For data synchronization delay, E For network connection stability, p 1 is the first weight, p 2 is the second weight, p 3 is the third weight, p 4 is the fourth weight, p 1 is the fifth weight; The cloud platform management backup node with the highest priority score is determined as the new cloud platform management master node.

5. The cloud platform management node switching control method according to claim 4, characterized in that: Methods for determining the hardware resource health, historical takeover success rate, data synchronization delay, and network connection stability include: according to B=Y- ( p 6 × B 1 +p 7 × B 2 +p 8 × B 3 ), determine the health of hardware resources of each cloud platform management backup node; in, B Manage the hardware resource health of backup nodes for the cloud platform, Y is the highest value of hardware resource health. B 1 is the CPU usage, B 2 is the memory usage, B 3 is the disk IO load rate, p 6 is the sixth weight, p 7 is the seventh weight, p 8 is the eighth weight; according to C=C 1 / C 2 ,determine the historical takeover success rate of each cloud platform management backup node; in, C Manage the historical takeover success rate of backup nodes for the cloud platform, C 1 is the number of successful takeovers, C 2 is the total number of attempts to manage the tube; according to D=1÷ ( 1+D 1 ), determine the data synchronization delay of each cloud platform management backup node; in, D Manage data synchronization delays of backup nodes for cloud platforms, D 1 is the average delay in seconds; according to E= ( p 9 × E 1 +p 10 × E 2 ), determine the network connection stability of each cloud platform management backup node; in, E Manage the network connection stability of backup nodes for the cloud platform, E 1 is the network connectivity rate, E 2 is the packet loss rate, p 9 is the ninth weight, p 10 The tenth weight.

6. The cloud platform management node switching control method according to claim 1, characterized in that: Bind the virtual IP address of the original cloud platform management master node to the new cloud platform management master node and complete the routing redirection, including: Unbind the original cloud platform management master node and virtual IP address; Bind the virtual IP address to the network interface of the new cloud platform management master node; Address mapping update notifications are sent to all devices in the network through Address Resolution Protocol (ARP) broadcasts to complete route redirection.

7. The cloud platform management node switching control method according to claim 1, characterized in that: Controlling the task switching of the original cloud platform management master node to the new cloud platform management master node for processing includes: Control the new cloud platform management master node to obtain tasks synchronized from the original cloud platform management master node; Control the new cloud platform management master node to process the tasks of the original cloud platform management master node.

8. A switching control device for a cloud platform management node, characterized in that: include: The acquisition module is used to obtain the status data of the cloud platform management master node; a processing module, configured to select a new cloud platform management master node from at least two groups of cloud platform management backup nodes when the status data indicates a failure; Bind the virtual IP address of the original cloud platform management master node to the new cloud platform management master node and complete the routing redirection; control the task switching of the original cloud platform management master node to the new cloud platform management master node for processing; and control the cloud platform management master node and the cloud platform management backup node to maintain data synchronization.

9. A computing device, characterized in that include: one or more processors; A storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.

10. A computing device readable storage medium, characterized in that: The computing device readable storage medium stores a program, which implements the method according to any one of claims 1 to 7 when executed by a processor.

Images