Inventory real-time monitoring and abnormity repairing system based on multi-node redundancy check

Abstract

The invention discloses an inventory real-time monitoring and abnormity repairing system based on multi-node redundancy check. According to the system, fragmentation and multi-replica redundancy guarantee of inventory data are achieved through a multi-node redundancy storage module, Hash fingerprints are generated through a real-time check monitoring module, and abnormity is found through quick comparison; the fault reason and the influence range are analyzed through the anomaly detection and positioning module, copy reconstruction and transaction rollback are completed through the automatic repair and rollback module, and the repair progress and the push alarm are displayed in real time through the visual monitoring and alarm module. According to the inventory real-time monitoring and abnormity repairing system based on the multi-node redundancy check, consistency maintenance and automatic repairing of inventory data are achieved, and continuous and stable operation of businesses is guaranteed. The system is characterized by comprising a multi-node redundancy storage module, a real-time verification monitoring module, an anomaly detection and positioning module, an automatic repair and rollback module and a visual monitoring and alarm module.

Description

Technical Field

[0001] This invention relates to a real-time inventory monitoring and anomaly repair system based on multi-node redundancy verification for inventory management scenarios. It belongs to the technical field of real-time monitoring and anomaly repair systems based on multi-node redundancy verification, and specifically relates to a system that achieves data sharding and multi-replica redundancy through a multi-node redundant storage module, generates hash fingerprints through a real-time verification monitoring module for rapid comparison and anomaly detection, analyzes the cause and impact of faults through an anomaly detection and location module, completes replica reconstruction and transaction rollback through an automatic repair and rollback module, and displays the repair progress and pushes alarms in real time through a visual monitoring and alarm module. This system achieves consistent inventory data maintenance and automated repair to ensure continuous and stable business operation. Background Technology

[0002] With the continuous development of e-commerce, logistics, and intelligent manufacturing, enterprises have placed higher demands on inventory management systems, especially in terms of data real-time performance, accuracy, and reliability. Existing inventory management systems generally use centralized databases or a single-master-multiple-slave architecture to record inventory data. When encountering node failures, network latency, or high-concurrency writes, issues such as inventory data loss, recording errors, duplicate deductions, or data synchronization delays can easily occur, leading to discrepancies between the recorded inventory and the actual inventory. This, in turn, affects order fulfillment and the overall operational efficiency of the enterprise. Current common solutions include regular data backups, manual inventory checks, master-slave database synchronization, and cache refresh. While these methods can partially alleviate the problems, they often suffer from shortcomings such as untimely error detection, long repair cycles, the need for manual intervention, and the inability to accurately pinpoint the source of the problem. Furthermore, they lack continuous monitoring and early warning of the health status of system nodes, often only triggering large-scale data recovery after a node has completely failed, which can easily cause business interruptions and the overwriting of normal data.

[0003] Publication No. CN106598791A discloses a machine learning-based method for preventative identification of industrial equipment faults. This method employs machine learning algorithms to continuously learn and update a state prediction model, enabling real-time monitoring and prediction of the operational status of industrial equipment. It detects early signs of abnormality, issues timely alarms, and prevents accidents. This invention not only uses typical fault cases throughout the entire lifecycle of industrial equipment as learning objects but also incorporates archival data, environmental conditions of the operating area, and especially reliability data of key components. This multi-faceted approach to assessing the reliability of industrial equipment leads to more accurate predictions. However, while the aforementioned system introduces inventory quantity and procurement cycle as auxiliary parameters for fault prediction, it does not address the consistency verification and real-time monitoring of the inventory data itself. In distributed or multi-node application environments, if node failures, network latency, or write conflicts occur, the system cannot proactively detect data loss, duplicate deductions, or delayed synchronization issues. This can easily lead to discrepancies between recorded and actual inventory, increasing the cost of subsequent manual verification and repair. Summary of the Invention

[0004] To improve the above situation, the present invention provides a real-time inventory monitoring and anomaly repair system based on multi-node redundancy verification. This system achieves data sharding and multi-replica redundancy through a multi-node redundant storage module, generates hash fingerprints through a real-time verification monitoring module and quickly compares them to detect anomalies, analyzes the cause and scope of the failure through an anomaly detection and location module, completes replica reconstruction and transaction rollback through an automatic repair and rollback module, and displays the repair progress and pushes alarms in real time through a visualization monitoring and alarm module. This system ensures the consistency maintenance and automated repair of inventory data, guaranteeing the continuous and stable operation of business.

[0005] The real-time inventory monitoring and anomaly repair system based on multi-node redundancy verification of this invention is implemented as follows: The real-time inventory monitoring and anomaly repair system based on multi-node redundancy verification of this invention includes a multi-node redundant storage module, a real-time verification monitoring module, an anomaly detection and location module, an automatic repair and rollback module, and a visual monitoring and alarm module. The feature is that the multi-node redundant storage module establishes signal interaction with the real-time verification and monitoring module; the real-time verification and monitoring module establishes signal interaction with the anomaly detection and location module; the anomaly detection and location module establishes signal interaction with the automatic repair and rollback module; the automatic repair and rollback module establishes signal interaction with the multi-node redundant storage module; the multi-node redundant storage module establishes signal interaction with the visual monitoring and alarm module; and the visual monitoring and alarm module establishes signal interaction with the operation and maintenance terminal. The multi-node redundant storage module is responsible for distributing and storing inventory data across multiple computing nodes, and ensuring data reliability through a redundancy mechanism. The multi-node redundant storage module includes a data sharding submodule, a replica management submodule, a write synchronization submodule, and a node balancing and scaling submodule.

[0006] The data sharding submodule performs hash sharding on inventory data according to multiple feature dimensions such as SKU number, warehouse location, and timestamp. It uses a consistent hashing algorithm to distribute the data evenly across different nodes to avoid overloading a single node or making it a performance bottleneck. Preferably, the data sharding submodule can automatically adjust the sharding range when the business scales up, and dynamically rebalance when adding or removing nodes, to ensure load balancing and data distribution stability within the cluster. The replica management submodule generates a primary replica and multiple redundant replicas for each data shard, and generates check blocks when necessary. The check blocks are generated using Reed-Solomon or LDPC erasure coding, enabling the original data to be recovered even if a certain number of replicas are lost, thereby significantly improving the system's fault tolerance. Preferably, the replicas are distributed across different physical nodes and can be distributed across data centers in different geographical locations as needed, so that the system can still provide complete data services even if a local data center fails.

[0007] The write synchronization submodule is responsible for ensuring data consistency across replicas. Each time inventory data changes, a multi-node write confirmation is performed using a distributed consensus protocol (Raft or Paxos). A success result is only returned to the business layer after a majority of nodes have completed the write. Simultaneously, the system generates a globally unique transaction sequence number for each write and records it in the write log for easy traceability and, if necessary, transaction replay. The node balancing and scaling submodule is used to maintain the overall stability of the system. When the system detects that the node load is close to the threshold, it will automatically trigger shard migration, migrating some shards to other idle nodes. When a node fails, it will immediately start the replica reconstruction process to restore the missing data on other healthy nodes. Preferably, the node balancing and scaling submodule supports online scaling up and down, and through dynamic migration and data rebalancing, the scaling up process does not affect the continuous operation of services. The real-time verification and monitoring module is used to continuously verify and validate the inventory data of each node, ensuring that anomalies are detected and repair processes are triggered as soon as possible. The real-time verification and monitoring module includes a hash generation submodule, a comparison submodule, and a hierarchical verification scheduling submodule. The hash generation submodule generates a corresponding hash fingerprint after each data write, using either the SHA-256 or BLAKE3 hash algorithm to ensure the uniqueness and collision resistance of the hash result. Preferably, to improve security, the system introduces a random salt value when generating hashes. Preferably, in addition to row-level hashing, the system also establishes shard-level and library-level Merkle tree structures to support fast consistency verification of batch data.

[0008] The comparison submodule periodically retrieves the latest hash value from the primary replica node and compares it with the local verification replica. The comparison process adopts an asynchronous execution mode and incremental comparison is achieved by recording the version number of the last comparison, thereby reducing computational overhead. When an inconsistency is found, the shard is immediately marked as abnormal and added to the repair task queue. Preferably, for large-capacity fragments, the comparison process uses a block-based approach to quickly locate erroneous data blocks.

[0009] The hierarchical verification and scheduling submodule divides data into hot data, warm data, and cold data based on the data access frequency. Hot data is verified at the second level, warm data at the minute level, and cold data at the hourly level, thereby ensuring consistency while also considering system performance. Preferably, when system resources are scarce, the scheduler will reduce the verification frequency of low-priority fragments to prioritize the real-time verification of critical data. The node health monitoring submodule is responsible for collecting the node's operating status, including CPU utilization, disk I / O latency, disk health indicators, network packet loss rate, etc. It calculates a node health score based on these indicators. When the health score falls below a threshold, the node is marked as a high-risk node, and replica migration or hardening measures are triggered in advance to prevent the fault from escalating. The anomaly detection and localization module is used to analyze inconsistencies found during the verification process and determine the root cause and scope of impact of the anomalies.

[0010] The anomaly detection and localization module includes an anomaly discrimination submodule, a version backtracking submodule, an impact analysis submodule, and a prediction and early warning submodule. The anomaly detection submodule analyzes transaction logs, node heartbeats, verification comparison results, and system monitoring indicators to determine whether the anomaly is caused by network latency, disk corruption, node crashes, or application write conflicts. Preferably, the anomaly detection submodule supports multiple sampling confirmations, which can effectively reduce false alarms caused by instantaneous network jitter. The version backtracking submodule maintains an incremental snapshot linked list for each shard and uses binary search to quickly locate the version that first encountered an anomaly when an anomaly occurs. Preferably, the version rollback submodule accurately determines the affected data range and only rolls back the necessary data, avoiding the loss of valid data caused by large-scale rollback. The impact analysis submodule calculates the number of SKUs, order quantity, and inventory value involved in the abnormal data, and generates a repair priority list based on customer priority and warehouse location importance to ensure that critical business operations are restored first, minimizing downtime losses. The predictive early warning submodule trains a machine learning model to perform pattern recognition on historical node performance data, enabling it to identify nodes that may fail in advance. When a node's disk I / O latency, CPU load, or other metrics remain abnormal, the system issues an early warning and proactively migrates the critical data shards it carries, reducing the cost and time required for future repairs. The automatic repair and rollback module enables the system to automatically repair itself, ensuring data consistency is restored without interrupting business operations. The automatic repair and rollback module includes a redundancy reconstruction submodule, a transaction rollback submodule, a bypass repair submodule, and a disaster recovery switching and retry submodule. When the redundancy reconstruction submodule detects that the primary or parity replica is corrupted, it reconstructs the missing data from other healthy replicas and erasure coding check blocks. The reconstruction process is executed in parallel by multiple nodes to shorten the recovery time. After the recovery is complete, the system regenerates the hash value and broadcasts it within the cluster to update the global consistency state. The transaction rollback submodule is used to roll back the sharded data to the most recent error-free snapshot, only undoing conflicting transactions and retaining the commit results of the remaining valid transactions. Preferably, undone transactions are re-executed using the transaction log replay function to ensure eventual data consistency. The bypass repair submodule ensures business continuity. During the repair process, read requests are redirected to a healthy replica or cache snapshot, and write requests are temporarily stored in a transaction queue and committed sequentially after the repair is complete. This ensures that the normal operation of the business layer is not affected. When a node or data center experiences a severe failure, the disaster recovery switching and retry submodule switches service traffic to a backup node or data center. Simultaneously, it retryes tasks that fail to be repaired multiple times and employs an exponential backoff algorithm to prevent concentrated retries from impacting the system. The visualization monitoring and alarm module provides operation and maintenance personnel with a real-time visualization interface and a multi-level alarm mechanism, enabling rapid response and closed-loop processing of faults. The visualization monitoring and alarm module includes a dashboard display submodule, a multi-level alarm submodule, an event tracking submodule, and an API interface submodule. The dashboard display submodule uses a visual panel to show node topology, shard distribution, node health status, global consistency score, repair time curve, and time trend of anomaly count. It also provides multi-dimensional filtering functions by time, SKU, and node to help operations and maintenance personnel quickly locate problems.

[0011] The multi-level alarm submodule classifies alarms into four levels—prompt, warning, critical, and fatal—based on the severity of the anomaly. Alarms of different levels are pushed through different channels, including system logs, email, instant messaging tools, SMS, and voice calls, to ensure that on-duty personnel receive notifications in a timely manner. The event tracking submodule records the entire process from anomaly discovery, location, repair to recovery completion, and automatically generates a closed-loop tracking link and audit report for subsequent compliance archiving and problem review analysis.

[0012] The API interface submodule provides a standardized RESTful interface for external enterprise resource management systems or business intelligence platforms to call, obtain real-time inventory health status, and push notifications to external systems when a repair task is completed or the node status changes, enabling automatic cross-system linkage and business recovery. Beneficial effects

[0013] I. Achieve high reliability and strong consistency of inventory data to ensure data integrity and security in a multi-node environment.

[0014] Second, it can detect and quickly locate inventory anomalies in milliseconds, significantly improving the real-time performance and accuracy of anomaly detection.

[0015] Third, automatically complete data repair and transaction rollback to ensure uninterrupted business operations and continuous system stability.

[0016] Fourth, it provides real-time visual monitoring and multi-level alarms, which facilitates rapid response and closed-loop processing by operation and maintenance personnel, thereby reducing management risks. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall system architecture of a real-time inventory monitoring and anomaly repair system based on multi-node redundancy verification according to the present invention. Figure 2 This is a schematic diagram of the data sharding and multi-replica storage structure of an inventory real-time monitoring and anomaly repair system based on multi-node redundancy verification according to the present invention. Figure 3 This is a schematic diagram of the workflow of a real-time inventory monitoring and anomaly repair system based on multi-node redundancy verification according to the present invention. Figure 4 This is a schematic diagram of the workflow of a real-time inventory monitoring and anomaly repair system based on multi-node redundancy verification according to the present invention. Detailed Implementation Example 1

[0018] The real-time inventory monitoring and anomaly repair system based on multi-node redundancy verification of this invention is implemented as follows: The real-time inventory monitoring and anomaly repair system based on multi-node redundancy verification of this invention includes a multi-node redundant storage module, a real-time verification monitoring module, an anomaly detection and location module, an automatic repair and rollback module, and a visual monitoring and alarm module. The feature is that the multi-node redundant storage module establishes signal interaction with the real-time verification and monitoring module; the real-time verification and monitoring module establishes signal interaction with the anomaly detection and location module; the anomaly detection and location module establishes signal interaction with the automatic repair and rollback module; the automatic repair and rollback module establishes signal interaction with the multi-node redundant storage module; the multi-node redundant storage module establishes signal interaction with the visual monitoring and alarm module; and the visual monitoring and alarm module establishes signal interaction with the operation and maintenance terminal. The multi-node redundant storage module is responsible for distributing and storing inventory data across multiple computing nodes, and ensuring data reliability through a redundancy mechanism. The multi-node redundant storage module includes a data sharding submodule, a replica management submodule, a write synchronization submodule, and a node balancing and scaling submodule.

[0019] The data sharding submodule performs hash sharding on inventory data according to multiple feature dimensions such as SKU number, warehouse location, and timestamp. It uses a consistent hashing algorithm to distribute the data evenly across different nodes to avoid overloading a single node or making it a performance bottleneck. Preferably, the data sharding submodule can automatically adjust the sharding range when the business scales up, and dynamically rebalance when adding or removing nodes, to ensure load balancing and data distribution stability within the cluster. The replica management submodule generates a primary replica and multiple redundant replicas for each data shard, and generates check blocks when necessary. The check blocks are generated using Reed-Solomon or LDPC erasure coding, enabling the original data to be recovered even if a certain number of replicas are lost, thereby significantly improving the system's fault tolerance. Preferably, the replicas are distributed across different physical nodes and can be distributed across data centers in different geographical locations as needed, so that the system can still provide complete data services even if a local data center fails.

[0020] The write synchronization submodule is responsible for ensuring data consistency across replicas. Each time inventory data changes, a multi-node write confirmation is performed using a distributed consensus protocol (Raft or Paxos). A success result is only returned to the business layer after a majority of nodes have completed the write. Simultaneously, the system generates a globally unique transaction sequence number for each write and records it in the write log for easy traceability and, if necessary, transaction replay. The node balancing and scaling submodule is used to maintain the overall stability of the system. When the system detects that the node load is close to the threshold, it will automatically trigger shard migration, migrating some shards to other idle nodes. When a node fails, it will immediately start the replica reconstruction process to restore the missing data on other healthy nodes. Preferably, the node balancing and scaling submodule supports online scaling up and down, and through dynamic migration and data rebalancing, the scaling up process does not affect the continuous operation of services. The real-time verification and monitoring module is used to continuously verify and validate the inventory data of each node, ensuring that anomalies are detected and repair processes are triggered as soon as possible. The real-time verification and monitoring module includes a hash generation submodule, a comparison submodule, and a hierarchical verification scheduling submodule. The hash generation submodule generates a corresponding hash fingerprint after each data write, using either the SHA-256 or BLAKE3 hash algorithm to ensure the uniqueness and collision resistance of the hash result. Preferably, to improve security, the system introduces a random salt value when generating hashes. Preferably, in addition to row-level hashing, the system also establishes shard-level and library-level Merkle tree structures to support fast consistency verification of batch data.

[0021] The comparison submodule periodically retrieves the latest hash value from the primary replica node and compares it with the local verification replica. The comparison process adopts an asynchronous execution mode and incremental comparison is achieved by recording the version number of the last comparison, thereby reducing computational overhead. When an inconsistency is found, the shard is immediately marked as abnormal and added to the repair task queue. Preferably, for large-capacity fragments, the comparison process uses a block-based approach to quickly locate erroneous data blocks.

[0022] The hierarchical verification and scheduling submodule divides data into hot data, warm data, and cold data based on the data access frequency. Hot data is verified at the second level, warm data at the minute level, and cold data at the hourly level, thereby ensuring consistency while also considering system performance. Preferably, when system resources are scarce, the scheduler will reduce the verification frequency of low-priority fragments to prioritize the real-time verification of critical data. The node health monitoring submodule is responsible for collecting the node's operating status, including CPU utilization, disk I / O latency, disk health indicators, network packet loss rate, etc. It calculates a node health score based on these indicators. When the health score falls below a threshold, the node is marked as a high-risk node, and replica migration or hardening measures are triggered in advance to prevent the fault from escalating. The anomaly detection and localization module is used to analyze inconsistencies found during the verification process and determine the root cause and scope of impact of the anomalies.

[0023] The anomaly detection and localization module includes an anomaly discrimination submodule, a version backtracking submodule, an impact analysis submodule, and a prediction and early warning submodule. The anomaly detection submodule analyzes transaction logs, node heartbeats, verification comparison results, and system monitoring indicators to determine whether the anomaly is caused by network latency, disk corruption, node crashes, or application write conflicts. Preferably, the anomaly detection submodule supports multiple sampling confirmations, which can effectively reduce false alarms caused by instantaneous network jitter. The version backtracking submodule maintains an incremental snapshot linked list for each shard and uses binary search to quickly locate the version that first encountered an anomaly when an anomaly occurs. Preferably, the version rollback submodule accurately determines the affected data range and only rolls back the necessary data, avoiding the loss of valid data caused by large-scale rollback. The impact analysis submodule calculates the number of SKUs, order quantity, and inventory value involved in the abnormal data, and generates a repair priority list based on customer priority and warehouse location importance to ensure that critical business operations are restored first, minimizing downtime losses. The predictive early warning submodule trains a machine learning model to perform pattern recognition on historical node performance data, enabling it to identify nodes that may fail in advance. When a node's disk I / O latency, CPU load, or other metrics remain abnormal, the system issues an early warning and proactively migrates the critical data shards it carries, reducing the cost and time required for future repairs. The automatic repair and rollback module enables the system to automatically repair itself, ensuring data consistency is restored without interrupting business operations. The automatic repair and rollback module includes a redundancy reconstruction submodule, a transaction rollback submodule, a bypass repair submodule, and a disaster recovery switching and retry submodule. When the redundancy reconstruction submodule detects that the primary or parity replica is corrupted, it reconstructs the missing data from other healthy replicas and erasure coding check blocks. The reconstruction process is executed in parallel by multiple nodes to shorten the recovery time. After the recovery is complete, the system regenerates the hash value and broadcasts it within the cluster to update the global consistency state. The transaction rollback submodule is used to roll back the sharded data to the most recent error-free snapshot, only undoing conflicting transactions and retaining the commit results of the remaining valid transactions. Preferably, undone transactions are re-executed using the transaction log replay function to ensure eventual data consistency. The bypass repair submodule ensures business continuity. During the repair process, read requests are redirected to a healthy replica or cache snapshot, and write requests are temporarily stored in a transaction queue and committed sequentially after the repair is complete. This ensures that the normal operation of the business layer is not affected. When a node or data center experiences a severe failure, the disaster recovery switching and retry submodule switches service traffic to a backup node or data center. Simultaneously, it retryes tasks that fail to be repaired multiple times and employs an exponential backoff algorithm to prevent concentrated retries from impacting the system. The visualization monitoring and alarm module provides operation and maintenance personnel with a real-time visualization interface and a multi-level alarm mechanism, enabling rapid response and closed-loop processing of faults. The visualization monitoring and alarm module includes a dashboard display submodule, a multi-level alarm submodule, an event tracking submodule, and an API interface submodule. The dashboard display submodule uses a visual panel to show node topology, shard distribution, node health status, global consistency score, repair time curve, and time trend of anomaly count. It also provides multi-dimensional filtering functions by time, SKU, and node to help operations and maintenance personnel quickly locate problems.

[0024] The multi-level alarm submodule classifies alarms into four levels—prompt, warning, critical, and fatal—based on the severity of the anomaly. Alarms of different levels are pushed through different channels, including system logs, email, instant messaging tools, SMS, and voice calls, to ensure that on-duty personnel receive notifications in a timely manner. The event tracking submodule records the entire process from anomaly discovery, location, repair to recovery completion, and automatically generates a closed-loop tracking link and audit report for subsequent compliance archiving and problem review analysis.

[0025] The API interface submodule provides a standardized RESTful interface for external enterprise resource management systems or business intelligence platforms to call, obtain real-time inventory health status, and push notifications to external systems when a repair task is completed or a node status changes, enabling automatic cross-system linkage and business recovery. In practice, inventory data is first sharded according to features such as SKU number, warehouse location, and timestamp. Primary and redundant replicas are then generated on multiple nodes to form a distributed redundant storage structure. When write, deduction, or update operations occur, the system instantly generates corresponding hash fingerprints and synchronizes them across nodes. A real-time verification module compares the hash values ​​of the primary and redundant replicas to promptly identify and mark inconsistent shards. Upon receiving anomaly information, the anomaly detection module retrieves transaction logs and version snapshots to trace the version from which the anomaly occurred, analyzes the cause of the anomaly, and determines the affected data range. This generates a priority order for repairs. Subsequently, the automatic repair module reconstructs the missing data in parallel from the healthy copy and the verification block based on the location results. If necessary, it performs transaction rollback to restore the data to the latest error-free state and writes it synchronously to the relevant nodes. The repair process is completed in the background, while the front-end business continues to provide query and deduction services through the healthy copy or cache snapshot to avoid business interruption. Throughout the entire operation, the system continuously transmits the node status, verification results, and repair progress to the visualization monitoring module for centralized display and pushes alarm information to the operation and maintenance terminal according to the anomaly level, forming a closed-loop process from anomaly discovery, analysis, repair to monitoring. The design of hashing and sharding inventory data according to SKU number, warehouse location and timestamp and generating primary and redundant replicas on multiple nodes can achieve balanced distribution of inventory data and multi-point redundancy protection. It can still reconstruct complete data when a node fails or a disk is damaged, which significantly improves the reliability of the system. The design of generating a hash fingerprint after each write operation and performing incremental comparison between the primary replica and the verification replica can detect data inconsistencies in milliseconds, promptly mark abnormal fragments and add them to the repair queue, effectively preventing the expansion of inventory errors. The design of using an incremental snapshot linked list to trace back to the first erroneous version and roll back only the affected data range can accurately locate the abnormal range, reduce the risk of irrelevant data being rolled back, and reduce the impact of the repair process on normal business. The design of switching read requests to a healthy copy or cache snapshot and temporarily storing write requests in a transaction queue during the repair process can ensure the continuous operation of front-end order placement, inventory query and other businesses, and make the repair process imperceptible to users. The design of displaying node topology, shard distribution, repair time curves and anomaly quantity trends through a visual panel and pushing alarm information in a hierarchical manner can help operation and maintenance personnel intuitively grasp the system status and respond quickly to anomalies, shortening the problem handling time. The design of automatically triggering shard migration when the node load approaches the threshold and performing online rebalancing when the node is expanded can dynamically distribute storage pressure and maintain the high efficiency and stability of the cluster operation. The system achieves the goal of ensuring consistent inventory data maintenance and automated repair to guarantee continuous and stable business operations. It can achieve inventory data sharding and multi-replica redundancy through multi-node redundant storage modules, generate hash fingerprints and quickly compare them to detect anomalies through real-time verification and monitoring modules, analyze the cause and scope of failures through anomaly detection and location modules, complete replica reconstruction and transaction rollback through automatic repair and rollback modules, and display the repair progress and push alarms in real time through visualization monitoring and alarm modules.

[0026] It should be noted that, unless otherwise explicitly specified and limited, the terms "placed," "connected," and "linked" should be interpreted broadly. For example, they can refer to fixed connections such as folded edges, rivets, pins, adhesives, and welds; detachable connections such as threaded connections, snap-fit ​​connections, and hinges; integral connections; electrical connections; direct connections; or indirect connections via an intermediate medium; or internal connections between two components. Those skilled in the art can understand the specific meaning of these terms in this invention based on the specific circumstances.

Claims

1. A real-time inventory monitoring and anomaly repair system based on multi-node redundancy verification, comprising a multi-node redundant storage module, a real-time verification monitoring module, an anomaly detection and location module, an automatic repair and rollback module, and a visual monitoring and alarm module, characterized in that: The multi-node redundant storage module is responsible for distributing inventory data across multiple computing nodes and ensuring data reliability through redundancy mechanisms. This module includes a data sharding submodule, a replica management submodule, a write synchronization submodule, and a node balancing and scaling submodule. The real-time verification and monitoring module continuously verifies and validates the inventory data on each node, ensuring that anomalies are detected and repair processes are triggered immediately. This module includes a hash generation submodule, a comparison submodule, and a hierarchical verification scheduling submodule. The anomaly detection and location module analyzes inconsistencies found during verification to determine the root cause and scope of impact. The anomaly detection and localization module includes an anomaly discrimination submodule, a version backtracking submodule, an impact analysis submodule, and a prediction and early warning submodule. The automatic repair and rollback module enables the system to automatically repair itself, ensuring data consistency is restored without interrupting business operations. The automatic repair and rollback module includes a redundancy reconstruction submodule, a transaction rollback submodule, a bypass repair submodule, and a disaster recovery switching and retry submodule. The visual monitoring and alarm module provides a real-time visual interface and a multi-level alarm mechanism for operation and maintenance personnel, enabling rapid fault response and closed-loop processing. The visual monitoring and alarm module includes a dashboard display submodule, a multi-level alarm submodule, an event tracking submodule, and an API interface submodule.

2. The real-time inventory monitoring and anomaly repair system based on multi-node redundancy verification according to claim 1, characterized in that... The data sharding submodule hashes the inventory data according to multiple feature dimensions such as SKU number, warehouse location, and timestamp. A consistent hashing algorithm is used to evenly distribute the data across different nodes to avoid overloading a single node or making it a performance bottleneck. The replica management submodule generates a primary replica and multiple redundant replicas for each data shard and generates check blocks when necessary. Check blocks are generated using Reed-Solomon or LDPC erasure coding, allowing the original data to be recovered even if a certain number of replicas are lost, thus significantly improving the system's fault tolerance. The write synchronization submodule is responsible for ensuring data consistency across replicas. Each time inventory data changes, a distributed consensus protocol (Raft) is used to synchronize the data. The system uses Paxos or other multi-node write confirmation mechanisms to ensure that a majority of nodes complete the write before returning a success result to the business layer. Simultaneously, the system generates a globally unique transaction sequence number for each write and records it in the write log for easy traceability and transaction replay when necessary. The node balancing and scaling submodule is used to maintain the overall stability of the system. When a node load is detected to be close to a threshold, the system automatically triggers shard migration, migrating some shards to other idle nodes. When a node fails, the replica reconstruction process is immediately initiated to restore the missing data on other healthy nodes.

3. The real-time inventory monitoring and anomaly repair system based on multi-node redundancy verification according to claim 1, characterized in that... The hash generation submodule generates a corresponding hash fingerprint after each data write, using SHA-256 or BLAKE3 hash algorithms to ensure the uniqueness and collision resistance of the hash result. The comparison submodule periodically obtains the latest hash value from the primary replica node and compares it with the local verification replica. The comparison process adopts an asynchronous execution mode and achieves incremental comparison by recording the version number of the last comparison, thereby reducing computational overhead. When an inconsistency is found, the fragment is immediately marked as abnormal and added to the repair task queue. The hierarchical verification scheduling submodule divides the data into hot data, warm data, and cold data according to the access frequency. Hot data is verified at the second level, warm data at the minute level, and cold data at the hourly sampling strategy, thereby ensuring consistency while taking into account system performance. The node health monitoring submodule is responsible for collecting the node's operating status, including CPU utilization, disk I / O latency, disk health indicators, network packet loss rate, etc. By calculating these indicators, a node health score is formed. When the health score is lower than the threshold, the node is marked as a high-risk node, and replica migration or hardening measures are triggered in advance to prevent the fault from escalating.

4. The real-time inventory monitoring and anomaly repair system based on multi-node redundancy verification according to claim 1, characterized in that... The anomaly detection submodule analyzes transaction logs, node heartbeats, verification comparison results, and system monitoring metrics to determine whether anomalies are caused by network latency, disk corruption, node downtime, or application write conflicts. The version backtracking submodule maintains an incremental snapshot linked list for each shard and uses binary search to quickly locate the version that first encountered the anomaly when it occurs. The impact analysis submodule calculates the number of SKUs, orders, and inventory amounts involved in the abnormal data and generates a repair priority list based on customer priority and warehouse location importance to ensure that critical business operations are restored first and reduce downtime losses. The prediction and early warning submodule trains a machine learning model to perform pattern recognition on historical node performance data to identify nodes that may fail in advance. When a node's disk I / O latency, CPU load, and other metrics remain abnormal, the system issues an early warning and proactively migrates the critical data shards it carries, reducing the cost and time of future repairs.

5. The real-time inventory monitoring and anomaly repair system based on multi-node redundancy verification according to claim 1, characterized in that... The redundancy reconstruction submodule reconstructs missing data from other healthy replicas and erasure coding check blocks when it detects damage to the primary or check replica. The reconstruction process is executed in parallel by multiple nodes to shorten recovery time. After recovery, the system regenerates hash values ​​and broadcasts them within the cluster to update the global consistency state. The transaction rollback submodule rolls back sharded data to the most recent error-free snapshot, only undoing conflicting transactions and retaining the commit results of other valid transactions. The bypass repair submodule ensures business continuity. During the repair process, read requests are redirected to healthy replicas or cached snapshots, and write requests are temporarily stored in the transaction queue and committed sequentially after repair, thus ensuring that the normal operation of the business layer is not affected. The disaster recovery switching and retry submodule switches business traffic to backup nodes or backup data centers when a node or data center experiences a serious failure. At the same time, it performs multiple retries on tasks that fail to repair and uses an exponential backoff algorithm to prevent concentrated retries from impacting the system.

6. The real-time inventory monitoring and anomaly repair system based on multi-node redundancy verification according to claim 1, characterized in that... The dashboard display submodule visually displays node topology, shard distribution, node health status, global consistency score, repair time curve, and time trend of anomaly quantity. It also provides multi-dimensional filtering functions by time, SKU, and node to help operations and maintenance personnel quickly locate problems. The multi-level alarm submodule classifies alarms into four levels—prompt, warning, severe, and critical—based on the severity of the anomaly. Alarms of different levels are pushed through different channels, including system logs, email, instant messaging tools, SMS, and voice calls, to ensure that on-duty personnel receive timely notifications. The event tracking submodule records the entire process from anomaly discovery, location, repair to recovery completion, and automatically generates a closed-loop tracking link and audit report for subsequent compliance archiving and problem review analysis. The API interface submodule provides a standardized RESTful interface for external enterprise resource management systems or business intelligence platforms to call, obtain inventory health status in real time, and push notifications to external systems when repair tasks are completed or node status changes, achieving automatic cross-system linkage and business recovery.

7. The real-time inventory monitoring and anomaly repair system based on multi-node redundancy verification according to claim 2, characterized in that... The data sharding submodule can automatically adjust the sharding range when the business scale expands and dynamically rebalance when adding or removing nodes to ensure load balancing and data distribution stability within the cluster. The replicas are distributed across different physical nodes and can be distributed across data centers in different geographical locations as needed to ensure that the system can still provide complete data services even if a local data center fails. The node balancing and scaling submodule supports online scaling up and down, and through dynamic migration and data rebalancing, the scaling process does not affect the continuous operation of the business.

8. The real-time inventory monitoring and anomaly repair system based on multi-node redundancy verification according to claim 3, characterized in that... When system resources are scarce, the scheduler will reduce the verification frequency of low-priority shards and prioritize the real-time verification of critical data. For large-capacity shards, the comparison process adopts a block-based approach to quickly locate erroneous data blocks. To improve security, the system introduces random salt values ​​when generating hashes. In addition to row-level hashes, the system also establishes shard-level and library-level Merkle tree structures to support fast consistency verification of batch data.

9. The real-time inventory monitoring and anomaly repair system based on multi-node redundancy verification according to claim 4, characterized in that... The version rollback submodule accurately determines the range of affected data and rolls back only the necessary data to avoid the loss of valid data caused by large-scale rollback. The anomaly detection submodule supports multiple sampling confirmations, which can effectively reduce false alarms caused by instantaneous network jitter. The undone transactions will be re-executed through the transaction log replay function to ensure the consistency of the final data.

10. The real-time inventory monitoring and anomaly repair system based on multi-node redundancy verification according to claim 1, characterized in that... The multi-node redundant storage module establishes signal interaction with the real-time verification and monitoring module, the real-time verification and monitoring module establishes signal interaction with the anomaly detection and location module, the anomaly detection and location module establishes signal interaction with the automatic repair and rollback module, the automatic repair and rollback module establishes signal interaction with the multi-node redundant storage module, the multi-node redundant storage module establishes signal interaction with the visualization monitoring and alarm module, and the visualization monitoring and alarm module establishes signal interaction with the operation and maintenance terminal.

Images