A container fault self-recovery method of a containerized platform of an underwater acoustic signal processor

Abstract

The application provides a container fault self-recovery method of a water acoustic signal processor containerized platform, comprising the following steps: obtaining hardware index parameters and service index parameters of any first node in the process of completing a water acoustic signal processing task; determining an abnormal probability of the first node based on the hardware and service index parameters; when the abnormal probability is higher than a warning threshold, determining a second node based on a pre-constructed node label registry, and preferentially querying a node with the same cluster ownership label as the first node cluster as the second node; the water acoustic signal stream input into the first node is synchronously forwarded to the second node, and a Pod image and a configuration file are loaded in the second node; when a failure of the first node is detected, the water acoustic signal stream in a time-sequenced cache queue is processed based on the Pod image and the configuration file in the second node. Through implementation of the application, the fault self-recovery capability can be improved, and the real-time performance of the water acoustic signal can be ensured.

Description

Technical Field

[0001] This invention belongs to the field of fault handling technology, and specifically relates to a container fault self-healing method for a containerized platform of underwater acoustic signal processor. Background Technology

[0002] With the technological advancements in fields such as marine exploration and underwater communication, underwater acoustic signal processors, as core equipment, are widely deployed on mobile platforms such as ships and unmanned underwater vehicles (UUVs), undertaking critical tasks such as underwater acoustic signal acquisition, filtering, spectrum analysis, and target identification. These mobile platforms have extremely stringent requirements for the real-time performance and reliability of underwater acoustic signal processing: on the one hand, the raw underwater acoustic signal data is large and time-sensitive, requiring preliminary preprocessing within microsecond delays to avoid signal attenuation and delay accumulation caused by transmission across hardware modules, otherwise it will affect the accuracy of target identification and the synchronization of underwater communication; on the other hand, mobile platforms face complex marine environments and interference such as equipment vibration and power supply fluctuations during navigation, which can easily lead to failures of the containers or hardware nodes carrying signal processing tasks, requiring rapid service recovery to avoid business interruption.

[0003] Existing container fault self-healing technologies are mainly developed for general IT scenarios, which are different from the local scheduling scenarios of containerized platforms for underwater acoustic signal processors. Since underwater acoustic signals generally require continuous processing, existing technologies are difficult to meet the real-time requirements for recovery after an underwater acoustic signal processing failure. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a container fault self-healing method for a containerized platform of underwater acoustic signal processor, so as to meet the real-time recovery requirements after an underwater acoustic signal processing failure.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] According to a first aspect, the present invention provides a container fault self-healing method for a containerized platform for underwater acoustic signal processing, comprising: acquiring hardware indicator parameters and business indicator parameters of any first node during the completion of an underwater acoustic signal processing task; determining the anomaly probability of the first node based on the hardware indicator parameters and business indicator parameters; when the anomaly probability is higher than a warning threshold, determining a second node based on a pre-built node label registry, prioritizing the query of nodes with the same cluster affiliation label as the first node and that are healthy and idle as the second node, and configuring nodes in the same physical domain with the same cluster affiliation label; synchronously forwarding the underwater acoustic signal stream input to the first node to the time-series cache queue of the second node, and loading the underwater acoustic signal stream processing Pod image and configuration file of the first node on the second node; when a fault is detected in the first node, processing the underwater acoustic signal stream in the time-series cache queue on the second node based on the pre-loaded underwater acoustic signal stream processing Pod image and configuration file.

[0007] According to the second aspect, this embodiment provides a container fault self-healing device for a containerized platform of underwater acoustic signal processors, comprising: a parameter acquisition module, used to acquire hardware indicator parameters and business indicator parameters of any first node during the completion of an underwater acoustic signal processing task; a probability determination module, used to determine the anomaly probability of the first node based on the hardware indicator parameters and business indicator parameters; a node determination module, used to determine a second node based on a pre-built node label registry when the anomaly probability is higher than a warning threshold, prioritizing the query of nodes with the same cluster affiliation label as the first node and that are healthy and idle as the second node, and configuring nodes in the same physical domain with the same cluster affiliation label; a forwarding module, used to synchronously forward the underwater acoustic signal stream input to the first node to the time-series cache queue of the second node, and load the underwater acoustic signal stream processing Pod image and configuration file of the first node on the second node; and a fault handling module, used to process the underwater acoustic signal stream in the time-series cache queue on the second node based on the pre-loaded underwater acoustic signal stream processing Pod image and configuration file when a fault is detected in the first node.

[0008] According to a third aspect, an embodiment of the present invention provides an electronic device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the steps of the container fault self-healing method of a containerized platform for an underwater acoustic signal processor as described in the first aspect or any embodiment of the first aspect.

[0009] According to a fourth aspect, embodiments of the present invention provide a computer storage medium storing computer instructions that, when executed by a processor, implement the steps of a container fault self-healing method for a containerized platform of an underwater acoustic signal processor as described in the first aspect or any embodiment of the first aspect.

[0010] This embodiment provides a container fault self-healing method for a containerized platform of underwater acoustic signal processors. By acquiring node hardware and service indicator parameters in real time to determine the probability of anomalies, healthy idle nodes with the same cluster affiliation label are pre-selected as backup nodes. The signal stream is forwarded synchronously and the processing environment is pre-loaded. In the event of a node failure, the system can quickly switch to the backup node to process the signal stream. Furthermore, by prioritizing the selection of nodes with the same cluster affiliation label, network latency and data packet loss caused by cross-domain transmission can be avoided. This effectively solves the stringent real-time requirements of underwater acoustic signal processing, avoids service interruptions caused by node failures, ensures the continuous and stable operation of key tasks such as underwater acoustic signal acquisition, filtering, and spectrum analysis, improves fault self-healing capabilities, guarantees real-time performance, and is suitable for complex application scenarios of mobile platforms such as ships and underwater unmanned vehicles.

[0011] Other advantages, objectives, and features of the invention will be set forth in the following description and will be apparent to those skilled in the art in some respects, or may be learned by practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0012] To make the objectives, technical solutions, and beneficial effects of this invention clearer, the following figures are provided for illustration:

[0013] Figure 1 This is a flowchart illustrating a specific example of a container fault self-healing method for a containerized platform of an underwater acoustic signal processor in this invention.

[0014] Figure 2 This is a schematic diagram of a module structure of a container fault self-healing device for a containerized platform of an underwater acoustic signal processor in this invention.

[0015] Figure 3 This is a schematic block diagram of a specific example of an electronic device in an embodiment of the present invention. Detailed Implementation

[0016] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can also refer to the internal connection of two components; and they can refer to a wireless connection or a wired connection. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0018] Furthermore, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0019] This invention provides a container fault self-healing method for a containerized platform of an underwater acoustic signal processor, such as... Figure 1 As shown, it includes:

[0020] S101, Obtain the hardware and service parameters of any first node during the underwater acoustic signal processing task.

[0021] S102, based on hardware indicator parameters and business indicator parameters, determine the anomaly probability of the first node;

[0022] S103, when the anomaly probability is higher than the warning threshold, the second node is determined based on the pre-built node label registry. Nodes with the same cluster affiliation label as the first node and that are healthy and idle are queried first as the second node. Nodes in the same physical domain are configured with the same cluster affiliation label.

[0023] S104, synchronously forward the underwater acoustic signal stream input to the first node to the time-sequential buffer queue of the second node, and load the underwater acoustic signal stream processing Pod image and configuration file of the first node on the second node;

[0024] S105: When a failure is detected in the first node, the second node processes the Pod image and configuration file based on the preloaded underwater acoustic signal stream, and processes the underwater acoustic signal stream in the time-sequential cache queue.

[0025] For example, by deploying a monitoring agent program on the first node, both hard metrics and business metrics are collected during the first node's completion of underwater acoustic signal processing tasks. The two types of metrics collected include CPU utilization, memory usage, hard disk read / write speed, network bandwidth utilization, temperature, etc.; and business metrics include underwater acoustic signal processing latency, data packet loss rate, task completion rate, etc.

[0026] Based on hardware and business metrics, the probability of anomalies in the first node can be determined by first standardizing the two types of metrics, then mapping each data point to the [0, 1] interval using a normalization formula, and then assigning weights to each metric parameter using a weighted summation algorithm. For example, CPU utilization 0.2, memory usage 0.2, hard disk read / write speed 0.1, network bandwidth utilization 0.1, temperature 0.1, processing latency 0.15, packet loss rate 0.1, and task completion rate 0.05. Thus, the probability of anomalies in the first node can be obtained.

[0027] When the anomaly probability exceeds the warning threshold, for example, 0.6, the node label registry query interface is invoked to prioritize nodes whose cluster affiliation label is the same as the first node and whose status is healthy and idle. If multiple nodes meet the criteria, nodes whose hardware type label perfectly matches the first node are initially selected as candidates. If no matching hardware type label is found, the node with the closest hardware performance is selected as the second node. The node label registry is stored in JSON format and includes node ID, hardware type label, cluster affiliation label, health status, and idle status. It should be noted that this embodiment introduces a cluster affiliation label, which is based on physical domain division. For example, ship deck area A and underwater hull area B belong to two different physical domains, and nodes located in the same ship deck area A are configured with the same label.

[0028] After the second node is determined, a dedicated data transmission channel is established between the first and second nodes. Within this channel, the underwater acoustic signal stream input to the first node can be synchronously forwarded to the second node in real time. The second node uses container orchestration tools, such as Kubernetes, to pull the underwater acoustic signal stream processing Pod image from the first node's private image repository. The image contains core algorithm modules such as signal acquisition, filtering, and spectrum analysis. Simultaneously, it obtains the first node's configuration file through a configuration center, which contains signal processing parameters such as the filtering cutoff frequency and the spectrum analysis window size. After loading, the Pod is in a ready state, waiting to receive processing tasks.

[0029] When a failure is detected in the first node, the second node processes the underwater acoustic signal stream in the time-series cache queue based on the pre-loaded underwater acoustic signal stream processing Pod image and configuration file. A failure in the first node can be identified as follows: the monitoring agent fails to send data to the time-series database for 10 consecutive seconds; or, the data packet loss rate in the business metrics suddenly rises above 50% for 5 seconds. Upon triggering a failure alarm, the second node immediately initiates the underwater acoustic signal stream processing procedure, calling the signal processing module in the pre-loaded Pod image, reading the signal stream data from the time-series cache queue, processing the data according to the same processing logic as the first node, and outputting the processing results to the downstream data receiving module in real time.

[0030] This invention provides a container fault self-healing method for a containerized platform of underwater acoustic signal processors. By acquiring node hardware and service indicator parameters in real time to determine the probability of anomalies, healthy idle nodes with the same cluster affiliation label are pre-selected as backup nodes. The signal stream is forwarded synchronously and the processing environment is pre-loaded. In the event of a node failure, the system can quickly switch to the backup node to process the signal stream. Furthermore, by prioritizing the selection of nodes with the same cluster affiliation label, network latency and data packet loss caused by cross-domain transmission can be avoided. This effectively solves the stringent real-time requirements of underwater acoustic signal processing, avoids service interruptions caused by node failures, and ensures the continuous and stable operation of critical tasks such as underwater acoustic signal acquisition, filtering, and spectrum analysis. It is suitable for complex application scenarios of mobile platforms such as ships and underwater unmanned vehicles.

[0031] As an optional implementation, synchronously forwarding the underwater acoustic signal stream input to the first node to the time-sequential buffer queue of the second node includes:

[0032] The input underwater acoustic signal stream is segmented according to the processing time sequence and labeled with a sequence tag to generate data segments with sequence tags. While storing the data segments in the second node's buffer queue, the checkpoint information of the first node's processing completion of each data segment is recorded and synchronized to the second node.

[0033] When a failure is detected in the first node, the second node processes the Pod image and configuration file based on the preloaded underwater acoustic signal stream, and processes the underwater acoustic signal stream in the time-series cache queue. This includes: when a failure is detected in the first node, the second node starts processing the data segment with the corresponding sequence tag in the time-series cache queue according to the latest checkpoint information received, and discards the data segment that has been completely processed by the first node.

[0034] For example, based on the smallest time unit of underwater acoustic signal processing, such as 20ms, the input underwater acoustic signal stream is segmented in time sequence. Each data segment corresponds to the signal collected within 20ms, forming a data segment. Then, a unique sequence label is generated for each data segment using the format of node ID-timestamp-segment number, such as Node001-20240520103000000-001, where Node001 is the first node ID, 20240520103000000 is the segment start timestamp accurate to milliseconds, and 001 is the segment number. The label is bound and stored with the corresponding data segment.

[0035] The first node updates its checkpoint information every time it completes 10% of the progress of a data segment, and synchronizes it to the second node in real time via the WebSocket protocol. Upon receiving the checkpoint information, the second node stores it in its local database, establishing a mapping relationship between sequence tags and checkpoint information for quick retrieval. The checkpoint information records the completion status of the first node's processing of each data segment, including the number of bytes processed, the current processing step, and the processing time, as well as the sequence tag and timestamp information.

[0036] When a failure is detected in the first node, the second node immediately queries its local database to obtain the latest checkpoint information related to the first node (selecting the record with the largest timestamp), and extracts the corresponding sequence tag and processing completion status. Based on the sequence tag in the latest checkpoint information, the second node locates the corresponding data segment in the time-series cache queue and continues processing from the unprocessed portion of that data segment; for data segments that have already been fully processed by the first node, they are directly deleted from the cache queue to avoid duplicate processing and improve processing efficiency.

[0037] This invention provides a container fault self-healing method for a containerized platform of underwater acoustic signal processor. The underwater acoustic signal stream is segmented and tagged with checkpoint information and in the event of a fault, the backup node can start processing from the corresponding data segment based on the latest checkpoint, while discarding the data segments that have been fully processed. This avoids the reprocessing of the signal stream, improves the processing efficiency after fault recovery, and accurately connects the signal processing processes before and after the fault, minimizing data loss and ensuring the temporal continuity and data integrity of underwater acoustic signal processing, further meeting the technical requirements for continuous underwater acoustic signal processing.

[0038] As an optional implementation, the second node is determined based on a pre-built node tag registry, including:

[0039] The system checks the node tag registry to determine if a target node exists that has the same hardware type tag and cluster affiliation tag as the first node and is healthy and idle. If multiple target nodes exist, they are evaluated based on a pre-built evaluation system to obtain an evaluation value for each target node. A preset number of candidate target nodes are selected based on the evaluation values. A benchmark test task is sent to each candidate target node. The benchmark test task is a key operation that simulates the first node's processing of underwater acoustic signal flow. The system receives the task completion data from each candidate target node and determines the second node from among the candidate target nodes based on the task completion data.

[0040] For example, the pre-built evaluation system may include hardware performance indicators, load status indicators, and historical stability indicators, as follows:

[0041] Hardware performance metrics include CPU clock speed, memory capacity, and hard drive read / write speed. CPU clock speed is scored out of 10 points: 3.0GHz and above earns 10 points, with 1 point deducted for every 0.2GHz decrease. Memory capacity is scored out of 10 points: 64GB and above earns 10 points, with 2 points deducted for every 16GB decrease. Hard drive read / write speed is scored out of 10 points: 500MB / s and above earns 10 points, with 2 points deducted for every 100MB / s decrease.

[0042] The load status indicators include CPU utilization and memory usage. CPU utilization is scored out of 10 points, with 10 points awarded for ≤30% and 2 points deducted for every 10% increase. Memory usage is scored out of 10 points, with 10 points awarded for ≤40% and 2 points deducted for every 10% increase.

[0043] Historical stability metrics include the fault-free operating time in the past 7 days and the fault recovery success rate. The fault-free operating time in the past 7 days is worth a maximum of 10 points, with 10 points awarded for 168 hours and 1 point deducted for every 24 hours less. The fault recovery success rate is also worth a maximum of 10 points, with 100% awarded for 10% and 1 point deducted for every 10% decrease.

[0044] The weights of hardware performance indicators, load status indicators, and historical stability indicators can be 0.4, 0.3, and 0.3, respectively. For each target node, the score of each indicator is calculated according to the above evaluation system. The scores of the indicators are multiplied by their corresponding weights and then summed to obtain the evaluation value of each target node. The top N target nodes in terms of evaluation value are selected as candidate target nodes, where N can be 3.

[0045] Next, benchmark tasks are sent to each candidate target node, recording the task start and completion times, and calculating the task completion time. Simultaneously, the CPU utilization, memory usage, and data processing accuracy of the nodes are monitored during task execution to generate task completion data. The benchmark task simulates key operations of the first node in processing aquatic signal flow, such as FFT spectral analysis. A 1024-point FFT operation is selected, with the input signal being a sine wave with a frequency of 1kHz and an amplitude of 1V, thereby generating a benchmark task package. This package should contain the input data, computational parameters, and the expected output format.

[0046] Finally, a comprehensive evaluation of the task completion data is conducted, prioritizing the candidate node with the shortest task completion time, lowest CPU usage, lowest memory consumption, and highest data processing accuracy as the second node. If multiple nodes have similar evaluation results, the final second node is determined by random selection.

[0047] This invention provides a container fault self-healing method for a containerized platform of underwater acoustic signal processors. The method accurately filters target nodes with matching hardware type, cluster affiliation, and healthy idle status through a node tag registry. The optimal backup node is selected by combining a multi-dimensional evaluation system and benchmark test tasks. This ensures that the backup node is highly compatible with the faulty node in terms of hardware performance, load status, historical stability, and key signal processing computing capabilities, thereby improving the overall reliability of the self-healing solution.

[0048] As an optional implementation, the anomaly probability of the first node is determined based on hardware and business performance parameters, including:

[0049] Based on hardware and business metrics, a pre-trained anomaly detection model is used to determine the anomaly probability of the first node itself; obtain the node propagation graph within the physical domain corresponding to the first node; determine the upstream nodes of the first node based on the node propagation graph; obtain the current state of each upstream node; determine the propagation anomaly probability of the first node based on the current state of each upstream node; and determine the anomaly probability of the first node based on its own anomaly probability and propagation anomaly probability.

[0050] For example, the anomaly detection model is constructed as follows: First, 10,000 sets of training data on hardware and business metrics in underwater acoustic scenarios are graded and labeled. Specifically, anomaly labels of 0 to 1 are assigned based on the degree of metric exceedance, alarm status, and business impact. For example, 0 is assigned to normal status, 0.2 to 0.5 to slightly exceed a single metric without alarms, 0.5 to 0.8 to exceed multiple metrics with minor alarms, and 0.8 to 1 to severely exceed core metrics with frequent alarms. Then, a lightweight CNN model adapted to this scenario is constructed. The input layer receives a 9-dimensional standardized metric vector, the hidden layer consists of two lightweight CNN convolutional layers, and the output layer uses one neuron with a sigmoid activation function. The loss function is replaced with mean squared error to achieve continuous mapping learning from the metric vector to the anomaly degree. The optimizer is Adam, and the model is trained for 100 epochs until the root mean square error (RMSE) of the test set is reached. Classification accuracy of 0.05 with a threshold of 0.6 95%; During the inference phase, the standardized first node index vector X is input into the trained model, and the model output layer directly outputs continuous values ​​in the interval [0,1] as its own anomaly probability. .

[0051] The directed weighted service dependency graph contains only all underwater acoustic signal processing nodes within the physical domain of the first node. This graph is a directed weighted acyclic graph, including nodes, directed edges, and weighted attributes. Nodes represent all underwater acoustic signal processing nodes within the physical domain. Directed edges represent direct data transmission / service dependencies between nodes, with arrows pointing from upstream to downstream nodes. Nodes without dependencies have no directed edges. The weighted attributes of each directed edge consist of three weighting parameters, all mapped to the [0,1] interval, including flow weight. Dependency type weight Historical fault propagation rate Among them, traffic weight Characterizes the flow rate of underwater acoustic signals transmitted from upstream nodes to the current node, as a proportion of the total inflow to the current node; dependent on type weights. Characterizing strong or weak dependencies where the upstream node is the first node; historical fault propagation rate. It represents the probability that an anomaly in the upstream node, as shown in historical data, will cause business anomalies in the current node.

[0052] The node propagation graph is stored in the central scheduling node of the physical domain in the form of an adjacency matrix. The element value in the matrix is ​​the weighted attribute set of the corresponding edge. The graph is updated in real time. When the node topology in the physical domain is adjusted, the traffic ratio changes, or the historical fault propagation rate is updated, the adjacency matrix is ​​corrected synchronously to ensure the accuracy of the graph.

[0053] By traversing the adjacency matrix of the node propagation graph, all valid upstream nodes of the first node are determined, and node hierarchies and attribute associations are completed to form a computable set of upstream nodes. Using the first node as the target node, the adjacency matrix of the node propagation graph within the physical domain is traversed to filter out all nodes with directed edges pointing to the first node (direct upstream) and nodes pointing to direct upstream nodes (indirect upstream), forming an initial upstream node list. This initial upstream node list is then filtered, removing nodes outside the physical domain, nodes marked as faulty, and nodes with no actual data transmission to the first node, retaining only valid upstream nodes.

[0054] Effective upstream nodes are classified according to the number of propagation hops, and core information extracted from the graph is associated with each upstream node, ultimately forming the upstream node set of the first node. ={ , ,..., The association information for each node is as follows: Includes node ID, hop count k, and traffic weight. Dependency type weight Historical fault propagation rate , where the hop count k represents the number of directed edges from the upstream node to the first node.

[0055] The propagation hop count classification rule is 1-hop upstream, 2-hop and above upstream; for example, the first node is the analysis node, and its upstream node set U includes 1-hop upstream (preprocessing node). k=1, =0.9, =1, =0.7), 2-hop upstream (collection node) k=2, =0.9, =1, =0.6), with no other valid upstream nodes.

[0056] Next, the multi-dimensional real-time status of each node in the upstream node set is collected in parallel, and the status standardization process is completed to form the status feature value of each upstream node.

[0057] Upstream node status collection dimensions: for each upstream node The current state is collected in four dimensions, specifically including:

[0058] Health / Abnormality Score , that is Its final abnormal probability is determined by The probability of its own abnormality and The propagation anomaly probability weighted fusion was obtained. ;

[0059] Performance index deviation , that is The deviation between the hardware and underwater acoustic service metrics and the normal threshold is mapped to [0,1] using Min-Max normalization. The closer it is to 1, the greater the performance deviation and the higher the risk of anomalies;

[0060] Alarm status It is a binary state. =1 means Currently triggering underwater acoustic service / hardware alarms (such as excessive packet loss rate, excessive CPU temperature). =0 indicates no alarms;

[0061] Change of status It is a binary state. =1 means There have been container deployments and configuration modifications in the past 24 hours. =0 indicates no changes.

[0062] The collected non-normalized state values ​​are processed using Min-Max normalization to ensure that the values ​​of all state dimensions are in the range [0,1], thus forming the final upstream node. Current state vector .

[0063] Then, through three sub-steps—single-link infectivity calculation, multi-link infectivity aggregation, and nonlinear correction—the probability of an upstream node abnormally infecting the first node is quantified and denoted as the propagation anomaly probability. , , specifically:

[0064] Step 1: Calculate the toxicity intensity of a single upstream node.

[0065] Toxicity intensity characterizes a single upstream node Its own abnormal propagation potential, that is The ability to spread, through The current state vector is calculated by weighting the vectors. The weights are set according to the importance of the underwater acoustic scene to ensure that they fit the actual business. Specifically:

[0066] ;

[0067] Where a, b, c, and d are weighting coefficients, and a+b+c+d=1.

[0068] Health / abnormality scoring by combining the characteristics of underwater acoustic signal processing It is the core element, with a weight of a=0.5; alarm status. It is a strong signal, with a weight c=0.3; performance index deviation value. It is an auxiliary signal with a weight of b=0.15; state change. It is a secondary signal with a weight d=0.05.

[0069] Calculated If the value exceeds the [0,1] interval, it will be automatically truncated to the interval boundary value. ∈[0,1], the closer the value is to 1, the more it represents The stronger the toxicity, the greater the potential for transmission.

[0070] Step 2: Calculate the upstream node Connection tightness with the first node

[0071] Connection tightness characterization The extent to which the anomaly can be transmitted to the first node, i.e. the strength of the dependency between the two, is calculated by fusing three weighted attributes extracted from the node propagation map, which fits the transmission characteristics of underwater acoustic signals.

[0072]

[0073] in, For traffic weight, For dependency type weights, The historical fault propagation rate is represented by values ​​in the [0,1] interval extracted from the graph. These three values ​​are multiplicative; if any attribute value is 0 (e.g., no traffic transmission w1=0), then... =0 means It has no actual dependency on the first node and no possibility of propagation; the closer the value is to 1, the stronger the connection between the two, and the easier it is for the exception to propagate.

[0074] Step 3: Calculate the path attenuation coefficient

[0075] The path decay coefficient characterizes the degree of attenuation of an anomaly's influence during propagation, and is only related to the upstream node. The propagation hop count k to the first node is related; the more hops, the more significant the attenuation.

[0076] 1. Jump upstream (direct upstream, k=1): =1, no decay, direct dependence on lossless anomaly propagation;

[0077] 2-Jump upstream (indirect upstream, k=2): =0.6, decay of 40%, the abnormal influence of indirect dependence is greatly reduced;

[0078] Upstream with 3 or more hops (k≥3): =0.3, attenuation 70%, anomalies in the far upstream layer have basically no significant impact on the first node;

[0079] Step 4: Calculate the single-link infectivity of a single upstream node.

[0080] Single-link infectivity characterizes a single upstream node The risk of a single anomalous propagation at the first node is a comprehensive reflection of toxicity intensity, connectivity tightness, and path decay coefficient, serving as the basic unit for subsequent aggregation.

[0081] = × × ;

[0082] Result constraints: ∈[0,1], the closer the value is to 1, the higher the risk of propagation from the upstream node to the first node; if =0 indicates that the node has no risk of propagation and can be removed from the aggregation.

[0083] Step 5: Aggregate and correct the infectivity of multiple links to obtain the probability of propagation anomalies.

[0084] The single-link infectivity of all upstream nodes is aggregated hierarchically, and then corrected using a nonlinear S-curve to finally obtain the propagation anomaly probability of the first node. .

[0085] First, the aggregation infectivity F is calculated by hierarchical aggregation. The hierarchy is based on the number of hops of the upstream nodes. The single-link infectivity with the same number of hops is aggregated first, and then aggregated sequentially from near to far (1 hop → 2 hops → 3 hops and above) according to the number of hops. The aggregation method is weighted summation, and the closer the number of hops, the higher the weight (which conforms to the characteristics of underwater sound propagation).

[0086] The aggregation formula is as follows:

[0087] ;

[0088] in The sum of the infectivity of the upstream nodes with one hop. The sum of the infectivity of the two upstream nodes. It is the sum of the infectivity of upstream nodes with 3 or more hops; , , For hierarchical weights, and + + =1, underwater sound scene setting =0.7、 =0.25、 =0.05;

[0089] Next, a nonlinear S-curve correction is performed. This correction serves two purposes: first, to achieve risk saturation; and second, to amplify key risks. The correction formula is as follows:

[0090] ,

[0091] The S-shaped curve has an inflection point at F=0.5, and when F<0.5... Slow growth, when F≥0.5 It improves rapidly, and the results naturally map to the [0,1] interval.

[0092] Finally, by weighted fusion, the anomaly probability of the first node itself is... With the probability of propagation anomalies By integrating the data, we obtain the final anomaly probability that comprehensively represents the anomaly risk of the first node. .

[0093] .

[0094] in The weight of its own abnormal probability. To propagate the weights of the anomaly probability, and + =1, It can be 0.6. It is 0.4.

[0095] This invention provides a container fault self-healing method for a containerized platform of underwater acoustic signal processors. It combines an anomaly detection model to determine the node's own anomaly probability, and simultaneously introduces node propagation graph analysis to analyze the impact of upstream node status on the current node. The method calculates the propagation anomaly probability and weights it to obtain the final anomaly probability. Compared to relying solely on node indicators to judge anomalies, this approach can more comprehensively and accurately assess node fault risks, identify potential faults caused by abnormal propagation from upstream nodes in advance, provide a more scientific decision-making basis for subsequent self-healing operations such as activating backup nodes, reduce misjudgments or omissions, and improve the accuracy of fault warnings.

[0096] As an optional implementation, a container fault self-healing method for a containerized platform of an underwater acoustic signal processor further includes:

[0097] The process involves acquiring the output results of the underwater acoustic signal stream from the time-series buffer queue processed by the second node, as well as the historical output results of the underwater acoustic signal stream processed by the first node before the failure occurred. First feature information representing continuous acoustic events is extracted from the output results of the second node. Second feature information representing continuous acoustic events is extracted from the historical output results of the first node. The first feature information and the second feature signal are subjected to time-series correlation analysis to obtain the continuity analysis result. If the continuity analysis result does not meet preset requirements, the process backtracks and locks the original data segment interval where the time-series break occurred. Based on preset rules, the process selects the original data segment interval for the third node to process, obtaining the processing result of the third node. The processing result of the third node is then correlated and fused with the output results of the underwater acoustic signal stream from the time-series buffer queue processed by the second node to obtain the underwater acoustic signal stream processing result.

[0098] For example, the second node processes the data into real-time underwater acoustic signal analysis data, which is stored in a distributed file system. The first node reads its historical output from a local backup database, also in JSON format, containing signal parameters corresponding to the second node's output. Based on the timestamps in the outputs, the second node's output is aligned with the first node's historical output in chronological order to ensure that the analysis focuses on continuous acoustic events within the same time period.

[0099] For the extraction of the first feature information, a signal processing algorithm is used to extract feature information representing continuous acoustic events from the output of the second node, including event duration, frequency variation trend, and amplitude fluctuation range. For the extraction of the second feature information, the same algorithm as for the first feature information is used to extract the corresponding second feature information from the historical output of the first node.

[0100] A dynamic time warping algorithm is used to calculate the similarity between the first and second feature information. The time series of the two feature information are aligned, the Euclidean distance between corresponding data points is calculated, the total distance is accumulated, and then normalization is applied to obtain the similarity. A similarity threshold of 0.8 is set. If the calculated similarity is greater than or equal to 0.8, the continuity analysis result meets the preset requirements, and the processing result of the second node is directly used as the final underwater acoustic signal flow processing result; if the similarity is less than 0.8, it is determined to be a temporal break, and subsequent data repair is required.

[0101] Based on the temporal correlation analysis results, the time interval with the lowest similarity (e.g., timestamp 20240520103005000-20240520103010000) is located. The corresponding original underwater acoustic signal data segment is then traced back to pinpoint the original data segment interval where the temporal break occurred. According to preset rules, such as prioritizing nodes with the same hardware type label as the first node, the highest historical processing accuracy, and the lowest current load, a third node is selected from the node label registry. If multiple nodes meet the criteria, the optimal node is selected through benchmark testing.

[0102] The third node loads the same Pod image and configuration file as the first node, processes the locked raw data segment range, and obtains the processing result. A data fusion algorithm is used to correlate and fuse the processing results of the third node with those of the second node to generate the final underwater acoustic signal stream processing result, ensuring data continuity and accuracy.

[0103] This invention provides a container fault self-healing method for a containerized platform of underwater acoustic signal processors. By extracting the feature information of continuous acoustic events in the output results of nodes before and after a fault and performing time-series correlation analysis, the method can promptly detect time-series breaks in signal processing after a fault switch. By selecting a third node to reprocess and fuse the results for the broken data segment interval, the method can repair the time-series breaks, ensure the continuity and accuracy of the final underwater acoustic signal stream processing results, and avoid target recognition accuracy reduction and underwater communication synchronization problems caused by time-series breaks.

[0104] As an optional implementation, synchronously forwarding the underwater acoustic signal stream input to the first node to the time-sequential buffer queue of the second node includes:

[0105] When the anomaly probability is lower than the target threshold but higher than the warning threshold, the underwater acoustic signal stream forwarded to the time-series buffer queue by the second node is a signal stream encoded with a lossy coding method using a high compression ratio; when the anomaly probability is higher than the target threshold, the underwater acoustic signal stream forwarded to the time-series buffer queue by the second node is a signal stream encoded with a lossless coding method.

[0106] For example, based on statistics from a large amount of historical fault data, when the anomaly probability... At a value of 0.8, the risk of node failure is extremely high, and data integrity must be guaranteed; at 0.6... abnormal probability At a threshold of 0.8, the risk of failure is moderate, and data can be appropriately compressed to save bandwidth. Therefore, the target threshold is set to 0.8, while the warning threshold remains at 0.6. Based on the relationship between the anomaly probability of the first node and these two thresholds, the corresponding encoding method is selected. A lossy encoding method with a high compression ratio can be the MP3 encoding algorithm, with a compression ratio set to 12:1; a lossless encoding method can be the FLAC encoding algorithm.

[0107] This invention provides a container fault self-healing method for a containerized platform of underwater acoustic signal processors. It dynamically selects the signal stream encoding method based on the node anomaly probability. When the anomaly probability is moderate, high-compression lossy encoding is used to save bandwidth resources; when the anomaly probability is extremely high, lossless encoding is used to ensure data integrity, achieving a balanced optimization between bandwidth resources and data integrity. In scenarios with varying node failure risks, it can both rationally utilize transmission resources and ensure that critical data is not lost based on the risk level, improving resource utilization efficiency and data reliability during the synchronous forwarding of underwater acoustic signal streams.

[0108] As an optional implementation, a container fault self-healing method for a containerized platform of an underwater acoustic signal processor further includes:

[0109] Analyze the hardware logs, container runtime logs, and signal processing business logs of the first node to locate the root cause of the fault. If the root cause is a hardware fault, mark the first node as pending maintenance and send a hardware maintenance alarm to the target port. If the root cause is a software fault, redeploy the Pod image and configuration files on the first node.

[0110] For example, if the analysis determines that the root cause of the fault is a hardware failure, such as the appearance of keywords like "hard disk badsector" or "CPU overheating shutdown" in the logs, the system automatically marks the first node as "pending maintenance" (updates the node status field in the node label registry) and generates a hardware maintenance alarm message, which is sent to the target port, which can be the corresponding port of the administrator's office computer. If the analysis determines that the root cause of the fault is a software failure, such as the appearance of keywords like "Pod crash loop back - off," "algorithm execution error," or "configuration fileerror" in the logs, the system first stops the faulty Pod on the first node and cleans up any remaining processes and files.

[0111] This invention provides a container fault self-healing method for a containerized platform of underwater acoustic signal processor. By analyzing multiple types of logs, the root cause of the fault is accurately located. Hardware faults are marked as requiring repair and alarms are sent to facilitate timely hardware maintenance. For software faults, Pod images and configuration files are redeployed to achieve rapid software-level repair.

[0112] This embodiment provides a container fault self-healing device for a containerized platform of an underwater acoustic signal processor, such as... Figure 2 As shown, it includes:

[0113] The parameter acquisition module 201 is used to acquire the hardware indicator parameters and business indicator parameters of any first node during the process of completing the underwater acoustic signal processing task.

[0114] The probability determination module 202 is used to determine the anomaly probability of the first node based on hardware indicator parameters and business indicator parameters.

[0115] The node determination module 203 is used to determine the second node based on the pre-built node label registry when the anomaly probability is higher than the warning threshold. It prioritizes querying nodes with the same cluster affiliation label as the first node and that are healthy and idle as the second node. Nodes in the same physical domain are configured with the same cluster affiliation label.

[0116] The forwarding module 204 is used to synchronously forward the underwater acoustic signal stream input from the first node to the time-sequential buffer queue of the second node, and load the underwater acoustic signal stream processing Pod image and configuration file of the first node on the second node.

[0117] The fault handling module 205 is used to process the Pod image and configuration file based on the preloaded underwater acoustic signal stream on the second node when a fault is detected in the first node, and to process the underwater acoustic signal stream in the time-sequential cache queue.

[0118] This application also provides an electronic device, such as... Figure 3 As shown, processor 501 and memory 502 are connected via a bus or other means.

[0119] Processor 501 can be a central processing unit (CPU). Processor 501 can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or combinations of the above types of chips.

[0120] The memory 502, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the container fault self-healing method of a containerized platform for an underwater acoustic signal processor in this embodiment of the invention. The processor executes various functional applications and data processing by running the non-transitory software programs, instructions, and modules stored in the memory.

[0121] Memory 502 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created by the processor, etc. Furthermore, the memory may include high-speed random access memory and non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory 502 may optionally include memory remotely located relative to the processor, which can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0122] The one or more modules are stored in the memory 502, and when executed by the processor 501, they perform actions such as... Figure 1 The embodiment shown illustrates a container fault self-healing method for a containerized platform of an underwater acoustic signal processor.

[0123] For specific details regarding the aforementioned electronic devices, please refer to the relevant documentation. Figure 1 The relevant descriptions and effects in the illustrated embodiments are for understanding purposes only and will not be repeated here.

[0124] This embodiment also provides a computer storage medium storing computer-executable instructions that can execute a container fault self-healing method for a containerized platform of an underwater acoustic signal processor in any of the above method embodiments. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk drive (HDD), or solid-state drive (SSD), etc.; the storage medium may also include combinations of the above types of memory.

[0125] Finally, it should be noted that the above preferred embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail through the above preferred embodiments, those skilled in the art should understand that various changes can be made to it in form and detail without departing from the scope defined by the claims of the present invention.

Claims

1. A container fault self-healing method for a containerized platform of an underwater acoustic signal processor, characterized in that, include: Obtain the hardware and service parameters of any first node during the underwater acoustic signal processing task. Based on hardware and business metrics, the probability of anomalies in the first node is determined. When the probability of an anomaly is higher than the warning threshold, the second node is determined based on the pre-built node label registry. Nodes with the same cluster affiliation label as the first node and that are healthy and idle are queried first as the second node. Nodes in the same physical domain are configured with the same cluster affiliation label. The underwater acoustic signal stream input to the first node is synchronously forwarded to the time-sequential buffer queue of the second node, and the underwater acoustic signal stream processing Pod image and configuration file of the first node are loaded on the second node. When a failure is detected in the first node, the second node processes the Pod image and configuration file based on the preloaded underwater acoustic signal stream, and processes the underwater acoustic signal stream in the time-sequenced cache queue. Obtain the output results of the underwater acoustic signal stream in the time-sequential buffer queue of the second node and the historical output results of the underwater acoustic signal stream before the failure occurred in the first node; Extract the first feature information representing continuous acoustic events from the output of the second node; Extract the second feature information representing continuous acoustic events from the historical output of the first node; The first feature information and the second feature signal are subjected to time-series correlation analysis to obtain the continuity analysis results; If the continuity analysis results do not meet the preset requirements, the original data segment interval where the time series break occurred is traced back and locked. According to the preset rules, the third node is selected to process the original data segment interval and the processing result of the third node is obtained. The processing result of the third node is correlated and fused with the output result of the underwater acoustic signal stream in the time-sequential buffer queue of the second node to obtain the underwater acoustic signal stream processing result.

2. The container fault self-healing method for a containerized platform of an underwater acoustic signal processor according to claim 1, characterized in that, The underwater acoustic signal stream input to the first node is synchronously forwarded to the time-sequential buffer queue of the second node, including: The input underwater acoustic signal stream is segmented according to the processing time sequence and labeled with sequence tags to generate data segments with sequence tags. While storing the data segments into the second node's cache queue, the checkpoint information of the first node's processing completion of each data segment is recorded, and the checkpoint information is synchronized to the second node. When a failure is detected in the first node, the second node processes the Pod image and configuration file based on the preloaded underwater acoustic signal stream, and processes the underwater acoustic signal stream in the time-series cache queue, including: When a failure is detected in the first node, the second node starts processing the data segment with the corresponding sequence tag in the time-series cache queue based on the latest checkpoint information received, and discards the data segment that has been fully processed by the first node.

3. The container fault self-healing method for a containerized platform of an underwater acoustic signal processor according to claim 1, characterized in that, The second node is determined based on a pre-built node label registry, including: Determine in the node label registry whether there exists a target node that is healthy and idle, with the same hardware type label and cluster affiliation label as the first node. When there are multiple target nodes, the multiple target nodes are evaluated based on the pre-built evaluation system to obtain the evaluation value of each target node; A preset number of candidate target nodes are selected based on the evaluation values; A benchmark task is sent to each candidate target node. The benchmark task is to simulate the key operations of the first node in processing the underwater acoustic signal flow. Receive task completion data from each candidate target node, and determine the second node from among the candidate target nodes based on the task completion data.

4. The container fault self-healing method for a containerized platform of an underwater acoustic signal processor according to claim 1, characterized in that, Based on hardware and business metrics, the probability of anomalies in the first node is determined, including: Based on hardware and business metrics, a pre-trained anomaly detection model is used to determine the anomaly probability of the first node. Obtain the node propagation graph within the physical domain corresponding to the first node; Based on the node propagation graph, determine the upstream node of the first node; Get the current state of each upstream node; Based on the current state of each upstream node, determine the probability of propagation anomaly in the first node; The probability of anomaly of the first node is determined based on its own anomaly probability and the probability of propagation anomaly.

5. The container fault self-healing method for a containerized platform of an underwater acoustic signal processor according to claim 1, characterized in that, The underwater acoustic signal stream input to the first node is synchronously forwarded to the time-sequential buffer queue of the second node, including: When the anomaly probability is lower than the target threshold but higher than the warning threshold, the underwater acoustic signal stream forwarded to the second node to the time-sequenced buffer queue is a signal stream encoded with a lossy coding method using a high compression ratio; When the anomaly probability is higher than the target threshold, the underwater acoustic signal stream forwarded by the second node to the time-sequenced buffer queue is a signal stream encoded using a lossless encoding method.

6. A container fault self-healing method for a containerized platform of an underwater acoustic signal processor according to any one of claims 1-5, characterized in that, Also includes: Analyze the hardware logs, container runtime logs, and signal processing service logs of the first node to locate the root cause of the fault; If the root cause of the fault is a hardware failure, the first node is marked as pending maintenance and a hardware maintenance alarm is sent to the target port. If the root cause of the failure is a software fault, then redeploy the Pod image and configuration files on the first node.

7. A container fault self-healing device for a containerized platform of an underwater acoustic signal processor, characterized in that, include: The parameter acquisition module is used to acquire the hardware and business performance parameters of any first node during the underwater acoustic signal processing task. The probability determination module is used to determine the anomaly probability of the first node based on hardware indicator parameters and business indicator parameters. The node determination module is used to determine the second node based on a pre-built node label registry when the anomaly probability is higher than the warning threshold. It prioritizes querying nodes with the same cluster affiliation label as the first node and that are healthy and idle as the second node. Nodes in the same physical domain are configured with the same cluster affiliation label. The forwarding module is used to synchronously forward the underwater acoustic signal stream input from the first node to the time-sequential buffer queue of the second node, and load the underwater acoustic signal stream processing Pod image and configuration file of the first node on the second node. The fault handling module is used to process the Pod image and configuration file based on the preloaded underwater acoustic signal stream on the second node when a fault is detected in the first node, and to process the underwater acoustic signal stream in the time-sequenced cache queue. Obtain the output results of the underwater acoustic signal stream in the time-sequential buffer queue of the second node and the historical output results of the underwater acoustic signal stream before the failure occurred in the first node; Extract the first feature information representing continuous acoustic events from the output of the second node; Extract the second feature information representing continuous acoustic events from the historical output of the first node; The first feature information and the second feature signal are subjected to time-series correlation analysis to obtain the continuity analysis results; If the continuity analysis results do not meet the preset requirements, the original data segment interval where the time series break occurred is traced back and locked. According to the preset rules, the third node is selected to process the original data segment interval and the processing result of the third node is obtained. The processing result of the third node is correlated and fused with the output result of the underwater acoustic signal stream in the time-sequential buffer queue of the second node to obtain the underwater acoustic signal stream processing result.

8. An electronic device, the device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor performs the steps of the container fault self-healing method of a containerized platform for an underwater acoustic signal processor as described in any one of claims 1-6.

9. A computer storage medium storing computer instructions thereon, characterized in that, When executed by the processor, this instruction implements the steps of the container fault self-healing method of the containerized platform of the underwater acoustic signal processor as described in any one of claims 1-6.

Images