Storage cluster arbitration method and device, electronic equipment and storage medium

Abstract

The invention discloses a storage cluster arbitration method and device, electronic equipment and a storage medium, and relates to the technical field of storage. Future operation data of a corresponding dimension factor and the future health degree of a storage node are predicted through operation data of a multi-dimension factor of each storage node; meanwhile, the current weight of the corresponding dimension factor is calculated according to the future operation data of each dimension factor, so that the current health degree of the corresponding storage node is obtained through calculation, and finally, a plurality of storage sites and a plurality of storage nodes of the storage cluster are arbitrated based on the current health degree of each storage node. According to the method, factors such as storage performance and data synchronization state are comprehensively considered, and through dynamic weight distribution and collaborative analysis, the accuracy and robustness of fault judgment of the storage system are improved, and the technical problems that related technologies are easily influenced by network fluctuation and arbitration judgment conditions cannot serve the performance of a cluster are solved; the technical effects of improving the arbitration stability, improving the service performance of the cluster and the like are achieved.

Description

Technical Field

[0001] The present application relates to the field of storage technology, and in particular to a storage cluster arbitration method, device, electronic device, and storage medium. Background Art

[0002] In an active-active cluster, two storage sites provide services simultaneously and ensure data consistency through data synchronization. However, a network interruption or node failure between the two storage sites can cause a "split-brain" phenomenon, or the inability to find the cluster configuration node, leading to cluster failure.

[0003] To address this issue, related technologies have introduced arbitration mechanisms to ensure that only one site continues to process write requests in the event of a network failure, maintaining cluster availability and data consistency. For example, IP arbitration deploys the IP arbitration file on a third server, allowing the storage to periodically send heartbeat signals to the arbitration server. If the original primary site stops sending heartbeat signals, a specific node selection mechanism is used to select the remaining node that can send heartbeat signals as the cluster's configuration node, ensuring cluster survival and business continuity.

[0004] Although relevant technologies effectively prevent split-brain, they are prone to misjudgment when network jitter or transient failures occur. In addition, they fail to fully consider factors such as storage performance and data synchronization status, making it difficult to optimize the business performance of the entire cluster. Summary of the Invention

[0005] The present application provides a storage cluster arbitration method, apparatus, electronic device, and storage medium to at least solve the technical problems in related technologies that are easily affected by network fluctuations and arbitration determination conditions cannot serve the performance of the cluster.

[0006] The present application provides a storage cluster arbitration method, wherein the storage cluster includes multiple storage sites, and the storage sites include multiple storage nodes, wherein the method includes: obtaining operating data of multi-dimensional factors of each storage node in each storage site; predicting future operating data of the corresponding dimensional factors and the future health of the storage nodes based on the operating data; calculating the current weight of the corresponding dimensional factors based on the future operating data of each dimensional factor; calculating the current health of the corresponding storage node based on the operating data, current weight and future health of each dimensional factor, and arbitrating the multiple storage sites and multiple storage nodes of the storage cluster based on the current health of each storage node.

[0007] The present application also provides a storage cluster arbitration device, including: an acquisition module for acquiring the operating data of the multi-dimensional factors of each storage node in each storage site; a prediction module for predicting the future operating data of the corresponding dimensional factors and the future health of the storage nodes based on the operating data; a calculation module for calculating the current weight of the corresponding dimensional factor based on the future operating data of each dimensional factor; an arbitration module for calculating the current health of the corresponding storage node based on the operating data, current weight and future health of each dimensional factor, and arbitrating multiple storage sites and multiple storage nodes of the storage cluster based on the current health of each storage node.

[0008] The present application also provides an electronic device, comprising: a memory for storing a computer program; and a processor for implementing the steps of any of the above-mentioned storage cluster arbitration methods when executing the computer program.

[0009] The present application also provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above storage cluster arbitration methods are implemented.

[0010] Through this application, since the future operating data of the multi-dimensional factors of each storage node can be used to predict the future operating data of the corresponding dimensional factors and the future health of the storage node, and the current weight of the corresponding dimensional factor is calculated based on the future operating data of each dimensional factor, the current health of the corresponding storage node is calculated. Finally, based on the current health of each storage node, arbitration is performed on multiple storage sites and multiple storage nodes of the storage cluster, and storage performance, data synchronization status and other factors are comprehensively considered. Through dynamic weight allocation and collaborative analysis, the accuracy and robustness of storage system fault judgment are improved, and the technical problems that related technologies are easily affected by network fluctuations and arbitration judgment conditions cannot serve the performance of the cluster are solved, achieving technical effects such as improving arbitration stability and improving the business performance of the cluster. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0012] Figure 1 A schematic diagram of a storage cluster arbitration method according to an embodiment of the present application;

[0013] Figure 2 A flowchart of an active-active cluster arbitration method based on multi-dimensional factors provided according to an embodiment of the present application;

[0014] Figure 3 A schematic diagram of the structure of a storage cluster arbitration device provided according to an embodiment of the present application;

[0015] Figure 4 A schematic diagram of the structure of an electronic device provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0016] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0017] It should be noted that, in the description of this application, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. The terms "first," "second," etc., in this application are used to distinguish similar objects, and are not used to describe a particular order or sequence.

[0018] To address the problem of "split brain" caused by network connection interruption or node failure between two storage sites, or the inability to find cluster configuration nodes, resulting in cluster failure, Related Technology 1 creates an arbitration volume on the storage side, maps the arbitration volume to the host, and monitors the communication between the host and storage sides through the arbitration driver module on the host side. By creating an arbitration volume on the storage side and mapping it to the host side, the host side configures the arbitration policy, and uses the arbitration volume as a medium to achieve autonomous arbitration of the storage system. However, this method is still limited to the judgment conditions of a single dimension; Related Technology 2 receives arbitration requests including arbitration information from each storage site in the active-active storage site when the storage loses connection; determines the target storage site for providing subsequent services based on the arbitration information in the arbitration requests received from each storage site; and controls the target storage site to provide subsequent services based on the determined target storage site. This aims to improve the stability of the active-active storage system in providing host and storage services, so as to effectively avoid host service interruptions. This method improves stability but also fails to consider multi-dimensional factors. Related Technology 3: When the first data center determines that communication with the second data center is disconnected and the arbitration device cannot arbitrate, the first data center checks the pre-set arbitration policy. If the arbitration policy determines that the first data center is the priority data center, service continues. If the arbitration policy determines that the first data center is not the priority data center, service is discontinued. This arbitration method utilizes a single-dimensional judgment criterion, is susceptible to network fluctuations, and fails to comprehensively consider factors such as storage performance and data synchronization status.

[0019] In summary, the above-mentioned related technologies use a single-dimensional judgment condition, which may be affected by factors such as network jitter and latency, and is prone to failure or misjudgment. Or, factors such as storage performance, data synchronization status, and business priority are not considered. Nodes with poor performance may be judged as healthy nodes and continue to provide write data services, causing business lag or even data loss.

[0020] In response to the above-mentioned deficiencies in the related technologies, embodiments of the present application provide a storage cluster arbitration method, apparatus, electronic device, and storage medium, which will be described in detail below.

[0021] In order to enable those skilled in the art to better understand the present application, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0022] Specifically, Figure 1 A flowchart of a storage cluster arbitration method provided by an embodiment of the present application. The storage cluster includes multiple storage sites, and the storage site includes multiple storage nodes, such as Figure 1 As shown, the method comprises the following steps:

[0023] In step S101 , operation data of multi-dimensional factors of each storage node in each storage site is obtained.

[0024] Among them, the multi-dimensional factors include network layer factors, storage layer factors and environmental factors. The operating data of each multi-dimensional factor are network layer data, storage layer data and environmental layer data respectively. The network layer data includes heartbeat signal strength and frequency, network delay, jitter, packet loss rate, and bandwidth utilization between sites. The storage layer data includes RAID (Redundant Array of Independent Disks) initialization progress, RAID reconstruction and copyback progress, SSD (Solid State Drive) wear, cache hit rate, storage system IOPS (Input / Output Operations Per Second), bandwidth, latency, and the copy rate and progress of remote replication between active-active sites. The environmental layer data includes storage node temperature, storage real-time PDU (Power Distribution Unit) current, and real-time UPS (Uninterruptible Power Supply) load rate.

[0025] It can be understood that the embodiment of the present application first needs to obtain the operating data of the network layer factors, storage layer factors and environmental factors of each storage node. These operating data serve as the basis for storage cluster arbitration, so that storage cluster arbitration can comprehensively consider factors such as network status, storage performance, and environment, making decisions more accurate.

[0026] In step S102 , future operating data of the corresponding dimension factor and the future health of the storage node are predicted based on the operating data.

[0027] It is understandable that after the embodiment of the present application obtains the operating data of the multi-dimensional factors of each storage node from the above steps, the future operating data of the corresponding dimensional factors and the future health of the storage node can be predicted based on these operating data. The specific method of prediction is as follows:

[0028] In an embodiment of the present application, predicting future operating data of corresponding dimensional factors and future health of storage nodes based on operating data includes inputting the operating data into a neural network model, and the neural network model outputting the future operating data of corresponding dimensional factors and future health of storage nodes. Before inputting the operating data into the neural network model, the method also includes: obtaining historical operating data of the storage cluster; generating a data set and real labels based on the historical operating data; and training the neural network model based on the data set and real labels.

[0029] Among them, the neural network model can be LSTM (Long Short-Term Memory), which can learn long-term dependent information and has great advantages in processing and predicting data with time components; the historical operation data is the historical operation data of multi-dimensional factors and the historical operation data of storage node health, and it is necessary to ensure that the collected historical operation data has a sufficient historical span to capture the changing trends of different dimensional factors and storage node health; the data set refers to a series of input data used to train the neural network model, and the true label refers to the known result or target value corresponding to each set of input data. The neural network model is trained using the data set, and the true label guides the model to learn the correct output, so that a neural network model that meets the requirements can be obtained.

[0030] It can be understood that in order to correspond to the future operating status of the dimensional factors and the health of the storage nodes, the embodiments of the present application first need to collect the historical operating data of the multi-dimensional factors and the historical operating data of the health of the storage nodes, and construct a training data set and corresponding real labels based on these historical data. The data set is used to describe the input features, and the real labels are used to represent the corresponding actual output results. The two are used to train the neural network model. After the training is completed, the current operating data is input into the model, and the model outputs the future operating trend of the corresponding dimensional factors and the future health of the storage nodes.

[0031] In step S103 , the current weight of each dimensional factor is calculated based on the future operation data of the corresponding dimensional factor.

[0032] Among them, the current weight of the corresponding dimensional factor is calculated based on the future operating data of each dimensional factor, and is dynamically calculated and allocated through a machine learning model. For example, for the storage layer factor, the future operating data indicator changes of the storage layer factor are predicted. If the predicted storage delay will exceed the set threshold, such as 10ms, the storage weight is temporarily increased, for example, to 0.4; for example, when a certain dimensional indicator changes by more than 50% within 100ms, the weight degradation is triggered. For example, when the network packet loss rate suddenly increases, the network weight is reduced, for example, to 0.1. It should be noted that the basic weight of each dimensional factor is usually pre-set based on operation and maintenance experience or historical data analysis as an initial reference value, and is not specifically limited here.

[0033] It can be understood that in order to more accurately evaluate the impact of each dimensional factor in the storage system on the overall health status, the embodiment of the present application can dynamically calculate and assign the current weight of each dimensional factor through a machine learning model. Specifically, the embodiment of the present application can dynamically adjust the weight value in the comprehensive evaluation based on the future operating data of each dimensional factor predicted according to the above steps, combined with its change trend and degree of abnormality. For example, in the storage layer, if it is predicted that the storage delay will exceed the set threshold in the future, the weight of the dimension will be temporarily increased to reflect its potential impact on performance; if a dimensional factor fluctuates greatly in a short period of time, such as the network packet loss rate increases by more than 50% within 100ms, it is considered that its stability has decreased. At this time, the weight degradation mechanism can be triggered to lower its weight. Through this dynamic weight adjustment mechanism, the changing trend of the real-time status of the storage cluster can be more finely reflected.

[0034] In step S104, the current health of the corresponding storage node is calculated according to the operating data, current weight and future health of each dimensional factor, and multiple storage sites and multiple storage nodes of the storage cluster are arbitrated based on the current health of each storage node.

[0035] It is understandable that the embodiment of the present application can calculate the current health of the corresponding storage node based on the operating data, current weight and future health of each dimensional factor obtained in the above steps. The health comprehensively considers factors such as network status, storage performance, and environment, and is more accurate and reliable. Based on the current health of each storage node, arbitration is performed on multiple storage sites and multiple storage nodes of the storage cluster, so that the arbitration judgment conditions serve the performance of the cluster and improve the accuracy and robustness of storage system fault judgment. Specifically, the method for calculating the current health of the storage node is as follows:

[0036] In an embodiment of the present application, the current health of the corresponding storage node is calculated based on the operating data of each dimensional factor, the current weight of the storage node, and the future health of the storage node, including: normalizing the operating data of each dimensional factor; calculating the score of the corresponding dimensional factor based on the normalized operating data; calculating the current health of the corresponding storage node based on the score of each dimensional factor, the current weight, and the future health of the storage node.

[0037] Among them, the normalization process uses the maximum and minimum normalization method to map the original data to the [0,1] interval. The maximum and minimum normalization method formula is: where x norm is the output value after normalization, x is the original data that needs to be normalized; x min is the minimum value of the dimension factor in the selected data set; x maxIt is the maximum value of the dimension factor in the selected data set, wherein the selected data set may be 70% of the data of the dimension factor in the historical operation data. It is set according to the actual situation and is not specifically limited here.

[0038] It can be understood that the embodiment of the present application performs standardization on the operating data of each dimensional factor. Specifically, the mean, standard deviation, minimum value, and maximum value of the operating data of each dimensional factor in the selected data set are calculated, and the maximum and minimum normalization method is used to convert the operating data of each dimensional factor into a numerical value within a unified range according to its historical maximum and minimum values, thereby eliminating the dimensional differences between different indicators. Based on the normalized data, the score corresponding to each dimensional factor is calculated. The score reflects the health level of the factor in the current state. Finally, the score of each dimensional factor, the dynamically adjusted current weight obtained in the above steps, and the future health of the storage node obtained by model prediction are combined to obtain the current health of the storage node through comprehensive calculation, thereby realizing the fusion analysis of the multi-dimensional data of the storage system and improving the accuracy and adaptability of the health assessment. Specifically, the method for calculating the current health of the storage node is as follows:

[0039] In this embodiment of the present application, the calculation formula for the current health of the storage node is:

[0040]

[0041] Among them, W i (t) is the current weight of the i-th dimension factor; S i (t) is the score of the i-th dimension factor; S i_max is the maximum value of the score of the i-th dimension factor in the selected data set; λ is the prediction factor adjustment coefficient; LSTM_predict(H(T+1)) is the future health of the storage node, which is predicted by the LSTM neural network model through the above steps; H is the current health of the storage node; T is the time point for calculating the health; (T+1) is the next time step starting from the current time point T. The length of each time step can be set according to actual needs, such as 3 seconds, 5 seconds, etc., which is not specifically limited here; n is the total number of dimension factors.

[0042] In an embodiment of the present application, arbitration is performed on multiple storage sites and multiple storage nodes of a storage cluster based on the current health of each storage node, including: obtaining the arbitration priority of the storage site; arbitrating based on the arbitration priority and the current health of each storage node; and controlling the storage cluster to perform a switching action according to the arbitration result, wherein the switching action includes at least one of switching between storage sites and switching between configuration nodes.

[0043] Among them, the arbitration priority of the storage site is to first determine whether there is a problem with the environmental factors, then determine whether site switching is required based on the health formula, and finally determine whether configuration node switching is required based on the health formula; the configuration node is the node responsible for managing and maintaining cluster configuration information, used to store and distribute cluster configuration data, and monitor the status of each node in the cluster in real time.

[0044] It can be understood that the embodiment of the present application arbitrates multiple storage sites and multiple storage nodes of the storage cluster based on the current health of each storage node. First, it is necessary to judge the environmental factors according to the arbitration priority. If the environment is high temperature or power outage, the health is directly set to 0, and the storage site is switched directly; if the environmental factors are not a problem, the health formula is used to determine whether site switching and node switching are required. By directly judging the health as 0 when the environment is abnormal, such as high temperature or power outage, and triggering site switching first, it is possible to quickly respond to serious physical layer failures, avoid service interruptions due to hardware damage or environmental problems, and thus ensure business continuity.

[0045] In an embodiment of the present application, the storage cluster is controlled to perform a switching action according to the arbitration result, including: if the arbitration result is a site switch, then control switching between storage sites; if the arbitration result is a node switch, then control switching configuration nodes.

[0046] Among them, switching between storage sites, that is, when the health status of a site is not suitable for continuing to carry critical loads, the workload is migrated to another site with better health status; switching configuration nodes, that is, when a problem occurs in the current configuration node, another storage node with good health status is selected to take over its responsibilities to ensure that functions such as configuration management and status synchronization are not affected.

[0047] It can be understood that the embodiment of the present application controls the storage cluster to execute the switching action according to the arbitration priority and the arbitration result of the health formula to confirm whether to perform site switching or node switching. For example, if the environment is high temperature or power outage, the health is directly set to 0, the arbitration result is site switching, and the storage site is directly controlled to switch; if there is no problem with the environmental factors, but the network delay is greater than 200 milliseconds and the storage IOPS is less than 1000, it is determined to be a joint network and storage failure, then the arbitration result is site switching, the storage site is controlled to switch, and at the same time, the configuration node is switched. If the network delay is less than 50 milliseconds and the storage IOPS is less than 1000, it means that there is no problem with the network but the performance is low. It is determined to be a single point failure of the storage and there is no need to switch the storage site. The arbitration result is node switching, and the switching configuration node is controlled.

[0048] In an embodiment of the present application, controlling switching between storage sites includes: if the number of the multiple storage sites is two, switching the current storage site to the remaining storage site; if the number of the multiple storage sites is greater than two, calculating the comprehensive health of the storage site based on the current health of the storage nodes of the remaining storage sites, selecting a target storage site from the remaining storage sites based on the comprehensive health, and switching the current storage site to the target storage site.

[0049] The comprehensive health can be obtained by taking a weighted average or summing the health of all storage nodes in the site to reflect the overall operating status of the entire site.

[0050] It can be understood that in the embodiment of the present application, when controlling the switching between storage sites, if there are only two storage sites in the cluster, when switching is required, the currently running storage site is directly switched to the other remaining site; when the number of storage sites exceeds two, it is necessary to comprehensively evaluate the current health of each storage node in the remaining sites, calculate the comprehensive health score of each remaining storage site, and select the target storage site with the best health status as the switching object based on this, thereby completing the switching operation from the current storage site to the target storage site.

[0051] In an embodiment of the present application, controlling switching between configuration nodes includes: selecting a storage node with the highest current health as a target configuration node; and switching the current configuration node to the target configuration node.

[0052] It can be understood that the embodiment of the present application selects the storage node with the highest current health, which means that the storage node has the best operating status and the most stable performance at the current moment, and is suitable as a new configuration node to ensure the reliable operation of the cluster. Therefore, when switching the configuration node, the storage node with the highest current health is selected as the new target configuration node to switch the configuration node.

[0053] According to the storage cluster arbitration method provided in the embodiment of the present application, the future operating data of the corresponding dimensional factors and the future health of the storage node can be predicted through the operating data of the multi-dimensional factors of each storage node. At the same time, the current weight of the corresponding dimensional factor is calculated based on the future operating data of each dimensional factor, thereby calculating the current health of the corresponding storage node. Finally, based on the current health of each storage node, arbitration is performed on multiple storage sites and multiple storage nodes of the storage cluster, comprehensively considering factors such as storage performance and data synchronization status, and through dynamic weight allocation and collaborative analysis, the accuracy and robustness of storage system fault judgment are improved, thereby achieving technical effects such as improving arbitration stability and improving the business performance of the cluster.

[0054] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus the necessary general hardware platform, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method.

[0055] The storage cluster arbitration method is further described below through a specific embodiment.

[0056] In order to solve the problem that single IP arbitration is affected by the network and cannot better serve the performance of the active-active cluster, this embodiment designs the following Figure 2 A dual-active cluster arbitration method based on multi-dimensional factors is shown.

[0057] Step S201: Obtain the operating data of each dimensional factor.

[0058] The technical core of this embodiment lies in collecting data on multi-dimensional factors, dynamically assigning weights to each factor through a machine learning model, arbitrating clusters based on multi-dimensional factors, and selecting the cluster's business sites and configuration nodes. Therefore, the operational data of the multi-dimensional factors must first be collected. Specifically:

[0059] Network layer factors require collecting information on the heartbeat signal strength and frequency of each node, network latency, jitter, packet loss rate, and bandwidth utilization between sites. Storage layer factors require collecting information on RAID initialization progress, RAID reconstruction and copyback progress, SSD wear, cache hit rate, storage system IOPS, bandwidth, and latency, and the copy rate and progress of remote replication between active-active sites. Environmental layer factors require collecting storage node temperature information, connecting to the data center's power supply system, and obtaining real-time data on storage PDU current and UPS load.

[0060] Step S202: Calculate the parameters of the running data of each dimensional factor and normalize them.

[0061] Because the operating data of various dimensional factors varies, the following uses the IOPS data type of the storage layer factor as an example to describe how to process this data and obtain a normalized score for the storage layer factor: The IOPS data for the past 70% of the storage layer factor is obtained as a training set to obtain the mean, standard deviation, minimum, and maximum values of the storage layer factor IOPS. Then, when real-time data is obtained, the real-time storage layer factor data is normalized using the previously calculated normalization parameters, such as the minimum and maximum values, using the maximum-minimum normalization method. Finally, a normalized score for the storage layer factor is obtained.

[0062] Step S203: LSTM trains the dynamic weights of factors at each layer.

[0063] Before calculating the dynamic weight, it is necessary to assign initial values to the weights of each dimensional factor. The initial values are pre-set based on operation and maintenance experience or historical data analysis.

[0064] After assigning initial values to each dimension factor, the LSTM algorithm is used to predict changes in the metrics over the next three seconds. For example, if the predicted storage latency exceeds a threshold (such as 10ms), the storage weight is temporarily increased to 0.4. If a dimension metric changes by more than 50% within 100ms, a weight reduction is triggered (for example, if the network packet loss rate suddenly increases, the network weight is reduced to 0.1).

[0065] Step S204: Calculate the health of each node according to the formula.

[0066] After calculating the weights of each factor, the health H of each node is calculated according to the following formula:

[0067]

[0068] Where W i (t) represents the dynamic weight of the i-th dimension, S i (t) represents the real-time score of the i-th dimension (normalized to 0-1), λ represents the prediction factor adjustment coefficient, which defaults to 0.2, and LSTM_predict(H(T+1)) represents the health predicted by the LSTM model three seconds in the future. If the ambient temperature exceeds 50°C or the node loses power, the health level H is reset to 0. If the storage layer latency exceeds 200ms, the node is considered to have poor performance (H<0.3).

[0069] Step S205: Determine whether the environmental factor of each storage node is lower than a threshold.

[0070] Perform hard judgment on the environment. Set an environmental factor judgment threshold. If the environmental factor is lower than the threshold, it indicates that the storage node is faulty and proceeds to step S207; if it is higher than the threshold, proceed to step S206.

[0071] Step S206: Determine whether the health level is lower than a threshold.

[0072] Perform health determination. Set a health determination threshold. If the health determination threshold is lower than this threshold, it means that the health of the storage node is too low, and proceed to step S207; if it is higher than this threshold, no change is made and the process ends.

[0073] Step S207: Find the healthiest node as the configuration node based on the health level.

[0074] Find the node with the highest health among the remaining nodes and set the healthiest node as the configuration node of the cluster.

[0075] In summary, this embodiment introduces a method for active-active cluster arbitration using multiple factors, making the arbitration of the active-active cluster more robust and reasonable, improving the service performance of the cluster, avoiding the risk of arbitration failure of a single IP, and enabling timely transfer when the performance of the configuration node is poor, thereby ensuring service stability.

[0076] The embodiment of the present application further provides a storage cluster arbitration device, Figure 3 A schematic diagram of the structure of the storage cluster arbitration device 10 provided in an embodiment of the present application is shown in FIG. Figure 3 As shown, the device includes: an acquisition module 301 , a prediction module 302 , a calculation module 303 and an arbitration module 304 .

[0077] Among them, the acquisition module 301 is used to obtain the operating data of the multi-dimensional factors of each storage node in each storage site; the prediction module 302 is used to predict the future operating data of the corresponding dimensional factors and the future health of the storage nodes based on the operating data; the calculation module 303 is used to calculate the current weight of the corresponding dimensional factor based on the future operating data of each dimensional factor; the arbitration module 304 is used to calculate the current health of the corresponding storage node based on the operating data, current weight and future health of each dimensional factor, and arbitrate multiple storage sites and multiple storage nodes of the storage cluster based on the current health of each storage node.

[0078] In an embodiment of the present application, the arbitration module 304 is further used to: normalize the operating data of each dimensional factor; calculate the score of the corresponding dimensional factor based on the normalized operating data; calculate the current health of the corresponding storage node based on the score of each dimensional factor, the current weight and the future health of the storage node.

[0079] In this embodiment of the present application, the calculation formula for the current health of the storage node is:

[0080]

[0081] Among them, W i (t) is the current weight of the i-th dimension factor; S i (t) is the score of the i-th dimension factor; S i_max is the maximum value of the score of the i-th dimension factor in the selected data set; λ is the prediction factor adjustment coefficient; LSTM_predic(H(T+1)) is the future health of the storage node; H is the current health of the storage node; T is the time point for calculating the health; (T+1) is the next time step starting from the current time point T; n is the total number of dimension factors.

[0082] In an embodiment of the present application, the arbitration module 304 is further used to: obtain the arbitration priority of the storage site; perform arbitration based on the arbitration priority and the current health of each storage node; and control the storage cluster to perform a switching action based on the arbitration result, wherein the switching action includes at least one of switching between storage sites and switching between configuration nodes.

[0083] In an embodiment of the present application, the storage cluster is controlled to perform a switching action according to the arbitration result, including: if the arbitration result is a site switch, then control switching between storage sites; if the arbitration result is a node switch, then control switching configuration nodes.

[0084] In an embodiment of the present application, controlling switching between storage sites includes: if the number of the multiple storage sites is two, switching the current storage site to the remaining storage site; if the number of the multiple storage sites is greater than two, calculating the comprehensive health of the storage site based on the current health of the storage nodes of the remaining storage sites, selecting a target storage site from the remaining storage sites based on the comprehensive health, and switching the current storage site to the target storage site.

[0085] In an embodiment of the present application, controlling switching between configuration nodes includes: selecting a storage node with the highest current health as a target configuration node; and switching the current configuration node to the target configuration node.

[0086] For descriptions of features in the embodiments corresponding to the storage cluster arbitration device, reference may be made to the relevant descriptions of the embodiments corresponding to the storage cluster arbitration method, which will not be detailed here.

[0087] According to the storage cluster arbitration device provided in the embodiment of the present application, the future operating data of the corresponding dimensional factors and the future health of the storage node can be predicted through the operating data of the multi-dimensional factors of each storage node. At the same time, the current weight of the corresponding dimensional factor is calculated based on the future operating data of each dimensional factor, thereby calculating the current health of the corresponding storage node. Finally, based on the current health of each storage node, arbitration is performed on multiple storage sites and multiple storage nodes of the storage cluster, comprehensively considering factors such as storage performance and data synchronization status, and through dynamic weight allocation and collaborative analysis, the accuracy and robustness of storage system fault judgment are improved, thereby achieving technical effects such as improving arbitration stability and improving the business performance of the cluster.

[0088] The embodiment of the present application also provides an electronic device, such as Figure 4 As shown, it includes a memory 401 and a processor 402. The memory 401 stores a computer program, and the processor 402 is configured to run the computer program to execute the steps in any of the above storage cluster arbitration method embodiments.

[0089] An embodiment of the present application further provides a computer-readable storage medium, in which a computer program is stored. The computer program is configured to execute the steps of any of the above storage cluster arbitration method embodiments when running.

[0090] In an exemplary embodiment, the computer-readable storage medium may include, but is not limited to, various media that can store computer programs, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk, or an optical disk.

[0091] An embodiment of the present application further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the steps of any of the above storage cluster arbitration method embodiments are implemented.

[0092] An embodiment of the present application further provides another computer program product, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of any of the above storage cluster arbitration method embodiments are implemented.

[0093] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the components and steps of each example according to their functions. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0094] The above is a detailed introduction to the storage cluster arbitration method, device, electronic device, and storage medium provided by the present application. This article uses specific examples to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only intended to help understand the method and core ideas of the present application. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of the present application, several improvements and modifications can be made to the present application, and these improvements and modifications also fall within the scope of protection of the claims of the present application.

Claims

1. A storage cluster arbitration method, characterized in that: The storage cluster includes a plurality of storage sites, each of which includes a plurality of storage nodes, wherein the method includes: Obtain operational data on multiple dimensions of each storage node in each storage site; Predicting future operating data of corresponding dimensional factors and future health of storage nodes based on the operating data; Calculate the current weight of the corresponding dimension factor based on the future operating data of each dimension factor; The current health of the corresponding storage node is calculated according to the operating data, current weight and future health of each dimensional factor, and the multiple storage sites and multiple storage nodes of the storage cluster are arbitrated based on the current health of each storage node.

2. The storage cluster arbitration method according to claim 1, wherein: The current health of the corresponding storage node is calculated based on the operating data of each dimension factor, the current weight of the storage node, and the future health of the storage node, including: Normalizing the operating data of each dimensional factor; Calculate the scores of the corresponding dimension factors based on the normalized operating data; The current health of the corresponding storage node is calculated based on the score of each dimensional factor, the current weight and the future health of the storage node.

3. The storage cluster arbitration method according to claim 2, wherein: The calculation formula for the current health of the storage node is: Among them, W i (t) is the current weight of the i-th dimension factor; S i (t) is the score of the i-th dimension factor; S i_max is the maximum value of the score of the i-th dimension factor in the selected data set; λ is the prediction factor adjustment coefficient; LSTM_predict(H(T+1)) is the future health of the storage node; H is the current health of the storage node; T is the time point for calculating the health; (T+1) is the next time step starting from the current time point T; n is the total number of dimension factors.

4. The storage cluster arbitration method according to claim 1, wherein: The arbitrating the multiple storage sites and multiple storage nodes of the storage cluster based on the current health of each storage node includes: Obtaining the arbitration priority of the storage site; performing arbitration based on the arbitration priority and the current health of each storage node; The storage cluster is controlled to perform a switching action according to the arbitration result, wherein the switching action includes at least one of switching between storage sites and switching between configuration nodes.

5. The storage cluster arbitration method according to claim 4, wherein: The controlling the storage cluster to perform a switching action according to the arbitration result includes: If the arbitration result is site switching, controlling the switching between storage sites; If the arbitration result is node switching, the control switches the configuration node.

6. The storage cluster arbitration method according to claim 5, wherein: The control of switching between storage sites includes: If the number of multiple storage sites is two, the current storage site is switched to the remaining storage site; If the number of multiple storage sites is greater than two, the comprehensive health of the storage sites is calculated based on the current health of the storage nodes of the remaining storage sites, and the target storage site is selected from the remaining storage sites based on the comprehensive health, and the current storage site is switched to the target storage site.

7. The storage cluster arbitration method according to claim 5, wherein: The control configuration switching between nodes includes: Select the storage node with the highest current health as the target configuration node; Switch the current configuration node to the target configuration node.

8. A storage cluster arbitration device, characterized in that: include: An acquisition module, used to obtain the operating data of the multi-dimensional factors of each storage node in each storage site; A prediction module, configured to predict future operation data of corresponding dimensional factors and future health of storage nodes based on the operation data; A calculation module is used to calculate the current weight of each dimensional factor based on the future operation data of each dimensional factor; An arbitration module is used to calculate the current health of the corresponding storage node based on the operating data, current weight and future health of each dimensional factor, and arbitrate multiple storage sites and multiple storage nodes of the storage cluster based on the current health of each storage node.

9. An electronic device, characterized in that: include: memory for storing computer programs; A processor, configured to implement the steps of the storage cluster arbitration method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, the steps of the storage cluster arbitration method according to any one of claims 1 to 7 are implemented.

Images