WO2020135889A1

WO2020135889A1 - Method for dynamic loading of disk and cloud storage system

Info

Publication number: WO2020135889A1
Application number: PCT/CN2019/130169
Authority: WO
Inventors: 黄华东; 夏伟强; 王伟; 林起芊
Original assignee: 杭州海康威视系统技术有限公司
Priority date: 2018-12-28
Filing date: 2019-12-30
Publication date: 2020-07-02
Also published as: CN111381766B; CN111381766A

Abstract

A method for dynamic loading of a disk and a cloud storage system. The method is applied to a cloud storage system. The cloud storage system comprises a management node and multiple storage nodes, and the multiple storage nodes access the same SAS switch. The method comprises: when detecting that a software failure occurs in a first storage node, the management node sends a disk loading instruction to a second storage node (S1); the second storage node loads the disk of the first storage node by means of the SAS switch after receiving the disk loading instruction (S2); the management node updates locally stored storage node information corresponding to the disk (S3). According to the method, the use of the SAS switch enables the storage nodes to access the disks of all the storage nodes on the switch, thereby enabling the disk of a failed storage node to be loaded by means of other storage nodes, implementing the dynamic loading of the disk, reducing the performance loss of system reconstruction, and improving the usability of object storage disks.

Description

Dynamic disk loading method and cloud storage system

This application requires the priority of the Chinese patent application filed on December 28, 2018 with the application number 201811625675.X and the invention titled "A method for dynamic disk loading and cloud storage system", the entire content of which is incorporated by reference in In this application.

Technical field

The present application relates to the field of data storage technology, in particular to a method for dynamically loading a disk and a cloud storage system.

Background technique

With the development of society, safety has increasingly become the focus of people's attention. The promotion of safe cities and other projects provides a certain guarantee for our safe life. Among them, security monitoring plays a vital role in safe cities and other projects. There are massive video data in security monitoring. Cloud storage provides flexible storage space for the storage of massive video data. The storage space of cloud storage needs to maintain the storage cluster, and the data is generally scattered in the storage cluster. That is to say, massive video data can be stored through the storage cluster.

Cloud storage can use copy mode or EC (Erasure Code, erasure code) mode to ensure data integrity. In a storage cluster, after a device fails, the data in the failed storage node needs to be recovered through the copy or EC data. That is refactoring. When the storage cluster size of cloud storage is large, storage node failures will become frequent. Especially when a part of the storage node failure is a software failure, such as service startup failure, operating system abnormality, etc., although the data in the failed storage node can be calculated through copy or EC data, it consumes the computing power of the storage cluster and increases Cluster burden.

Summary of the invention

Embodiments of the present application provide a method for dynamically loading a disk and a cloud storage system, which can reduce system resource consumption caused by data reconstruction. The technical solution is as follows:

In one aspect, a method for dynamically loading a disk is provided, which is applied to a cloud storage system. The cloud storage system includes a management node and multiple storage nodes. The multiple storage nodes access the same SAS switch. The method includes:

When the management node detects that the first storage node of the plurality of storage nodes has a software failure, it sends a disk load instruction to the second storage node of the plurality of storage nodes;

After receiving the disk loading instruction, the second storage node loads the disk of the first storage node through the SAS switch.

In a possible implementation manner of the present application, the management node updates storage node information corresponding to the locally stored disk.

In a possible implementation manner of the present application, the method further includes:

When the management node receives the read request to read the data of the disk, the management node sends the read request to the second storage node according to the updated information of the storage node corresponding to the locally stored disk.

The second storage node reads the data in the disk through the SAS switch according to the received read request.

In a possible implementation manner of the present application, when the management node receives a write request to write data to the disk, according to the updated locally stored information of the storage node corresponding to the disk, the management node will write The request is sent to the second storage node;

The second storage node writes data to the disk through the SAS switch according to the received write request.

In a possible implementation manner of the present application, loading the disk of the first storage node through the SAS switch includes:

The second storage node updates the index information of the disk in the first storage node to the database of the second storage node.

In a possible implementation manner of the present application, the management node updating the storage node information corresponding to the locally stored disk includes:

The management node correspondingly updates the storage node information of the disk and the second storage node to a local database.

In a possible implementation manner of the present application, before the management node updates the storage node information corresponding to the locally stored disk, the method further includes:

The management node receives the message that the disk is loaded successfully from the second storage node.

On the other hand, a cloud storage system is also provided. The cloud storage system includes: a management node and multiple storage nodes, the multiple storage nodes accessing the same SAS switch, and the multiple storage nodes include a first storage Node and second storage node, where:

The management node is configured to send a disk load instruction to the second storage node when a software failure is detected in the first storage node;

The second storage node is configured to load the disk of the first storage node through the SAS switch after receiving the disk loading instruction.

In a possible implementation manner of the present application, the management node is further used to update storage node information corresponding to the locally stored disk.

In a possible implementation manner of the present application, the management node is further configured to, when receiving a read request to read the data of the disk, according to the updated storage node information corresponding to the disk stored locally , And send the read request to the second storage node;

The second storage node is also used to read the data in the disk through the SAS switch according to the received read request.

In a possible implementation manner of the present application, the management node is further configured to, when receiving a write request to write data to the disk, update the locally stored storage node information corresponding to the disk , Send the write request to the second storage node;

The second storage node is further configured to write data to the disk through the SAS switch according to the received write request.

In a possible implementation manner of the present application, the second storage node is also used to update the index information of the disk in the first storage node to the database of the second storage node.

In a possible implementation manner of the present application, the management node is further configured to correspondingly update the storage node information of the disk and the second storage node to a local database.

In a possible implementation manner of the present application, the management node is further configured to receive a message that the second storage node successfully loads the disk.

On the other hand, a disk dynamic loading device is proposed, including:

One or more processors, a storage device that stores one or more programs;

When the one or more programs are executed by the one or more processors, the one or more processors implement the disk dynamic loading method.

On the other hand, a computer-readable storage medium is also proposed, on which a computer program is stored, which implements the disk dynamic loading method when the computer program is executed by a processor.

On the other hand, a computer program product containing instructions is provided, which, when run on a computer, causes the computer to implement the disk dynamic loading method described in the above aspect.

The method for dynamically loading the disk of the present application accesses the same SAS switch through the storage node, and the storage node can access the disks of all storage nodes connected to the SAS switch, so that in the case of a storage node software failure, loading through other storage nodes The disk of the failed storage node can realize the dynamic loading of the disk, reduce the performance loss of system reconstruction, and improve the availability of the object storage disk.

BRIEF DESCRIPTION

By reading the detailed description of alternative embodiments below, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing alternative embodiments, and are not considered as limitations to the present application. Furthermore, throughout the drawings, the same reference symbols are used to denote the same components. In the drawings:

FIG. 1 shows a first overall flowchart of a method for dynamically loading a magnetic disk according to an embodiment of the present application.

FIG. 2 shows a schematic structural diagram of a storage node accessing a SAS switch according to an embodiment of the present application.

FIG. 3 shows a second overall flowchart of a method for dynamically loading a disk according to an embodiment of the present application.

FIG. 4 shows a first schematic flowchart of an MDS drift disk according to an embodiment of the present application.

FIG. 5 shows a second schematic flowchart of an MDS drift disk according to an embodiment of the present application.

FIG. 6 shows a third schematic flow chart of an MDS drift disk according to an embodiment of the present application.

detailed description

Hereinafter, exemplary embodiments of the present disclosure will be described in more detail with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

In this application, the following terms are defined as follows:

Database: Database (Data, Base, DB) refers to a collection of related, structured data that is reasonably stored on a computer's storage device. A database contains various contents, including tables, views, fields, indexes, etc.

Video positioning: This application refers to the time entered by the user. The system can quickly find the stored video data corresponding to this time according to the relevant information recorded in the database.

Byte: Data storage is in the unit of "Byte" (Byte), and every 8 bits (bit, abbreviated as b) form a byte (Byte, abbreviated as B), which is the smallest level of information unit.

Video stream: includes the video data to be transmitted, which can be processed as a stable and continuous stream through the network.

Object storage: The object storage system is a massive, safe, highly reliable and easily expandable cloud storage service provided to users. Instead of organizing files into a directory hierarchy, it stores files in a flat container organization and uses unique IDs to retrieve them. The result is that object storage systems require less metadata to store and access files than file systems, and they also reduce the overhead of managing file metadata due to storing metadata. The object storage system provides services for users through the platform-independent RESTFUL protocol and supports convenient storage and management of massive objects through the web. The object storage system can store arbitrary objects in a durable and highly available system. Applications and users can use simple APIs (Application Programming Interface) to access data in the object storage; these are usually based on the state of table attributes Transfer (REST) architecture, but there are also interfaces for programming languages.

OSD (Object-Based Storage Device): This solution represents a storage node and is a module for reading and writing objects in an object storage system. The OSD stores data to the tracks and sectors of the disk, and combines several tracks and sectors to form an object, and provides access to the data to the outside world through this object.

MDS (Metadata Server): It is the management node in the object storage system, which stores the index information of the object, including the name of the object, the specific location information of the object, and the last modification time of the object.

Allocation of resources: This solution refers to MDS allocating storage resources for object writing, and specifically refers to allocating OSD and object disks.

File object: Responsible for file access operations. After obtaining the file object, you can use the file object to read the data on the disk. The file object is uploaded to the cloud storage by the user at a time, and the upload is completed in one interaction using the PUT protocol.

Cluster technology: A cluster is a group of independent computers interconnected by a high-speed network. They form a group and are managed as a single system. When a client interacts with the cluster, the cluster acts as an independent server. The cluster configuration is used to improve availability and scalability.

Disk loading: Cloud storage persists data to multiple disks, which are media for storing data in cloud storage. Each disk usually includes multiple partitions. In the Linux operating system, disk loading refers to mounting the disk of a device (usually a storage device) to an existing directory. Specifically, if you want to access files in a disk of a storage device, you must Mount the partition where the file is located on an existing directory, and then access the file by accessing this directory. The disk can only be read and written after it is loaded by cloud storage.

Disk drift: The disk drifts between OSDs, which means that the read and write control of the disk is switched from one OSD to another OSD.

Reconstruction: The process of recovering damaged data blocks can be calculated through valid data blocks and check blocks in EC data.

SAS (Serial Attached SCSI (Small Computer System Interface, small computer system interface) serial connection SCSI) switch: a switch that uses the SAS protocol for disk discovery and simulated network communication. After a storage node is connected to a SAS switch, you can discover and use disks in all storage nodes connected to the switch.

FIG. 1 is a schematic flowchart of a method for dynamically loading a magnetic disk provided by this embodiment, and each step is described in detail below. The method for dynamically loading a magnetic disk may be applied to a cloud storage system. The cloud storage system includes a management node and multiple storage nodes, and the multiple storage nodes are connected to the same SAS switch.

Before introducing this method, first introduce the structural framework of multiple storage nodes provided by this application to access the same SAS switch, please refer to FIG. 2, in an embodiment of this application, the cloud storage system may include multiple Management nodes, the multiple management nodes form a management cluster, as shown in Figure 2, the signaling ports of each management node MDS1, MDS2, MDS3...MDSN in the management cluster are interconnected with ordinary Gigabit switches, through the mutual The interconnection realizes signaling exchange. Similarly, the multiple storage nodes form a storage cluster, and each storage node OSD1, OSD2, 0SD3...OSDN signaling ports in the storage cluster are interconnected with ordinary Gigabit switches, and the signaling exchange is realized through the interconnection between each other. At the same time, the data ports of each storage node OSD1, OSD2, 0SD3...OSDN of the storage cluster are interconnected through SAS switches, and data exchange between them is realized through interconnection.

Next, the cloud storage system includes a management node MDS as an example for description. The signaling exchange between the management node MDS and the ordinary gigabit switch is a two-way exchange. The signaling can be transmitted bidirectionally between the management node MDS and the ordinary gigabit switch; the signaling exchange between the storage node OSD and the ordinary gigabit switch is Bidirectional exchange, signaling can be bidirectionally transmitted between the storage node OSD and the ordinary gigabit switch; the data exchange between the storage node OSD and the SAS switch is also a bidirectional exchange, and the data can be bidirectionally transferred between the storage node OSD and the SAS switch.

Because SAS switches use the SAS protocol for disk discovery and simulated network communication, when a storage node is connected to the SAS switch, it can discover and use the disks in all storage nodes connected to the switch. By connecting the storage node OSD in the cloud storage system to the SAS switch, the storage node OSD can access the disks of other storage nodes connected to the SAS switch.

Specifically, as shown in FIG. 1, the method for dynamically loading a disk provided by this application may include the following steps:

S1. When the management node detects that the first storage node has a software failure, it sends a disk load instruction to the second storage node.

Suppose that a storage node has a software level failure, such as a failed service start or an abnormal operating system. For convenience of description, this storage node is referred to as a faulty storage node. Here, the faulty storage node is also referred to as a first storage node. . After the storage node has a software failure, the failed storage node cannot report the heartbeat to the management node MDS. The management node MDS considers the failed storage node offline. At this time, the management node MDS requests other storage nodes connected to the SAS switch to try to load the failed storage node. For the convenience of description, the other storage node is referred to as a second storage node, that is, the management node sends a disk load instruction to the second storage node to instruct the second storage node to load the first storage node. Disk.

S2. After receiving the disk loading instruction, the second storage node loads the disk of the first storage node through a SAS switch.

After receiving the disk loading instruction, the second storage node loads the disk in the first storage node. After the second storage node is successfully loaded, the data in the disk in the first storage node can be normally read by the second storage node, of course, the data in the disk of the first storage node can also be written by the second storage node , Thereby avoiding the process of data recovery.

In a possible implementation manner of the present application, the second storage node updates the index information of the disk in the first storage node to the database of the second storage node.

In this application, the disk index information in the first storage node can be sent to the second storage node through the SAS switch, and the second storage node copies the disk index information in the first storage node to the database of the local node for The purpose of the update is to use this disk index information to read the data in the disk in the first storage node that has a software failure in the future.

That is, the management node MDS can dynamically adjust the disks in the failed storage node to other storage nodes for read, write, and load according to the status of the storage node. For example, if the management node MDS does not find the storage node abnormal, it reads and writes the data in the disk normally; and when the management node MDS finds a storage node abnormal, it requests another storage node in the same switch to load the disk of the failed storage node , Through the other storage node to normally read and write data in the disk of the failed storage node, to achieve disk drift.

It can be seen from the above process that the management node MDS realizes disk drift according to the status of the storage node, that is, when a storage node has a software failure, the read and write permissions of the disk drift from the failed storage node to the normal storage node in the storage cluster. After the disk drifts, all disk read and write requests are executed through the normal storage node. The normal storage node uses the drifted disk like a local disk. Therefore, the normal storage node can access the disk in the failed storage node normally through the SAS switch, and the disk in the failed storage node can be normally loaded. In this way, the data in the disk of the failed storage node can still be read and written normally without using copy mode or EC mode for recovery.

S3. The management node updates locally stored storage node information corresponding to the disk.

The disk is a disk in the first storage node, that is, a disk that drifts to the second storage node.

In a possible implementation manner of the present application, the management node updates the disk and storage node information of the second storage node to the local database. Among them, the storage node information is used to uniquely indicate a storage node.

Further, before the management node updates the storage node information corresponding to the locally stored disk, the method may further include: the management node receives a message that the second storage node successfully loads the disk. That is to say, after determining that the second storage node successfully loads the disk in the first storage node, the management node updates the above-mentioned disk and the storage node information of the second storage node to the local database.

For example, after the second storage node successfully loads the disk in the first storage node, it will send a corresponding message that the disk is successfully loaded to the management node. After receiving the message that the disk is loaded successfully, the management node updates the information of the disk and the storage node of the second storage node to the local database of the management node as a record, so that if the first storage node fails again next time When the disk needs to be loaded again, it is not necessary to search or find a new storage node to load the disk, and a second storage node can be directly assigned to load the disk.

After the above steps, after the software level of the storage node is abnormal, the disk in the faulty storage node with software level abnormality can be successfully loaded and read and written by other storage nodes. Data read and write does not need to be reconstructed to restore data, avoiding Unnecessary calculations. Moreover, after the storage node is abnormal, the reading and writing of data in the entire cloud storage system will not have much performance impact.

After the failed storage node returns to normal, the management node MDS may request the second storage node to unload the loaded disk. For example, after the failed storage node returns to normal, the management node MDS may first request the second storage node to unload the disk of the loaded failed storage node, and then request the failed storage node to load the disk, so that the local disk of the failed storage node can be replaced by the The failed storage node takes over itself, thereby dispersing the pressure on the storage disk of the storage node in the system.

In this way, this application uses disks to drift between storage nodes inside the object storage to realize dynamic loading of the disks. After a software failure occurs on the storage nodes, the storage nodes can be drifted through SAS switches to continue to access the disks of the failed storage nodes Data to improve disk availability.

In a possible implementation manner of the present application, as shown in FIG. 3, the disk dynamic loading method may further include:

S4. When the management node receives the read request to read the data of the disk, the management node sends the read request to the second storage node according to the updated storage node information corresponding to the locally stored disk.

As an example, the second storage node reads the data in the disk through the SAS switch according to the received read request.

As an example, when the management node receives a write request to write data to the disk, it sends the write request to the second according to the updated storage node information corresponding to the disk stored locally. Storage node

Therefore, the disks of the failed storage node can be read and written normally after being loaded by other storage nodes. After the disk of the failed storage node is successfully loaded by the normal storage node, subsequent disk data reading and writing can be performed through the normal storage node of the loaded disk. The SAS switch allows the storage node to access the disks of other storage nodes in the same switch, just like The same as accessing the local disk.

Correspondingly, this application also proposes a cloud storage system. The cloud storage system includes: a management node and multiple storage nodes, the multiple storage nodes accessing the same SAS switch, and the multiple storage nodes include a first storage Node and second storage node, where:

As an example, the management node is also used to update storage node information corresponding to the locally stored disk.

As an example, the management node is further configured to, when receiving a read request to read the data of the disk, update the read request according to the updated storage node information corresponding to the disk stored locally Sent to the second storage node;

As an example, the second storage node is further configured to read the data in the disk through the SAS switch according to the received read request.

As an example, the management node is further configured to, when receiving a write request to write data to the disk, update the write request according to the updated storage node information corresponding to the disk stored locally Sent to the second storage node;

As an example, the second storage node is also used to update the index information of the disk in the first storage node to the database of the second storage node.

As an example, the management node is also used to update the disk and the storage node information of the second storage node to the local database.

As an example, the management node is further configured to receive a message that the second storage node successfully loads the disk.

As shown in FIG. 4, for ease of understanding, a specific example is used to explain the specific implementation of the management node MDS provided in this embodiment to request other storage nodes in the switch to load the disk of the failed storage node, that is, the management node MDS implements the drift disk Steps can include:

A1. The storage node OSD1 is abnormal.

Suppose that the software level of the storage node OSD1 fails, such as service startup failure, abnormal operating system, etc. In this case, the disk and the data on the disk are normal, and the disk can still be accessed.

A2. The management node MDS requests the storage node OSD2 to load the disk of the storage node OSD1.

After a software failure occurs on the storage node OSD1, the storage node OSD1 cannot report the heartbeat to the management node MDS. The management node MDS considers that the storage node OSD1 is offline. At this time, the management node MDS requests the other storage node OSD2 to try to load the storage node OSD1 disk. After the storage node OSD2 is successfully loaded, the disk data in the storage node OSD1 can be normally read by other storage nodes OSD2. Of course, the other storage node OSD2 can also write data to the disk, thereby avoiding the data recovery process.

Specifically, according to the state of the storage node, the management node MDS can dynamically adjust the disk to other storage node OSD2 for reading, writing, and loading. For example, if the management node MDS finds that the storage node is abnormal, it reads and writes the disk data normally; and when the management node MDS finds that the storage node OSD1 is abnormal, it requests the storage node OSD2 in the same switch to load the disk of the storage node OSD1, through the storage node OSD2 normally reads and writes the disk data of storage node OSD1 to realize disk drift.

It can be seen from the above process that the management node MDS implements disk drift according to the state of the storage node. When a software failure occurs on the storage node OSD1, the read and write permissions of the disk drift from the faulty storage node OSD1 to the normal storage node OSD2 in the storage cluster.

A3. The storage node OSD2 successfully loads the disk of OSD1.

After the disk drifts, all disk read and write requests are made through the normal storage node OSD2. The storage node OSD2 uses the drifted disk like a local disk.

Therefore, the storage node OSD2 can normally access the disk in OSD1 through the SAS switch, and the disk in OSD1 can be normally loaded.

The disk of the faulty storage node OSD1 can be read and written normally after being loaded by other storage nodes OSD2. After the disk of the faulty storage node OSD1 is successfully loaded by the normal storage node OSD2, the subsequent reading and writing of the disk data can be performed by the storage node OSD2 loading the disk. The SAS switch allows the storage node OSD2 to access other storage nodes OSD1 in the same switch. Disk, just like accessing a local disk.

After the above steps, after the software level of the storage node OSD1 is abnormal, the disk can be successfully loaded and read and written by other storage nodes OSD2. Data read and write does not need to be reconstructed to restore data, avoiding unnecessary calculations. Moreover, after the storage node OSD1 is abnormal, the reading and writing of data in the entire cloud storage system will not have much performance impact.

After the faulty storage node returns to normal, the MDS requests other storage nodes to unmount the loaded disk. For example, after the failed storage node OSD1 returns to normal, the management node MDS may first request the other storage node OSD2 to unload the disk of the failed storage node OSD1, and then request the storage node OSD1 to load the disk, so that the local disk of the storage node OSD1 The storage node OSD1 can take over the reading and writing by itself, thereby dispersing the pressure of the operating disk of the storage node OSD in the system.

As shown in FIG. 5, in another optional embodiment, when multiple storage nodes in the SAS switch have a software failure, the management node MDS requests the other storage nodes in the SAS switch to load the disks of the failed storage node, thereby managing the nodes The steps of MDS drift disk are as follows:

B1. The storage nodes OSD1 and OSD3 are abnormal.

Assume that the software level of the storage nodes OSD1 and OSD3 is faulty, such as service startup failure and abnormal operating system. In this case, the disk and the data on the disk are normal, and the disk can still be accessed.

B2. The management node MDS requests the storage node OSD2 to load the disks of the storage nodes OSD1 and OSD3.

After a software failure occurs on the storage node, the storage nodes OSD1 and OSD3 cannot report the heartbeat to the management node MDS. The management node MDS considers the storage node OSD1 and OSD3 to be offline. At this time, the management node MDS requests the other storage node OSD2 to try to load the storage node OSD1 and The disk in OSD3, after the other storage node OSD2 is loaded successfully, the data in the disk in the failed storage node can be read normally through the other storage node OSD2, of course, the other storage node OSD2 can also write data to the disk, thereby avoiding The process of data recovery.

Specifically, according to the state of the storage node, the management node MDS can dynamically adjust the disk to other storage nodes for read-write loading. For example, if the management node MDS finds that the storage node is abnormal, it reads and writes the disk data normally; and when the management node MDS finds that the storage nodes OSD1 and OSD3 are abnormal, it requests the storage node OSD2 in the same switch to load the storage nodes OSD1 and OSD3. Disk, through the storage node OSD2 to read and write data in the disks of the storage nodes OSD1 and OSD3 normally, to achieve disk drift.

It can be seen from the above process that the management node MDS implements disk drift according to the state of the storage node. When multiple storage nodes OSD1 and OSD3 have a software failure, the read and write permissions of the disk automatically drift from the failed storage node OSD1 and OSD3 to the normal storage node OSD2 in the storage cluster.

B3. The storage node OSD2 successfully loads the disks in the storage nodes OSD1 and OSD3.

Therefore, the storage node OSD2 can normally access the disks in OSD1 and OSD3 through the SAS switch, and the disks in OSD1 and OSD3 can be normally loaded.

As shown in FIG. 6, in another optional embodiment, when multiple storage nodes in the SAS switch have a software failure, the management node MDS requests the other multiple storage nodes in the switch to load the disks of the failed storage node, thereby managing The steps for implementing drift disk on the node MDS are as follows:

C1. The storage nodes OSD1 and OSD3 are abnormal.

C2. The management node MDS requests the storage nodes OSD2 and OSD4 to load the disks of the storage nodes OSD1 and OSD3.

After a software failure occurs on the storage node, the storage nodes OSD1 and OSD3 cannot report the heartbeat to the management node MDS. The management node MDS considers that the storage nodes OSD1 and OSD3 are offline. At this time, the management node MDS requests other storage nodes OSD2 and OSD4 to try to load the storage node. After the disks of OSD1 and OSD3, and other storage nodes OSD2 and OSD4 are loaded successfully, the data in the disk in the failed storage node can be read normally by other storage nodes OSD2 and OSD4. Of course, the OSD2 and OSD4 can also write data to the disk. This avoids the process of data recovery.

Specifically, according to the state of the storage node, the management node MDS can dynamically adjust the disk to other storage nodes for loading and reading and writing. For example, if the management node MDS does not find the storage node abnormal, it reads and writes the data in the disk normally; and when the management node MDS finds that the storage node OSD1, OSD3 is abnormal, requests the storage node OSD2, OSD4 in the same switch to load the storage node OSD1 , OSD3 disk, through the storage node OSD2, OSD4 normally read and write data in the storage node OSD1, OSD3 disk, for example, the storage node OSD2 can normally read and write data in the storage node OSD1 disk, and, through storage The node OSD4 normally reads and writes data in the disk of the storage node OSD3, thereby realizing disk drift.

It can be seen from the above process that the management node MDS implements disk drift according to the state of the storage node. When multiple storage nodes OSD1 and OSD3 have a software failure, the read and write permissions of the disk drift from the failed storage nodes OSD1 and OSD3 to the normal storage nodes OSD2 and OSD4 in the storage cluster.

C3. The storage nodes OSD2 and OSD4 successfully load the disks of the storage nodes OSD1 and OSD3.

After the disk drifts, all disk read and write requests are made through the normal storage nodes OSD2 and OSD4. The storage nodes OSD2 and OSD4 use the drifted disks as if they were local disks.

Therefore, the storage nodes OSD2 and OSD4 can normally access the disks in OSD1 and OSD3 through the SAS switch, and the disks in OSD1 and OSD3 can be normally loaded.

In other embodiments of the present application, a disk dynamic loading device is proposed. The disk dynamic loading device may be the aforementioned management device, or may also be the aforementioned storage node, which may include:

One or more processors, a storage device that stores one or more programs;

In other embodiments of the present application, a computer-readable storage medium is also provided, on which a computer program is stored, and when the computer program is executed by a processor, the disk dynamic loading method is implemented.

The above is a detailed introduction to a method and device for dynamically loading a disk provided by the present application. In this application, the disk is drifted between storage nodes inside the object storage to realize the dynamic loading of the disk. After the software of the storage node fails, Through the SAS switch, the drift of the storage node can be realized, and the disk data of the failed storage node can be continuously accessed to improve the availability of the object storage disk. This document uses specific examples to explain the principles and implementation of this application. The descriptions of the above examples are only used to help understand the method and core ideas of this application; at the same time, for those of ordinary skill in the art, based on this application There will be changes in the specific implementation and application scope of the idea.

The above is only the preferred specific implementation of this application, but the scope of protection of this application is not limited to this, any person skilled in the art can easily think of changes or changes within the technical scope disclosed in this application. Replacement should be covered within the scope of protection of this application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

A method for dynamically loading a magnetic disk is characterized by being applied to a cloud storage system. The cloud storage system includes a management node and multiple storage nodes. The multiple storage nodes access the same SAS switch. The method includes:

When the management node detects a software failure of the first storage node of the plurality of storage nodes, it sends a disk load instruction to the second storage node of the plurality of storage nodes;

After receiving the disk loading instruction, the second storage node loads the disk of the first storage node through the SAS switch.
The method according to claim 1, wherein the method further comprises:

The management node updates storage node information corresponding to the locally stored disk.
The method according to claim 2, wherein the method further comprises:

When the management node receives the read request to read the data of the disk, the management node sends the read request to the second storage according to the updated storage node information corresponding to the disk stored locally node;

The second storage node reads the data in the disk through the SAS switch according to the received read request.
The method according to claim 2, wherein the method further comprises:

When the management node receives a write request to write data to the disk, it sends the write request to the second storage according to the updated storage node information corresponding to the disk stored locally node;

The second storage node writes data to the disk through the SAS switch according to the received write request.
The method according to claim 1, wherein loading the disk of the first storage node through the SAS switch includes:

The second storage node updates the index information of the disk in the first storage node to the database of the second storage node.
The method according to claim 2, wherein the management node updating the storage node information corresponding to the locally stored disk includes:

The management node correspondingly updates the storage disk information of the disk and the second storage node to a local database.
The method according to claim 1, wherein before the management node updates the storage node information corresponding to the locally stored disk, the method further comprises:

The management node receives the message that the disk is loaded successfully from the second storage node.
A cloud storage system, characterized in that the cloud storage system includes: a management node and a plurality of storage nodes, the plurality of storage nodes are connected to the same SAS switch, and the plurality of storage nodes include a first storage node and a Two storage nodes, where:

The management node is configured to send a disk load instruction to the second storage node when it detects that the first storage node has a software failure;

The second storage node is configured to load the disk of the first storage node through the SAS switch after receiving the disk loading instruction.
The system of claim 8, wherein:

The management node is also used to update locally stored storage node information corresponding to the disk.
The system of claim 9, wherein:

The management node is further configured to, when receiving a read request to read the data of the disk, send the read request to all of the disks according to the updated storage node information corresponding to the disk stored locally Describe the second storage node;

The second storage node is also used to read the data in the disk through the SAS switch according to the received read request.
The system of claim 9, wherein:

The management node is further configured to, when receiving a write request to write data to the disk, send the write request to all the disks according to the updated storage node information corresponding to the disk stored locally Describe the second storage node;

The second storage node is further configured to write data to the disk through the SAS switch according to the received write request.
The system according to claim 8, characterized in that:

The second storage node is also used to update the index information of the disk in the first storage node to the database of the second storage node.
The system according to claim 9, characterized in that:

The management node is also used to correspondingly update the storage node information of the disk and the second storage node to a local database.
The system according to claim 8, characterized in that:

The management node is also used to receive a message that the second storage node successfully loads the disk.