CN103019894B - Reconstruction method for redundant array of independent disks - Google Patents

Abstract

This application discloses a method for rebuilding a Redundant Array of Independent Disks (RAID), including: A. The controller of the RAID system finds that the first disk in the RAID system cannot respond to IO operations, and turns off the power of the first disk separately, and Start a timer of a predetermined duration; B, during the timing of the timer, the RAID system performs normal read and write operations, and records all the stripe numbers that have been written during this period; C, the timer is overtime and opened The power supply of the first disk, power on the first disk; D, after the first disk is powered on, perform a read and write test operation on the first disk; E, judge whether the first disk is read and write normally, if so, execute F, otherwise Execute step G; F, restore the data in the corresponding stripe of the first disk according to all the stripe numbers recorded during the power-off period of the first disk, and end this process after the recovery; G, mark the first disk as For a bad disk, replace the first disk with a second disk as a hot spare disk, perform calculations based on the data and parity checks of other disks in the RAID system, and write the calculation results into the second disk.

Description

A Reconstruction Method of Independent Redundant Disk Array

technical field

The present application relates to the field of computer storage technology, in particular to Redundant Array of Independent Disks (RAID) technology, and in particular to a reconstruction method of Redundant Array of Independent Disks.

Background technique

RAID is a technology that combines multiple independent disks in different ways to form a disk group (logical disk), thereby providing higher storage performance and data redundancy protection than a single disk. The principle of RAID technology is to store data and corresponding parity information on each disk that makes up the RAID system, and store parity information and corresponding data on different disks. When the data on a disk of the RAID system is damaged, use the remaining data and corresponding parity information to recover the damaged data. As the foundation and key component of the network storage system, RAID is famous for its fast speed, large volume and high reliability. After the emergence of RAID technology, the application requirements in various fields such as industry, military affairs, and education are very extensive, and the research on RAID technology has always been a hot spot in the industry.

The different ways to form a disk array are called RAID levels (RAID Levels). For example, common RAID levels include RAID0, RAID1, RAID5, and RAID6. Different RAID levels provide different data protection schemes.

Taking RAID5 composed of 4 disks as an example, only one hard disk is allowed to fail. When one disk fails, RAID5 does not have the function of data redundancy protection, so when a faulty disk occurs, it needs to be replaced as soon as possible. When the faulty hard disk is replaced, the disk controller will use the data and parity on the normal disk to perform calculations, and write the calculation results to the new disk after replacement. This process is called RAID reconstruction.

The purpose of rebuilding is to make RAID regain the function of data redundancy protection. When a RAID disk fails, disk array manufacturers generally use the hot spare disk technology to automatically rebuild the RAID. Hot spare disk technology, in simple terms, is to designate a disk for the RAID as a hot spare disk when creating a RAID system. When a member disk of the RAID system fails, the hot spare disk can automatically replace the failed disk and trigger RAID. reconstruction. As the name implies, when a "hot" spare disk replaces a failed disk, the read and write services on the RAID system do not need to be interrupted, that is, when the RAID system is rebuilt, the read and write operations on the RAID system can still be performed.

In the prior art, when an input/output (IO) request of the upper layer cannot be responded to by a member disk of the RAID system, it is generally considered that the member disk has failed, and the RAID system will automatically start a reconstruction process. The reconstruction operation of the RAID system is expensive and takes a long period, which affects the performance of normal data IO. Generally, during the reconstruction period, if another disk fails, the RAID system will directly crash, which makes the RAID system very fragile. Therefore, it should be avoided as much as possible. Rebuild operation.

Contents of the invention

The present application provides a RAID rebuilding method, which can reduce the probability of RAID rebuilding as much as possible.

A method for rebuilding a RAID provided in an embodiment of the present application includes:

A, the controller of the RAID system finds that the first disk in the RAID system cannot respond to the IO operation, turns off the power supply of the first disk separately, and starts a timer of a predetermined duration;

B. During the timing of the timer, the RAID system performs normal read and write operations, and records all stripe numbers that have been written during this period;

C, the timer is overtime, turn on the power supply of the first disk, and power on the first disk;

D. After the first disk is powered on, perform a read and write test operation on the first disk;

E. Determine whether the first disk is read and written normally, if so, execute F, otherwise execute step G;

F. Restore the data in the corresponding stripe of the first disk according to all the stripe numbers recorded during the power-off period of the first disk, and end this process after the restoration is completed;

G. Mark the first disk as a bad disk, replace the first disk with the second disk as a hot spare disk, calculate according to the data and parity of other disks in the RAID system, and write the result of the calculation into the second disk disk.

Preferably, the read and write test operations include:

D1. Check whether the first disk is online and has been loaded into the operating system by the driver. If not online, the first disk is a bad disk; if online, continue to step D2;

D2. Send the SCSI command "TEST UNIT READY" to the disk to check whether the disk is ready for reading and writing; if it cannot be read and written, the disk is a bad disk; if it can execute step D3;

D3. Write the RAID metadata corresponding to the first disk recorded in the operating system to the corresponding metadata position of the disk. If the writing fails, then it is determined that the first disk is a bad disk. If the writing is successful, continue to perform step D4;

D4. Perform a read operation on the RAID metadata of the first disk. If the read is successful, the first disk is confirmed as a good disk, and if the read fails, it is determined that the first disk is a bad disk.

It can be seen from the above technical solutions that when a disk in the RAID system cannot respond to IO operations, it is first powered off. During the power-off period, the application layer is allowed to read and write to the RAID normally, and writes Then, power on the disk to test whether it can read and write normally. If so, start to restore the data in the corresponding stripe of the disk according to the record of all stripe numbers that have been written; otherwise , marks the disk as bad, and starts the normal rebuild process. In this way, the disks of the RAID system can be restored to normal without rebuilding in most cases.

Description of drawings

FIG. 1 is a flow chart of a method for rebuilding an independent redundant disk array provided by an embodiment of the present application.

Detailed ways

In most cases, the IO request of the upper layer cannot be responded by a member disk of the RAID system, and it is not that the disk as the member disk is really damaged. According to the statistics of the disk manufacturer Seagate, when the disk cannot respond to IO requests, 95% of the cases are caused by software errors such as firmware and verification. In these cases, the disk can still be valid through simple repair operations; only 5% of the cases Next, it is due to a real disk corruption. Therefore, if the rebuilding process is started on the RAID system when the disk is not really damaged, the use and maintenance costs of the RAID system will be greatly increased.

This application provides a rebuilding method of an independent redundant disk array, and its basic idea is: a circuit for controlling the disk to be powered off and on separately is provided on each disk slot interface of the RAID system; When the disk cannot respond to the IO operation, it is first powered off. During the power-off period, the application layer is allowed to read and write to the RAID normally, and all the stripe numbers of the write operation occur during this period; and then the disk is powered on. Processing, test whether it can read and write normally, if so, start to restore the data in the corresponding stripe of the disk according to all the recorded stripe numbers that have been written; otherwise, mark the disk as a bad disk, and start the regular reconstruction process.

In order to make the technical principles, features and technical effects of the technical solution of the present application clearer, the technical solution of the present application will be described in detail below in conjunction with specific embodiments.

The process flow of a method for rebuilding an independent redundant disk array provided in the embodiment of the present application is shown in Figure 1, including the following steps:

Step 101: The controller of the RAID system finds that a certain disk in the RAID system cannot respond to the IO operation, turns off the power of the disk individually, cuts off the power of the disk, and starts a timer with a predetermined duration. Hereinafter, this disk is referred to as a first disk.

Step 102: During the timing of the timer (that is, during the power-off period of the first disk), the RAID system performs normal read and write operations, and records all stripe numbers that have undergone write operations during this period.

Step 103: When the timer times out, turn on the power of the first disk, and power on the first disk.

Step 104: After the first disk is powered on, perform read and write test operations on the first disk.

In the embodiment of this application, the read and write test performs the following operations:

D1. Check whether the first disk is online and has been loaded into the operating system by the driver. If not online, the first disk is a bad disk; if online, continue to step D2;

D2. Send the SCSI command "TEST UNIT READY" to the disk to check whether the disk is ready for reading and writing; if it cannot be read and written, the disk is a bad disk; if it can execute step D3;

D3. Write the RAID metadata corresponding to the first disk recorded in the operating system to the corresponding metadata position of the disk. If the writing fails, then it is determined that the first disk is a bad disk. If the writing is successful, continue to perform step D4;

D4. Perform a read operation on the RAID metadata of the first disk. If the read is successful, the first disk is confirmed as a good disk, and if the read fails, it is determined that the first disk is a bad disk.

Step 105: Judging whether the first disk is read and written normally, if yes, go to step 106, otherwise go to step 107.

Step 106: Recover the data in the corresponding stripe of the first disk according to all the stripe numbers recorded during the power-off period of the first disk, and end the process after the restoration is completed.

Step 107: mark the first disk as a bad disk, replace the first disk with the second disk as a hot spare disk, calculate according to the data and parity of other disks in the RAID system, and write the result of the calculation into the first disk In the second disk.

The above descriptions are only preferred embodiments of the application, and are not intended to limit the scope of protection of the application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the technical solutions of the application are It should be included within the protection scope of this application.

Claims (1)

1. a method for reconstructing of raid-array RAID, is characterized in that, comprising:

The controller of A, RAID system finds that the first disk in this RAID system cannot respond I/O operation, closes separately the power supply of the first disk, and starts the timer of a scheduled duration;

B, during described timer timing, RAID system carries out normal read-write operation, and all bar reel numbers of write operation occurred record during this period;

C, described timer expiry, open the power supply of the first disk, powers on to the first disk;

After D, the first disk power on, the first disk is done and carries out readwrite tests operation; Described readwrite tests operation comprises:

D1, check the first disk whether online and by drive load in operating system, if online, the first disk is low-quality disk; Continue to perform step D2 if online;

D2, to this disk send " TEST UNIT READY " this scsi command chkdsk whether be ready to read and write; If cannot read and write, disk is low-quality disk; If can step D3 be performed;

D3, RAID metadata corresponding for the first disk recorded in operating system is write the position of these disk corresponding element data, if write failure, then judge that the first disk is low-quality disk, continued to perform step D4 if write as merit;

D4, do read operation to the first disk RAID metadata, if the merit of being read as, the first disk confirms as well to coil, and reading failure then judges that the first disk is low-quality disk;

E, judge whether the first disk is read and write normally, if so, performs F, otherwise perform step G;

F, according to the generation of the first disk turnoff time interocclusal record all bar reel numbers of write operation, recover data in the corresponding band of the first disk, be recovered rear process ends;

G, be low-quality disk by the first disk label, the second disk as HotSpare disk is replaced the first disk, calculate according to the data of other disks in RAID system and parity checking, the result of calculating is write in described second disk.