WO1997027530A1

WO1997027530A1 - Magnetic disk subsystem

Info

Publication number: WO1997027530A1
Application number: PCT/JP1996/000094
Authority: WO
Inventors: Takeki Okamoto; Minoru Yoshida; Hirozumi Matoba
Original assignee: Hitachi, Ltd.
Priority date: 1996-01-22
Filing date: 1996-01-22
Publication date: 1997-07-31

Abstract

A disk array subsystem having magnetic disk drives provided with a diagnostic function for maintenance of media. The subsystem sets up a substitutive block in a magnetic disk drive when a defective block of which the data may become unreadable is detected by the diagnosis and writes the data of the defective block in the substitutive block when the system can read out the data of the defective block. When system cannot read out the data of the defective block, the subsystem restores the data of the defective block by recovering the data by using the data of the parity group of the disk drives and setting up a substitutive block in a magnetic disk drive, and then, writing the recovered data in the substitutive block.

Description

Details

Title of Invention Magnetic Disk Sub, Technical Field

The present invention relates to a magnetic disk subsystem, and more particularly, to a disk array subsystem that performs preventive maintenance of a medium. Background art

In a disk subsystem equipped with a magnetic disk drive, the magnetic disk subsystem spontaneously performs diagnosis for the purpose of preventive maintenance of media asynchronously with input / output requests to / from a central processing unit.

The magnetic disk subsystem described in Japanese Patent Application Laid-Open Publication No. 100-801,1993 stores the diagnostic information internally or reports it to the central processing unit to provide maintenance personnel with deterioration and failure of the media. Of a magnetic disk subsystem by setting a replacement block for a failed block that may become unreadable or copying the data of the failed block to the replacement block and recovering it. Performance is enhanced.

In addition, in a disk array subsystem that provides redundancy by applying RAID technology to multiple magnetic disk devices, if a data failure within the redundancy is detected during input / output processing with the central processing unit. However, Patters on, DA et al., A Case for Redundant Arrays of Inexpensive Disks (RA ID) ", Report no. UCB / CSD87 / 391, Computer Science Div. University of Cal ifornia, Berkeley, 1987.

However, the diagnostics for preventive maintenance of the media cannot be read. When this block is detected, the data of the failed block cannot be recovered.

In particular, in areas where the frequency of input / output processing with the central processing unit is low, the frequency of detection of a failure block due to the input / output processing with the central processing unit is low, so to improve the data integrity performance of the magnetic disk subsystem. If a block that cannot be read is detected during preventive maintenance diagnosis, a means for recovering the data of the failed block is required. Disclosure of the invention

In order to solve the above-mentioned problems, a magnetic disk subsystem according to the present invention includes a means for detecting an unreadable failure block by a diagnosis for the purpose of preventive maintenance of a medium voluntarily or instructed by a control device, Means for recovering the block data using the parity group data and saving it to the memory buffer, means for setting a replacement block for the detected fault block, and recovery data for the faulty block saved to the memory buffer Means for writing to the replacement block.

Further, the magnetic disk subsystem according to the present invention has means for detecting a failure block which may become unreadable due to a diagnosis intended for preventive maintenance of the medium voluntarily or in accordance with an instruction of the control device, The data of the detected unreadable block at risk is read into the memory buffer, a replacement block is provided at the drive at which the block at risk of unreadable is detected, and the data in the memory buffer is stored in the replacement block. Writing allows recovery of data from blocks that may be unreadable.

According to the present invention, when a failure block that cannot be read is detected by the preventive maintenance function of the magnetic disk drive, the parity group is Writes data to a replacement block when it detects a block that could become unreadable due to data recovering the data in the failed block or due to preventive maintenance of the magnetic disk drive. This makes it possible to recover from a failure block before input / output processing with the central processing unit. As a result, the frequency of detecting and recovering from a failed block during input / output processing with the central processing unit can be reduced, thereby improving the processing performance of the disk subsystem. In addition, there is an effect that data maintenance performance is improved even in an area where the frequency of input / output processing with the central processing unit is low. BRIEF DESCRIPTION OF THE FIGURES

FIG. 1 is a diagram showing a configuration of a disk subsystem according to an embodiment of the present invention.

FIG. 2 is a diagram showing a flow chart of the preventive diagnosis processing according to the embodiment of the present invention.

FIG. 3 is a diagram showing a configuration of a magnetic disk drive according to an embodiment of the present invention. FIG. 4 is a diagram C showing a configuration of a disk subsystem of another embodiment of the present invention. BEST MODE FOR CARRYING OUT THE INVENTION

Next, an embodiment of a magnetic disk subsystem for performing data recovery during preventive maintenance according to the present invention will be described.

FIG. 1 is a configuration diagram of a disk subsystem according to an embodiment of the present invention. One or more host interfaces 1 1 0 are provided between the central processing unit 101 and the host interface control circuit 127 provided in the control unit 102. Connected by The host interface control circuit 127 and the main processor 120 are connected by a main processor / host interface control circuit interface 170. Further, the host interface control circuit 127 and the memory buffer 122 are connected by a host interface control circuit-memory buffer interface 171.

Further, SCS I protocol control circuits 125 and 126 for performing SCS I (Small Computer System interface) protocol control are provided in the control device 102. The SCSI protocol control circuits 125, 126 and the magnetic disk drives 150, 151, 152, 153 are connected by SCSI 111, 112, 113, 114 . The SCS I protocol control circuits 125, 126 and the main processor 120 are connected by the main processor-SCS I protocol control circuit interfaces 142, 143.

Further, in the control device 102, data transfer control circuits 123 and 124 for controlling high-speed data transfer with the memory buffer 122 are provided. The data transfer control circuits 123 and 124 and the memory buffer 122 are connected by the data transfer control circuit ′ and the memory buffer interfaces 146 and 147. Further, the data transfer control circuits 123, 124 and the SCS I protocol control circuits 125, 126 are connected by data transfer control circuit / SCS I protocol control circuit interfaces 148, 149. Further, the data transfer control circuits 123 and 124 and the main processor 120 are connected by a main processor / data transfer control circuit.

Further, a parity generation circuit 121 for creating parity data is provided in the control circuit 102. Parity generation circuit 1 2 1 and main processor It is connected to the main processor by an interface 140 between the main processor and the parity generation circuit. Further, the parity generation circuit 122 and the memory buffer 122 are connected by an interface 141 between the parity generation circuit and the memory buffer.

And, in the memory buffer 122, the memory buffer areas 150, 131, 132, 133 corresponding to the magnetic disk drives 150, 151, 152, 153 respectively are provided. Provide.

FIG. 2 is a flowchart of an embodiment of preventive maintenance according to the present invention. The main processor 120 uses the SCS I protocol control circuit 125 to issue an SCS I command to the drive to cause the magnetic disk drive 150 to instruct the control unit or report the result of the media diagnosis performed voluntarily by the drive. Yes (step 201). In response to the diagnosis, there is a risk that the contents of the diagnostic result received from the drive may make it unreadable. In order to read the data of the failed block 160 into the area 130 in the memory buffer, the transfer between the magnetic disk drive and the SCSI protocol control circuit is performed using the SCSI protocol control circuit 125 using the SCSI protocol. The data transfer between the magnetic disk drive and the memory buffer is performed using the data transfer control circuit 123 for the transfer between the protocol control circuit and the memory buffer (step 203). If the process of reading data from the magnetic disk drive to the memory buffer is successful (step 204), the main processor 120 uses the SCS I protocol control circuit 125 to control the magnetic disk drive 150. Issue a SCSI command to set alternate block 16 1 for block 160 (step 207). Then, the main processor 120 stores the area 1 in the memory buffer. To write 30 data to the replacement block 161 of the magnetic disk drive 150, the transfer between the memory buffer and the SCS I protocol control circuit uses the data transfer control circuit 123, and the SCS I protocol control circuit The transfer between the hard disk drive and the magnetic disk drive uses the SCSI protocol control circuit 125 to transfer the data between the memory buffer and the magnetic disk drive (step 208) to recover the failed block 160.

Further, in step 203, when the reading process from the magnetic disk drive to the memory buffer ends abnormally and the data of the fault block 160 cannot be read (step 204), the main processor 120 sets the fault block 160 The data of the blocks 162, 163, and 164 forming the parity loop are read out to the memory buffer areas 131, 132, and 133 corresponding to the magnetic disk drives 151, 152, and 153 (step 205). In the data transfer between the magnetic disk drive and the memory buffer, the transfer between the magnetic disk drive and the SCSI protocol control circuit uses the SCS I protocol control circuits 125 and 126, and the SCS I protocol control The transfer between the circuit and the memory buffer uses the data transfer control circuits 146 and 147. The main processor 120 uses the parity generation circuit 121 to read the data in the areas 131, 132, and 133 in the memory buffer, and recovers the data of the unreadable failure block 160 using the RAID function. Write to memory buffer area 130 (step 206). Then, the main processor 120 uses the SCS I protocol control circuit 125 to issue an SCS I command to the magnetic disk drive 150 to set the replacement block 161 for the failure block 160 (step 207). Main processor 120 is memory Data is transferred between the memory buffer and the magnetic disk drive to write the data in the buffer area 130 to the replacement block 161, and the data in the failed block 160 is recovered (step 208). . In the data transfer between the memory buffer and the magnetic disk drive, the data transfer between the memory buffer and the SCSI protocol control circuit uses the data transfer control circuit 123, and the data transfer between the memory buffer and the magnetic disk drive uses the SCSI protocol control circuit. The transfer to and from the magnetic disk drive uses the SCSI protocol control circuit 125.

The magnetic disk subsystem of the above-described embodiment is configured so that each magnetic disk diagnoses a medium. However, the main processor performs the medium diagnosis based on data read from each magnetic disk drive. You can.

Before setting the replacement block 161, write the data or dummy data of the block 160 saved or restored in the memory buffer area 130 to the block buffer 160, and write it to the memory buffer area 130. Re-reading may be performed to check whether writing and reading can be performed normally. Alternatively, the drive itself may check whether the reading of the write data of block 160 is normal.

FIG. 3 is a configuration diagram of the magnetic disk drive according to the embodiment of the present invention. A SCSI protocol control circuit 320 for performing SCSI protocol control is provided in the magnetic disk drive 302, and one or more SCSI devices are provided between the controller 301 and the SCSI protocol control circuit 330. Connect by 1 0. Further, the SCSI protocol control circuit 302 is connected to the main processor 322 provided in the magnetic disk drive 302 by a main processor-SCSI protocol control circuit interface 330. Memory bar provided in the magnetic disk drive 302 It is connected to the buffer 321 by the SCSI protocol control circuit / memory buffer interface 332. In addition, a medium read / write control circuit 323 for controlling reading / writing to / from the medium is provided in the magnetic disk drive 302, and a medium read / write control circuit It is connected by the buffer-to-buffer interface 333, and is connected to the medium 324 by the medium read / write control circuit and the medium 334. In the medium diagnosis of the present invention, the main processor 322 performs a medium diagnosis using the medium read / write control circuit 323 when there is no IZ〇 request from the control device 301, and checks the diagnosis result. This is done by writing to memory buffer 3 2 1 or medium 3 2 4.

FIG. 4 is a configuration diagram of another embodiment of the magnetic disk subsystem according to the present invention. One or more host adapters 403 for controlling data transfer with the central processing unit 401 are provided in the disk subsystem 402. Also, one or more disk adapters 405 for controlling data transfer with the magnetic disk drive group 406 are provided in the disk subsystem 402. Also, a cache memory 430 for temporarily saving data in the disk subsystem 402 and a shared memory 435 for communication between the host adapter 403 and the disk adapter 405 Is provided. The host adapter 403, disk adapter 405 and cache adapter 404 are connected to buses 480 and 481. Also, the magnetic disk drive group 406 may be provided in the disk subsystem 402 in some cases. The host adapter 403 includes host interface control circuits 412, 413 for controlling data transfer between the main processor 414 and the central processing unit 401, and a host interface control circuit. Data transfer for high-speed data transfer between 4 1 2, 4 1 3 and cache memory 4 3 0 Transmission control circuits 415 and 416 are provided. The central processing unit 401 and the host interface control circuits 412, 413 are connected by host interfaces 410, 411. The host interface control circuits 412, 413 and the main processor 414 are connected by the main processor-host interface control circuit interfaces 418, 420, and the data transfer control circuits 415, 4 The connection to 16 is made by the interface between host interface control circuit and data transfer control circuit. The data transfer control circuits 415, 416 and the main processor 414 are connected by the main processor-data transfer control circuit interface 419, 421. Furthermore, the data transfer control circuits 415, 416 and the bus 480 are connected by data transfer control circuits and cache memory interfaces 424, 425, and the cache adapter 404 in the cache adapter 404 is connected via the bus 480. Transfer data with the cache memory 430. The main processor 414 and the node 481 are connected by a main processor-bus interface 423, and data transfer with the shared memory 435 in the cache adapter 404 via the bus 481.

SCSI protocol control circuits 444 and 445 that perform SCS I protocol control are provided in the disk adapter 405, and the magnetic disk drives 470, 471, 472, and 473 are connected by SCSI 456, 457, 458, and 459. . The SCS I protocol control circuits 444 and 445 and the main processor 443 provided in the disk adapter 405 are connected by the main processor and SCS I protocol control circuit interfaces 454 and 455, and cached. A data transfer control circuit for controlling high-speed data transfer with the cache memory 430 in the adapter 404 and a data transfer control circuit between the data transfer control circuits 440 and 441 Connected by SCSI protocol control circuit interface 450, 453. The data transfer control circuits 440, 441 and the main processor 443 are connected by interfaces 451, 452 between the main processor and the data transfer control circuit. In addition, the data transfer control circuit 440, 441 and the bus 480 are connected by the data transfer control circuit-bus interface 444, 448, and cached via the bus 480. Data transfer with memory 430 is performed. The main processor 443 and the bus 481 are connected by a main processor-bus interface 449, and data transfer with the shared memory 435 is performed via the bus 481. Also, a parity generation circuit 442 for generating parity data is provided in the disk adapter 405, and the interface between the parity generation circuit and the main processor 443 is provided by an interface 460 between the main processor and the parity generation circuit. Connecting. Further, the parity generation circuit 442 and the bus 480 are connected by a parity generation circuit / bus interface 446, and are connected to the cache memory 430 via the bus 480. Perform data transfer.

The cache memory 430 and the bus 480 are connected by the cache memory-to-noise interface 436, and are connected to the host adapter 403 and the disk adapter 404 via the bus 480. Perform data transfer. The shared memory 435 and the bus 481 are connected by the shared memory-bus interface 347, and data transfer between the host adapter 403 and the disk adapter 404 via the bus 481. I do. In addition, the cache memory 430 includes cache slots 431, 432, 433, 433, and 437, which correspond to the magnetic disk drives 470, 471, 472, and 473, respectively. Is provided.

The operation of preventive maintenance according to another embodiment of the present invention will be described with reference to the flowchart of FIG. This will be described using a yat.

Using the SCSI protocol control circuit 444, the main processor 443 issues an SCS I command to the magnetic disk drive 470 to cause the magnetic disk drive 470 to instruct the control device or report the result of the media diagnosis performed voluntarily (step 201). ). In response to this, if the diagnosis result indicates that there is a risk that the contents of the diagnostic result received from the drive may make the block unreadable, the main processor 443 sends the failure block 474 from the disk drive 470 if it is diagnosed that there is a failure block 474 (step 202). In order to read the data from the hard disk drive to the cache slot 431 in the cache memory 430, the transfer between the magnetic disk drive and the SCS I protocol control circuit is performed using the SCSI protocol control circuit 444, and the SCS I protocol is used. For the transfer between the control circuit and the cache memory, the data transfer control circuit 440 is used to transfer data between the magnetic disk drive and the cache memory via the bus 480 (step 203). If the process of reading data from the magnetic disk drive to the cache memory has been performed normally (step 204), the main processor 443 sends a failure block 474 to the magnetic disk drive 470 using the SCS I protocol control circuit 444. Issue an SCS I command to set a replacement block 475 for the command (step 207). Then, the main processor 443 writes the data of the cache slot 431 to the replacement block 475 of the magnetic disk drive 470, so that the transfer between the cache memory and the SCS I protocol control circuit is performed by the data transfer control circuit 440. The transfer between the SCSI protocol control circuit and the magnetic disk drive is performed using the SCSI protocol control circuit 444, and the data transfer between the cache memory and the magnetic disk drive via the bus 480. (Step 208), and then Use lock 475 as an alternative to obstacle block 474.

In step 203, if the read processing from the magnetic disk drive to the cache memory ends abnormally and the data of the fault block 474 cannot be read (step 204), the main processor 443 sets the fault block of the magnetic disk drive 470 to failure. The data of blocks 476, 477, and 478 forming a parity group with 474 are read out to the cache slots 432, 433, and 434 corresponding to the respective magnetic disk drives 471, 472, and 473 (step 205). In the data transfer between the magnetic disk drive and the cache memory, the transfer between the magnetic disk drive and the SCS I protocol control circuit uses the SC SI protocol control circuits 444 and 445, and the SC SI protocol control circuit The transfer between the circuit and the memory buffer uses the data transfer control circuits 440 and 441. The main processor 443 reads the data of the cache slots 432, 433, and 434 into the parity generation circuit 442, and outputs the data. The data of the unreadable failure block 474 is recovered using the RAID function of the parity generation circuit 442, and the recovered data of the failure block 474 is written from the parity generation circuit to the cache slot 431 (step 206). Then, the main processor 443 uses the SCSI protocol control circuit 444 to issue a SCSI command to the magnetic disk drive 470 to set the replacement block 475 to the failure block 474 (step 207). The main processor 443 transfers the data from the cache memory to the magnetic disk drive in order to write the data in the cache slot 431 to the replacement block 475 (step 208). In the data transfer between the memory buffer and the magnetic disk drive, the transfer between the memory buffer and the SCSI protocol control circuit uses the data transfer control circuit 440 and the SCS The transfer between the I protocol control circuit and the magnetic disk drive uses the SCSI protocol control circuit 444. Then, the replacement block 4 7 5 is used as an alternative block of the obstacle block 4 7 4.

The magnetic disk subsystem of the above-described embodiment is configured such that each magnetic disk performs a medium diagnosis. The main processor 443 executes the medium diagnosis based on data read from each magnetic disk drive. You may try tT.

Before setting the replacement block 475, write the data or dummy data of the block 474 saved or restored in the cache slot 431 to the block 474, and read it back to the cache slot 431. It may be checked whether or not the writing and reading can be performed normally.

Also, before setting the replacement block, damage the data or dummy data of the block 474 saved or restored to the cache slot 431 to the replacement block 474.Check whether the drive can read normally. Let me check.

According to the present invention, when a failed block that cannot be read is detected by preventive maintenance of the magnetic disk drive, the data of the failed block is recovered by the data of the parity group, or When a block that may become unreadable due to preventive maintenance is detected, the data in that block is written to a replacement block, so that the faulty block can be recovered before I / O processing with the central processing unit. This can reduce the frequency of detecting and recovering from a failure block during input / output processing with the central processing unit, thereby improving the processing performance of the disk subsystem. Also, there is an effect that data maintenance performance is improved even in an area where the frequency of input / output processing with the central processing unit is small.

Claims

The scope of the claims

(1) A magnetic disk sub which has a storage medium having a plurality of areas for storing data, and has a plurality of magnetic disk drives forming a parity group, and a control device interposed between a central processing unit and the magnetic disk drive A system for diagnosing the storage medium asynchronously from the central processing unit, a unit for recovering data in a failure area by using data in another area of the parity group, and setting a replacement area for the failure area. Means for recovering data of the faulty area by the recovery means when the faulty area is detected by the diagnostic means, setting a replacement area of the faulty area by the setting means, and setting the replacement area in the replacement area. A magnetic disk subsystem for writing recovery data.

2. The magnetic disk subsystem according to claim 1, wherein the magnetic disk drive has the diagnostic unit.

(3) The magnetic disk subsystem writes the recovery data to the fault area before setting the replacement area of the fault area by the setting means, and the diagnostic means determines whether the data written to the fault area can be read correctly. 2. The magnetic disk subsystem according to claim 1, wherein a check is performed to determine whether the area is a normal area and the replacement area is not set.

4. The magnetic disk sub-system according to claim 1, wherein the diagnostic unit diagnoses the replacement area after writing the recovery data to the replacement area.

5 A plurality of magnetic disk drives having a storage medium having a plurality of blocks for storing data, forming a parity group, and a central processing unit. 5 and a memory for temporarily storing data transferred between the central processing unit and the magnetic disk drive, and data for a failed block to another block of the parity group. A control unit having a control unit configured to recover the data from data of the disk, and a means for diagnosing the storage medium asynchronously from the central processing unit. Means for setting a replacement block of the block, and when a failure block is detected by the diagnosis means, data of another block of the parity group in which the failure block has occurred is read into the memory, and The control unit recovers the data of the faulty block based on the data in the memory, and the setting unit changes the replacement area of the faulty block. Magnetic disk subsystem set, and wherein the writing 該回 recovery data in alternating charges Proc.

6. The magnetic disk subsystem according to claim 5, wherein the magnetic disk drive includes the diagnosis unit.

7 The magnetic disk subsystem writes the recovery data to the failure block before setting the replacement block of the failure block by the setting unit, and the diagnostic unit writes the recovery data to the failure block. 6. The magnetic disk subsystem according to claim 5, wherein it is checked whether or not the data can be read correctly, and if the data can be read correctly, the block is regarded as a normal block, and the replacement block is not set.

8. The magnetic disk subsystem according to claim 5, wherein the diagnosis unit diagnoses the replacement block after writing the recovery data to the replacement block.

9 A plurality of magnetic disk drives having a storage medium having a plurality of blocks for storing data and forming a parity group, and a central processing unit. Memory that temporarily holds data transferred between the central processing unit and the magnetic disk drive, and another block of the parity group. A magnetic disk subsystem comprising: a control unit having a control unit configured to recover the data from the central processing unit, wherein the means performs diagnosis of the storage medium asynchronously from the central processing unit; Means for setting a replacement block of a block, and when a failure block is detected by the diagnosis means, data of another block of the parity group in which the failure block has occurred is read out to the memory, and The controller recovers the data of the fault block based on the data and stores the data in the memory. An alternate block of the block is set, the data of the memory is written to the alternate block, and when the block which may become unreadable by the diagnostic means is detected, the data of the block is stored in the memory. A magnetic disk subsystem, comprising: reading, setting a replacement block of the block by the setting means, and writing data in the memory to the replacement block.

10. The magnetic disk subsystem according to claim 9, wherein said magnetic disk drive has said diagnosis means.

11 1 The magnetic disk subsystem writes the data in the memory buffer to the faulty block or the block which may become unreadable before setting the replacement block by the setting means, and A check is made to see if the data written in the failed block can be read correctly, and if it can be read correctly, the block is regarded as a normal block, and the replacement block is not set. The magnetic disk subsystem according to item 9. 12. The magnetic disk subsystem according to claim 9, wherein the diagnostic unit diagnoses the replacement block after writing the recovery data to the replacement block.