JPH0713867A - Disk subsystem - Google Patents

Abstract

PURPOSE:To prevent unwritten data from being held on a cache memory in case of trouble of writing from the cache memory to a disk device as to the write-after type disk subsystem. CONSTITUTION:Disk devices 107 and 109 are provided with preliminary areas 108 and 110 where alternate tracks for the writing trouble are assigned, and a memory 104 which stores control information for managing the alternate tracks at the time of the trouble is incorporated in a controller 101. Data which are sent from a CPU 106 to the cache memory 102 are written in the alternate tracks in the preliminary areas 108 and 110 from the cache memory 102 if the writing trouble occurs. When preliminary area alternate tracks overflow, the writing from the CPU 106 to the cache memory 102 is inhibited and data are written directly in the disk devices 107 and 109 from the CPU 106.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a disk subsystem such as a magnetic disk device adopting a write-after system, and more particularly to data when a disk device failure occurs during writing from a cache memory to the disk device. The present invention relates to a disk subsystem including a recovery process.

[0002]

2. Description of the Related Art Generally, in a computer system,
As a means for eliminating the access gap between the CPU and the magnetic disk device, an access speed of the magnetic disk device is improved by interposing a cache memory between the CPU and the magnetic disk device. This is because by holding the data in the cache memory, it is possible to avoid mechanical operations such as positioning and waiting for rotation that are access necks of the magnetic disk device.

Conventionally, the following two methods are generally known as a writing method of a cache memory.

First, after the data writing from the upper CPU to the cache and the magnetic disk device is completed,
This is a write-through method of reporting the end to the upper CPU.

The second is a write-after system in which after writing from the upper CPU to the cache memory, a completion report is sent to the upper CPU, and then writing from the cache memory to the magnetic disk device is performed.

In the first write-through method, since the completion report is made after writing to the magnetic disk device, the access speed is no different from that of the conventional magnetic disk device, but a failure occurs when writing to the magnetic disk device. When
Since the error report to the upper CPU can be immediately made in synchronization with the upper CPU, failure recovery is easy. On the other hand, in the second write-after method, since the end is reported when the data writing to the cache memory is completed, the access speed can be the same as the writing speed of the cache memory, which is very high. Therefore, when a failure occurs during writing to the magnetic disk device, the failure report is not synchronized with the upper CPU, which makes it difficult to recover data.

An example of the latter write-after type is known as an IBM 3990 control device, the outline of which is "IBM 3990 Storage C".
Intro Reference "(reference 1).

[0008]

According to the above-mentioned document 1,
"Pinned Data" refers to data in the cache memory that cannot be written due to a failure in writing data from the cache memory to the magnetic disk device.
It is defined as "a", and "P"
The presence of "inned Data" is reported to the upper CPU. The data once recognized as "Pined Data" is not subject to the write processing to the magnetic disk device, and the subsequent recovery action is taken. It is held in the cache memory up to.

As a recovery method for this "Pinned Data", the operator uses the "Pinn Data"
If an "ed Data" notification is found, the following recovery procedure is generally followed.

(1) The writing from the cache memory to the magnetic disk device is performed again. Writing is done by command, and the operator will attempt recovery actions.

(2) If the above writing fails,
A file including "Pinned Data" or a magnetic disk device volume is read (dumped), and if the read is successful, the failure factor is repaired,
"Pinned" held in the cache memory
Data can be recovered by discarding “Data” and recovering (restoring) the dumped data in file units or volume units.

(3) When the reading of the magnetic disk device volume fails in the above (2), not only the cause of failure is repaired but also full recovery from the dump data periodically backed up must be tried. I won't.
Therefore, the backup operation of the user for a failure becomes complicated.

As described above, when a failure occurs during the writing process, it is temporarily set to "PinnedData",
The method of holding the data in the cache memory has a problem in that when recovering the data later, it is necessary to manually clean up the data, and the operation form becomes complicated.

Therefore, an object of the present invention is to solve the above problems and to provide a "Pinned Data" when a failure occurs during writing from the cache memory to the disk device.
An object of the present invention is to provide a disk subsystem that can easily perform recovery without holding unwritten data as a ″ in a cache memory.

[0015]

To achieve the above object, the present invention has a cache memory between a host device (CPU) and a disk device, and when data is written from the host device to the cache memory. In the write-after type disk subsystem that reports the end to the upper device and then writes from the cache memory to the disk device,
A control means is provided for writing data in the cache memory to a spare area of the disk device when a disk device failure occurs when writing data from the cache memory to the disk device.

Further, a storage means for storing management information (spare area write control information) indicating that data is written in the spare area is provided in the disk controller, and a read or write command for this data is located at the upper position. When the device is present, the spare area write control information is used to directly read or write data from the spare area.

Furthermore, when the spare area overflows, a means is provided for immediately writing to the disk device in response to a write command from the host device and directly writing to the disk device. It is configured so that unwritten data due to a disk device failure is not retained inside.

[0018]

The operation based on the above configuration will be described.

According to the present invention, in the write-after type disk subsystem, a spare area is provided on the disk device, and when a failure occurs at the time of writing data from the cache memory to the disk device, the cache memory is controlled by the control means. Since the above data is written (saved) in this spare area, if a failure occurs when writing to the disk device, the data will not be written to the disk device and will remain in the cache memory forever. Inconvenience can be prevented. The spare area may be provided in the same disk device in which the failure has occurred, or in the disk device group under the disk control device.

Since the location of the data can be recognized by referring to the spare area write control information (spare area management information) indicating that the data has been written in the spare area, the host device (CPU) on the spare area can recognize it. Direct access including data is possible.

Further, when the spare area overflows, immediately in response to a write command from the host device, the write to the cache memory is suppressed, and the host device directly writes to the disk device.
When the spare area overflows and a write failure to the disk device occurs thereafter, alternate writing cannot be performed because there is no alternate track, and unwritten data (so-called Pinned Data) on the cache memory.
It is possible to avoid the situation where a) remains.

In this way, when a failure occurs in writing from the cache memory to the disk device, unwritten data (Pinn) is written in the cache memory.
Since edData) is not held, data recovery can be easily performed.

[0023]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram of a magnetic disk subsystem according to an embodiment of the present invention. In the figure, the control device 101
Is the cache memory 102 and the microprocessor 10.
3, a control memory 104, a data transfer circuit 105,
When a data write command is issued from the upper CPU 106, the data is transferred to the cache memory 102 via the data transfer circuit 105, and a completion report is sent to the CPU 106. After that, the control device 101 determines that the upper CPU 1
Data is written from the cache memory 102 to the magnetic disk device 107 via the data transfer circuit 105 asynchronously with 06. This write operation is hereinafter referred to as destaging. Cache memory 102
Also, the control memory 104 is backed up by a non-volatile memory, and guarantees data in the event of a cache failure such as power failure and
Control information can be handed over when the power of the control device 101 is turned on and off.

The magnetic disk device 107 is provided with a spare area 108, which is used as a data writing area when data cannot be written due to a failure of the magnetic disk device during the destaging.

The magnetic disk device 109 is also provided with a similar spare area 110. These spare areas have a plurality of alternative tracks of unwritable data therein and are also used for backup of the control information in the control memory 104.

FIG. 2 is a conceptual diagram of a control information table stored in the control memory 104 of FIG. This table has spare area maps 201 to 203 corresponding to the respective magnetic disk devices under the control device. Reserve area map 20
1 is composed of a spare area counter 204 indicating the number of tracks used as alternative tracks in the magnetic disk device, a track address 205 of a faulty track, a magnetic disk device number 206 in which the alternative track exists, and an alternative track address 207. It

In FIG. 1, when the magnetic disk failure occurs during the destaging processing for a certain track address of the magnetic disk device 107 from the cache memory 102 and writing becomes impossible, the microprocessor 1
03 assigns an alternative track to the faulty track. At this time, the control information shown in FIG. 2 is registered in the control memory 104 from the microprocessor 103. In FIG. 2, the spare area counter 204 is counted up and the faulty track address 205, the magnetic disk device number 206, and the alternative track address 207 are registered. As an alternative track, a track in the spare area 108 of the magnetic disk device is assigned, and the destaging process is tried on this alternative track.
Thereafter, in this magnetic disk device, the alternative track address in the spare area is used as the official track address for the faulty track address.

Next, the operation for the read / write command from the upper CPU 106 for the track assigned as the alternative track will be described. When receiving the read / write command, the microprocessor 103 determines whether the track address is an alternative track registered in the control memory 104. If it ’s an alternative truck,
Positioning operation is performed for the spare area device address (magnetic disk device number) 206 and the alternative track address 207 of the spare area map in the control memory of FIG.
Execute read / write instructions. Further, with respect to the write command, the spare area counter 204 is referred to, and it is checked whether or not the upper limit value (the number of tracks) assigned to the magnetic disk device as an alternative track has overflowed (whether there is a space or not). . If a magnetic disk drive failure occurs when the upper limit value is overflowed, data cannot be written because there is no alternative track. Therefore, when the upper limit value overflows, the data writing operation from the host device to the cache memory 102 is suppressed (that is, the write-after operation is stopped) in response to the host CPU write command, and the host CPU directly , Write data to the magnetic disk device. by this,
It is possible to prevent a situation in which the writing is disabled and the alternative track cannot be assigned to the spare area due to a failure during the destage operation from the cache memory.

In the case where a failure occurs in a part of the disk such as dust adhering to the disk surface or partial flaw, a spare area alternative track can be provided on the same magnetic disk device. On the other hand, when the failure extends to one magnetic disk device as a whole due to a defect in the head, it is preferable that the alternative track in the spare area is on another magnetic disk device. Also, even if the spare area of one magnetic disk device overflows (is full), if there is a spare area in another magnetic disk device, it is better to be able to use it. Therefore, the alternative track in the spare area does not necessarily have to be assigned to the same magnetic disk device as the failed magnetic disk device. The spare area device address 206 in the control memory shown in FIG. 2 becomes valid when the alternative track of the magnetic disk device is assigned to the spare area of another disk device under the control of the control device. This is a means for avoiding the inability to access the alternative track when a failure occurs in the magnetic disk device.

The control memory 104 is composed of a non-volatile memory, and is used for the purpose of preventing the control information from being lost due to power-off or the like and inheriting the control information when the power of the control device is turned ON / OFF. To be It is also possible to have a backup of this control information in the spare area of the magnetic disk device.

In the data recovery process, the data including the replacement track may be temporarily stored in the cache memory and then stored in the regular track of the disk device again.

The present invention can be applied not only to magnetic disk devices but also to optical disk devices and magneto-optical disk devices.

[0034]

As described above in detail, according to the present invention, in the write-after type disk subsystem, when a spare area is provided on the disk device and a data write failure from the cache memory to the disk device occurs. In addition, since the control means saves the data in the cache memory to this spare area, there is an effect that unwritten data does not remain in the cache memory even when a write failure occurs in the disk device. can get. In addition, writing to the disk device in response to a write command from the host device (CPU) is performed asynchronously with the host device. However, the write failure that occurs at this time is not reported to the host device, There is also an effect that data write recovery processing can be easily performed.

The spare area write control information indicating that the data has been written in the spare area is stored, and the location of the data can be recognized by referring to this control information. The effect that the data on the disk can be directly accessed is obtained.

Further, when the spare area overflows, writing to the cache memory is suppressed in response to a write command from the host device, and data is written directly from the host device to the disk device, so the spare area overflows. Even in this case, it is possible to obtain the effect that the unwritten data due to the disk device failure is not held in the cache memory.

In this way, since the unwritten data due to the writing failure of the disk is not held in the cache memory, the writing recovery in the disk control device facilitates the failure recovery processing for the unwritten data. The effect that it can be performed is also obtained.

[Brief description of drawings]

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

FIG. 2 is a configuration diagram of control information stored in a control memory used in an embodiment of the present invention.

[Explanation of symbols]

 101 Control Device 102 Cache Memory 103 Microprocessor 104 Control Memory 105 Data Transfer Circuit 106 CPU 107,109 Magnetic Disk Device 108,110 Spare Area 201-203 Spare Area Map 204 Spare Area Counter 205 Fault Track Address 206 Magnetic Disk Device Number 207 Alternative Track address

Claims (3)

[Claims] 1. A write-after system having a cache memory between a host device and a disk device, and when data is written from the host device to the cache memory, an end report is sent to the host device, and then the cache memory writes the data to the disk device. In the disk subsystem described above, when a disk device failure occurs when writing data from the cache memory to the disk device, there is provided control means for writing data in the cache memory to a spare area of the disk device. The disk subsystem to do. 2. A spare area write control information indicating that data is written in the spare area is provided, and the spare area write is performed when there is a read or write command for this data from a host device. 2. The disk subsystem according to claim 1, wherein the disk subsystem is configured to directly read or write data from the spare area using control information. 3. A means for immediately suppressing writing from the host device to the cache memory when the spare area overflows, and directly writing from the host device to the disk device. Alternatively, the disk subsystem described in 2.