JP2009104369A - Disk sub-system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To dispense with any spare disk in a conventional RAID configuration, and to prevent data lost by restoring original data even when failure occurs in a plurality of disk drives. <P>SOLUTION: In a plurality of disk drives of RAID configurations, a data region for writing and reading normal data and a spare region for writing and reading data as the alternate of the data region are specified. When the necessity of the copy of the data region of a certain disk drive is generated, the data of the data region of the first disk drive are automatically copied to the spare region of the other disk drive according to the circumstances of the failure of the plurality of disk drives. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a disk subsystem, and more particularly to reduction of data lost when a failure occurs in a disk drive in a disk subsystem comprising a disk array.

  A disk subsystem with a RAID configuration includes a parity disk, and when a failure occurs in one disk drive under the RAID configuration, data can be restored using the parity disk. However, if a plurality of disk drives fail, the error cannot be repaired and data loss occurs.

In Patent Document 1 (Japanese Patent Laid-Open No. 2005-122338), in a RAID-structured disk array having spare disks, the possibility of occurrence of a failure is estimated for each disk drive, and a spare from a disk drive having a high possibility of occurrence of failure A technique for reducing data lost due to failure of multiple disks by automatically copying data to the disk is disclosed.
In legacy systems, it is not uncommon for the area of a disk drive used by the system to be limited, and the unused area tends to increase as the storage capacity of the disk drive increases.

JP 2005-122338 A

  In the disk subsystem having the spare disk as in the above-described prior art, in order to continuously operate even when a plurality of disk drives fail, a spare disk is required, so that the hardware cost increases. In addition, when data is saved to a spare disk, normal user operations may be affected.

  An object of the present invention is to eliminate the need for a spare disk in a conventional RAID configuration, and to copy data to a spare area defined in a disk drive in a RAID configuration, so that the copy data is used when a failure occurs in the disk drive. The original data is restored to reduce data loss.

A disk subsystem according to the present invention is a disk subsystem having a plurality of disk drives with a RAID configuration and processing means for performing control for writing and reading data to and from the disk drive group.
A data area for writing and reading normal data to the plurality of disk drives, and a spare area for writing and reading data as an alternative to the data area,
A managing means for managing the states of the plurality of disk drives, and when the data area of a certain disk drive (first disk drive) needs to be copied from the state of the disk drives by the managing means, the first disk Copy execution means for automatically copying the data in the data area of the drive to the spare area of another disk drive (second disk drive) is configured as a disk subsystem.

  In a preferred example, the disk drive group includes a plurality of primary and secondary disk drives in a mirror configuration, and the data area and the spare area are formed in each disk drive, and the management means includes a pair of disk drives. A management table for managing the first flag indicating one of the defective states (blocked disk) and the second flag indicating the necessity of updating data in the paired disk drive is held in the memory and managed. The copy execution means refers to the management table and copies data from the data area of the first disk drive to the spare area of the second disk drive.

  Preferably, the copy execution means is in a first data copy mode in which data copy is performed in a lump when power is turned off, or not when power is turned off, according to the value of the first flag in the management table. In response to the second flag, a second data copy mode is implemented in which data in the data area of the first disk drive is copied to the spare area of the second disk drive in block units.

  Preferably, the capacity of the spare area is equal to or larger than that of the data area.

  According to the present invention, a spare disk in a conventional RAID configuration is not required, and data can be copied using a storage area of a disk drive having a RAID configuration as a spare area. When a failure occurs in the disk drive, the data lost can be reduced by restoring the original data using the data copied to the spare area.

A preferred embodiment will be described below with reference to the drawings.
FIG. 1 shows an example of a disk subsystem having a RAID configuration.
The disk subsystem 1 executes various related programs to perform data processing accompanying data reading / writing on the disk, and disk control controlware (C / W) for disk control held in the memory 21. 4), an automatic data copy program 5 for performing data copy control on a spare area related to the present invention, a data cache 6, a disk operation status management table 7 held in the nonvolatile memory 22, a SCSI controller 9, And a disk drive group 8 connected via a SCSI bus 91. The disk subsystem 1 is connected to a host system (hereinafter simply referred to as a host) 10 via a main bus 11.

  The disk drive group 8 has a RAID configuration including a plurality of disk drives 100 to 105, and each disk drive is connected to the SCSI controller 9 via the SCSI bus 91. Here, the disk drives 100 and 103, 101 and 104, and 102 and 105 are paired to form a mirror configuration.

  The storage area of each of the disk drives 100 to 105 is divided into a data area A (for example, 100A) and a spare area B (for example, 100B). The storage capacity has a relation of spare area ≧ data area. The data area A stores the processing data of the host 10, and the spare area B stores the processing data instead when the data area A cannot be used due to a failure.

  The disk control C / W 4 reads or writes data recorded in the data area or spare area of each disk drive of the disk drive group 8 as required by an instruction from the host 10. The data cache 6 temporarily stores data read from the disk drive group 8. In response to a data request from the host 10, if the target data is stored in the data cache 6, it is read out from the data cache 6 and transmitted to the host 10.

FIG. 2 shows an outline of the operation at the time of starting from the host 10 to the disk drive group 8.
Step S1 is a data writing process. In this case, data is written in both the data area 100A of the disk drive 100 and the data area 103Å of the disk drive 103. Step S2 is data read processing. In the case of data reading, data is read from the data area 100A of the disk drive 100 or the data area 103A of the disk drive 103.

  Step S3 is a writing process to the data areas of the disk drive 101 and the disk drive 104, as in step S1. Step S4 is a read process from the data area of the disk drive 101 or the disk drive 104, as in step S2.

Step S5 is a process of copying the data area 100A of the disk drive 100 to the spare area 101B of the disk drive 101.
Step S6 is a process of copying data from the data area 101A of the disk drive 101 to the spare area 102B of the disk drive 102, as in step S5. Step S7 is a process of copying data from the data area 102A of the disk drive 102 to the spare area 100B of the disk drive 100, as in step S5.

FIG. 3 shows an example of the disk operation status management table 7.
The disk operation status management table 7 includes a blocked disk management flag 300 having 1 to N pairs, a data copy destination disk drive number management table 301 for each pair, and an update necessary flag 302 for LBA (Logical Block Address) of all data areas. Remember and manage. Here, it is assumed that there are a total of N disk pairs, and each disk has a total of M blocks, each of which is designated by an LBA. The capacity of each block is, for example, 512 bytes.

  A blocked disk refers to a case where one of the paired disk drives is defective. When the blocked disk management flag 300 is “1”, there is a blocked disk, and when it is “0”, there is no blocked disk. In the illustrated example, since the first pair and the Nth pair are “0”, there is no blocked disk, indicating that both disks are good. On the other hand, the second and third pairs are “1”, indicating that there is a blocked disk. What is characteristic in this embodiment is that when a disk drive in a pair relationship has a blocked disk, it is determined that a failure is likely to occur in the pair, and data is copied to the spare area.

Further, when the update necessary flag 302 is “1”, it means that data update of the corresponding block is necessary, and when it is “0”, it means that data update is unnecessary.
In the illustrated example, the management table for the first pair of disk drives 100 and 103 has data update flags for the second pair of disk drives 101 and 104, and the block number of the second pair of disk drives is 1, 2, 4 indicates that the block needs to be updated. Similarly, in the data update flag of the third pair of disk drives 102 and 105 associated with the second pair of disks 101 and 104, the blocks with block numbers 1 and k need to be updated.

FIG. 4 shows an execution process of the automatic data copy program 5 in a normal state.
It is determined whether or not the host 10 is on standby, and automatic data copying is started when the host 10 is on standby (step S10).
When automatic data copying is executed, it is determined whether there is a blocked disk by referring to the blocked disk management flag 300 in the disk operation status management table 7 (step S11). As a result of the determination, if there is a pair of blocked disks (that is, flag 300 is “1”), one of the pair disks is defective, so it is determined that a failure is likely to occur and data is copied to the spare area. Do it, but don't immediately copy data automatically. On the other hand, if the result of the determination is that there is no blocked disk pair (that is, the flag 300 is “0”), automatic data copying is performed.

  Since automatic data copying is performed when there is no access request to the target disk drive from the host 10, one disk drive is operating when there is a blocked disk pair, and all automatic data copying is completed. If the disk drive fails when not in use, data loss occurs. For this reason, automatic data copy is executed in batch for those with a blocked disk pair. Automatic data copying performed in a batch is executed when the power is turned off (step S12).

If the result of determination in step S11 is that there is no blocked disk pair, refer to the data copy destination disk drive number management table 301 in pairs #L (1 to N) recorded in the disk operation status management table 7, and LBA # It is determined whether the update necessity flag 302 of P (1 to M) is “1” (step S13). If the update necessary flag 302 is “1”, the update is necessary, so the data is read from the primary side or the secondary side of the corresponding disk drive pair #L (step S14).
Then, the read data is written to the disk drive pair #L that needs to be updated (the primary disk drive indicated by the data copy destination disk drive number management table 301 for each pair in the disk operation status management table 7) (step S15). .

  In this case, the unit of read and write data is an LBA unit. The read and write data units are performed in LBA units in order to smoothly interrupt the automatic data copy process when there is an access request from the host 10 to the disk subsystem 1. The unit is LBA unit.

Next, the update necessary flag 302 of LBA # P for which the update of the disk operation status management table 7 has been normally completed is set from “1” to “0” (step S16). This completes automatic data copying.
If the update necessity flag 302 in step S13 is “0”, no update is necessary, so the count is incremented to the next LBA # P (step S17). Then, it is determined whether the counted up LBA # P is the final LBA # (M + 1) (step S18). If it is not the final LBA # (M + 1), the automatic data copy is terminated. On the other hand, in the case of the final LBA # (M + 1), it is determined whether the pair #L in the disk operation status management table 7 is the final pair # (N + 1) (step S19). As a result of the determination, if the final pair # (N + 1), the automatic data copy is terminated. If not the final pair # (N + 1), the pair #L is counted up and the automatic data copy is terminated (step S20). ).

  The above processing operation is repeated until the final pair # (N + 1) and the final LBA # (M + 1). When the final pair # (N + 1) is reached, the execution is started again from the first pair #L. When there is an access request from the host 10, the automatic data copy is interrupted until there is no access request from the host 10 again.

FIG. 5 shows an execution process of the automatic data copy program 5 when the power is turned off.
This process is executed as the process in step S12 of FIG. In this case, automatic data copying is performed, and the system power is turned off after the processing is completely completed.
First, it is determined whether the update necessary flag 302 of the pair #L (1-N) LBA # P (1-M) recorded in the disk operation status management table 7 is “1” (step S21). As a result, since the update is necessary when the update necessary flag 302 is “1”, data is read from the primary block of the corresponding disk drive pair #L (step S22). If there is a blocked disk drive, it is read from one block of the disk drive that is not a blocked disk drive.

  Then, the read data is written to one block on the primary side of the disk drive pair # that needs to be updated (indicated by the data copy destination disk drive number management table 301 for each pair in the disk operation status management table 7) (step S23). ). If there is a blocked disk drive, it is written to the disk drive that is not a blocked disk drive.

Next, the update necessary flag 302 of LBA # P for which the update of the disk operation status management table 7 has been normally completed is set from “1” to “0” (step S24). When the update necessary flag 302 is “0”, there is no need for updating, so the next LBA #P is counted up (step S25).
Next, it is determined whether the counted up LBA # P is the final LBA # (M + 1) (step S26). If the result is not the final LBA # (M + 1), the process is repeated until the final LBA # (M + 1) is reached.

Then, it is determined whether the pair #L in the disk operation status management table 7 is the last pair # (N + 1) (step S27). As a result, in the case of the final pair # (N + 1), the automatic data copy is terminated. On the other hand, if it is not the final pair # (N + 1), the automatic data copy is repeated until the final pair # (N + 1) is reached.
Finally, when the automatic data copy is completed, the system power is turned off (step S29). The disk drive for which automatic data copying has been completed replaces the blocked disk drive, and can continue to be used as a RAID configuration.

  According to this embodiment, a recording area of a disk drive having a RAID configuration can be effectively used by defining it as a spare area. In addition, since the data can be automatically copied when the CPU is not using the disk drive, the backup data can be saved without affecting the normal operation of the user.

The figure which shows the structure of the disk subsystem in one Example. The figure which shows the operation | movement at the time of starting to the disk array 11 from the host 2. FIG. The figure which shows the example of the disk operation condition management table. The flowchart which shows the operation | movement at the normal time in one Example. The flowchart which shows the operation | movement at the time of the power-off in one Example.

Explanation of symbols

1: Disk subsystem 10: Host system 11: Main bus
3: CPU 4: Disk control C / W 5: Automatic data copy program 6: Data cache 7: Disk operation status management table 8: Disk drive group 9: SCSI controller 91: SCSI bus.

Claims (4)

In a disk subsystem having a plurality of disk drives with a RAID configuration and processing means for performing control for writing and reading data to and from the disk drive group,
A data area for writing and reading normal data to the plurality of disk drives, and a spare area for writing and reading data as an alternative to the data area,
Management means for managing the status of the plurality of disk drives;
When it is necessary to copy the data area of a certain disk drive (first disk drive) from the state of the disk drive by the management means, data in the data area of the first disk drive is transferred to another disk drive ( Copy execution means for automatically copying to the spare area of the second disk drive). The disk drive group is composed of a plurality of primary and secondary disk drives in a mirror configuration, and the data area and the spare area are formed in each disk drive,
The management means includes a first flag indicating that one of the paired disk drives is defective (blocked disk), a second flag indicating the necessity of updating data in the paired disk drive, Manage the management table to manage
2. The disk subsystem according to claim 1, wherein the copy execution means copies data from the data area of the first disk drive to the spare area of the second disk drive with reference to the management table. The copy execution means is in a first data copy mode in which data copy is performed at a time when the power is turned off according to the value of the first flag in the management table, or when the power is turned off. 3. A second data copy mode is implemented in which data in the data area of the first disk drive is copied to the spare area of the second disk drive in block units according to a flag. Disk subsystem. 4. The disk subsystem according to claim 1, wherein the spare area has a capacity equal to or larger than that of the data area.

Images