JP4799277B2 - Capacity expansion method in RAID device and RAID device - Google Patents

Description

  The present invention relates to a capacity expansion method and a RAID apparatus provided with capacity expansion means in a disk array apparatus [hereinafter referred to as a RAID (Redundant Array of Inexpensive Disks) apparatus) that manages data redundantly using a plurality of storage devices.

A RAID device employs a redundant configuration in order to increase the reliability of stored data and the reliability of the device, and RAID0, RAID1, RAID0 + 1, RAID2, RAID3, RAID4, RAID5, and the like are known.
For example, RAID 0 uses a striping configuration in which data is distributed and stored in units of blocks, and RAID 1 uses a mirroring configuration in which the same data is stored on a plurality of disks. A RAID group that uses both the striping configuration and the mirroring configuration and has a mirroring relationship between the primary disk group and the secondary disk group that are formed of the striped configuration disk group is referred to herein as RAID 0 + 1.
The redundancy is set based on the user's system configuration and the like. However, there are cases where it is desired to change the redundancy, capacity, and RAID level after the system operation is started.
If the system is stopped, it is possible to easily change the redundancy and the like. However, without stopping the system, the redundancy, capacity, and RAID can be maintained in an active state (while receiving I / O from the host). Various techniques for changing the level have been proposed in the past (see, for example, Patent Document 1 and Patent Document 2).

As described above, the function of expanding the capacity of a RAID group or converting the RAID level by adding a disk to an already existing RAID group in an active state is called LDE (Logical Device Expansion), In general, a new volume is created in the expanded part of the RAID group.
Japanese Patent Laid-Open No. 2004-213064 JP 2005-107841 A

In the above-described LDE, there are cases where volumes in the new configuration (RAID group after LDE) and the old configuration (RAID group before LDE) overlap. The processing is performed by performing Read as the old configuration, storing the data temporarily in the buffer, and then performing Write as the new configuration. However, in the above process, the data is temporarily stored in the buffer. There are moments when there is only data.
More specifically, since it is not known how much data has been written to the actual disk during Write, the data in the area that requested Write on the disk is unreliable. For unavoidable reasons, if the buffer is not backed up at the time of a power failure or if a hardware memory failure occurs, the data in the buffer may be lost if the write to the disk fails.
The present invention has been made to solve the above-described conventional problems, and the object of the present invention is to add an even number of disks from a RAID group that performs mirroring such as RAID 1 or RAID 0 + 1, thereby increasing the capacity. When performing expansion to RAID 0 + 1, it is possible to execute LDE while temporarily eliminating the state in which no data exists on the disk.

In the present invention, the above problem is solved as follows.
(1) A RAID group of RAID 1 or RAID 0 + 1, which is composed of 2m (m is a natural number) disk group, and m disk groups and other m disk groups are in a mirroring relationship, and each of m disks In a RAID device having a RAID group consisting of disk groups in which the group is striped, 2n (n is a natural number) disk groups are added to the 2m disk group as follows to expand the capacity.
In the old RAID group, add n disks to the m disk group, which is the primary disk group, so that the new RAID group becomes one of the disk groups that have a mirroring relationship. After the completion of this process, the remaining n disk groups are added to the other m disks that are the secondary disk groups that have not been processed as described above, and a mirroring relationship is established with respect to the primary disk group. Have it.
Since RAID 1 or RAID 0 + 1 is mirrored, even if one of the mirror data is lost, the RAID group only loses redundancy and data on the disk remains.
In the present invention, this is utilized to create a state in which data on the disk is stored as an intermediate state of the entire LDE process, with one of the mirrors stored in the old configuration and one of the mirrors in the new configuration. It is possible to realize a situation in which data exists above.
(2) When expanding the capacity in (1), the primary disk group and the secondary disk group are not expanded at the same time for the same logical block address.
(3) In the above (2), the secondary disk group is expanded so as to follow the expansion of the primary disk group.
(4) In the above (2), the primary disk group is expanded from the smaller or larger logical block address, and the secondary disk group is expanded from the larger or smaller logical block address.

In the present invention, the following effects can be obtained.
(1) During capacity expansion, as an intermediate state of the entire LDE process, the data on the disk is stored with one of the mirrors in the old configuration and one of the mirrors in the new configuration. Since the existing situation can be realized, even if the battery backup is not performed at the time of a power failure or when a hardware memory failure occurs, data is not lost even if writing to the disk fails.
(2) During the above capacity expansion, the secondary disk group is expanded so as to follow the expansion of the primary disk group, for example, so that the primary disk group and the secondary disk group do not expand the same logical block address at the same time. Thus, the expansion process of the primary disk group and the secondary disk group can be performed at the same time, and the processing time can be shortened.

FIG. 1 is a diagram showing a schematic configuration of a RAID apparatus according to an embodiment of the present invention.
In the figure, 1 is a RAID controller, 2-1 to 2-n are disks for storing data, and the RAID controller 1 is connected to the host 3 directly or via a network device. A large amount of data is read from and written to the disk at high speed and randomly as required.
The RAID controller 1 includes a buffer 1a for temporarily storing data when data is read and written, and capacity expansion means 1b for performing capacity expansion processing at the time of capacity expansion, and holds the definition of RAID group configuration information 1c. To do. FIG. 1 shows only the functional configuration according to the present invention. For details of the hardware configuration of the RAID controller, refer to, for example, the above-mentioned Patent Document 1.

The disks 2-1 to 2-n are composed of 2m (m is a natural number) disk groups as described above, and the m disk groups and the other m disk groups are in a mirroring relationship. The disk group is configured as a RAID 1 or RAID 0 + 1 RAID group consisting of striped disk groups. In the present invention, when the disk capacity is expanded, 2n disks are added to the disk group, A RAID group of RAID 0 + 1 composed of (m + n) disks is configured.
For example, as shown in FIG. 2 (a), two disks 2-3 and 2-4 are added to a pair of mirrored disks 2-1 and 2-2, and mirrored and striped. A RAID 0 + 1 RAID group comprising two disks 2-1, 2-3 and 2-2, 2-4 is constructed.
Alternatively, as shown in FIG. 2B, two disks 2-9 and 2-10 are added to the RAID 0 + 1 RAID group to form a RAID 0 + 1 RAID group.

Hereinafter, the processing in the capacity expansion unit 1b when the disk capacity is expanded by adding 2n disks to a RAID group of RAID 0 + 1 composed of 2m disks as described above will be described.
FIG. 3 is a diagram illustrating the LDE process according to the embodiment of this invention.
FIG. 3A shows a state before LDE, in which the primary disk groups 2-1 to 2-4 and the secondary disk groups 2-5 to 2-8 are striped, respectively. -4 and disk groups 2-5 to 2-8 have a mirroring configuration.
From this state, as shown in FIG. 3B, one of the two additional disks, one disk 2-10 is used to expand the secondary disk group 2-5 to 2-8 (this state is intermediate). Called state). At this stage, the primary disk groups 2-1 to 2-4 are not changed, and the added disk 2-10 is not in a mirroring relationship.
Next, as shown in FIG. 3C, the remaining one of the two additional disks is added to expand the primary disk group 2-1 to 2-4. Mirror the created disk.

FIG. 4 is a flowchart showing the processing procedure.
First, the configuration definition is changed to a state during secondary extension (step S1). Next, the progress is confirmed (step S2), and if there is an area before expansion, the primary disk group 2-1 to 2-4 is read with the old configuration mapping, and the secondary disk group 2-5 to 2-8. The new configuration is written (steps S3 and S4).
Next, the progress is updated (step S5), the process returns to step S2, and this process is repeated until there is no area before expansion.
When the above processing is completed to the end, the process goes to step S6, and the configuration definition is changed to a state of primary extension. Next, the progress is confirmed (step S7), and if there is an area before expansion, the secondary disk group 2-5 to 2-8 is read by mapping of the new configuration, and the primary disk group 2-1 to 2-4 is read. The new configuration mapping is performed (steps S8 and S9). Next, the progress is updated (step S10), and the process returns to step S7, and this process is repeated until the area before expansion disappears. When the process is completed to the end, the configuration definition is changed to the extended word state, and the process ends.

In the above embodiment, a case has been described in which only one of the disks having a mirroring relationship is expanded and then the other is expanded. Next, while managing so that data always exists on the disk, An improved example of the above embodiment that can shorten the processing time will be described.
The outline of this embodiment is that both the disk groups are expanded in parallel while managing so that one and the other of the disk groups having a mirroring relationship are not expanded at the same time for the same LBA (logical block address). It is a method of going.
In the following, “follow-up method” and “front-back method” will be described.

(1) Follow-up method In this method, the secondary disk group is expanded so as to follow the expansion of the primary disk group, and both are managed so as not to process the same LBA area. FIG. 5 shows an outline of the extension processing by the follow-up method. FIG. 5 shows a case where RAID0 + 1 is configured by adding two disks 2-3 and 2-4 to RAID1 composed of mirroring disks 2-1 and 2-2.
First, as shown in FIG. 5A, one of the two additional disks, one disk 2-4, is added to expand the area R1 (the area R1 is in an intermediate state). At this stage, the disk 2-1 is not changed.

Next, as shown in FIG. 5B, the remaining one of the two additional disks 2-3 is added, and the expanded area R1 and the added disk 2-3 are added. The region R1 ′ is mirrored (regions R1 and R1 ′ are in the final state). At the same time, the area R2 is expanded for the added disk 2-4 (the area R2 is in an intermediate state).
Further, similarly to the above, as shown in FIG. 5C, the expanded area R2 of the added disk 2-4 and the area R2 ′ of the added disk 2-3 are mirrored (areas R2, R2). 'Is the final state). At the same time, the area R3 is expanded for the added disk 2-4 (the area R3 is in an intermediate state).
Thereafter, the same processing is performed until the expansion processing for all areas is completed.

Here, in the follow-up method, specifically, the following two methods can be considered.
The first method is the same as when the expansion of the primary disk group is completed at the same time the primary disk group is moved to the next LBA range as the primary after the expansion of the primary disk group is completed for a certain LBA range. In this method, secondary expansion is started for the LBA range.
In the second method, the expansion of the primary disk group and the expansion of the secondary disk group are individually managed in parallel, and the expansion is performed in parallel. This is a method of waiting for a certain period of time for the secondary process to start.
In this case, after waiting, it is checked whether the corresponding LBA area is being expanded in the primary disk group. If the process is in progress, the process waits for a fixed time. If not, the secondary disk group is processed. To start.

(2) Front-rear system
In this method, the primary disk group is expanded from the smaller LBA to the larger one, and the secondary disk group is expanded from the larger LBA to the smaller LBA.
FIG. 6 shows an overview of the expansion process using the front-rear system. FIG. 6 shows a case where RAID 0 + 1 is configured by adding two disks 2-3 and 2-4 to RAID1 composed of mirrors 2-1 and 2-2, as in FIG.
First, as shown in FIG. 6A, one of the two additional disks, one disk 2-4, is added, and the area R1 is expanded from the larger LBA (area R1 is in an intermediate state). At this stage, the disk 2-1 is not changed.

Next, as shown in FIG. 6B, the remaining one of the two additional disks 2-3 is added, and the expanded area R1 and the added disk 2-3 are added. The region R1 ′ is mirrored (regions R1 and R1 ′ are in the final state). At the same time, the area R2 of the added disk 2-3 is expanded from the smallest LBA (the area R2 is in an intermediate state).
Further, similarly to the above, as shown in FIG. 6C, the expanded area R2 of the added disk 2-3 and the area R2 ′ of the added disk 2-4 are mirrored (areas R2, R2). 'Is the final state). At the same time, the area R3 of the added disk 2-3 is expanded from the larger LBA (area R3 is in an intermediate state).
Thereafter, the same processing is performed until the expansion processing for all areas is completed.
In this case, since the primary and secondary disk groups may try to process the same LBA range near the midpoint of the RAID group, the LBA range that either the primary or secondary is trying to process is the other. If the expansion is in progress, it is necessary to wait until it is finished.

It is a figure which shows schematic structure of the RAID apparatus of the Example of this invention. It is a figure explaining the case where a disk is added to RAID1 or RAID0 + 1 and capacity is expanded. It is a figure explaining the LDE process of the Example of this invention. It is a flowchart which shows the LDE processing procedure of the Example of this invention. It is a figure explaining the follow-up system in the Example of this invention. It is a figure explaining the back-and-forth system in the Example of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 RAID controller 1a Buffer 1b Capacity expansion means 1c Configuration information 2-1 to 2-n Disk 3 Host

Claims (5)

A RAID group consisting of 2m (m is a natural number) disk groups, where m disk groups and other m disk groups are in a mirroring relationship, and each m disk group is striped. In addition, 2n (n is a natural number) disk groups are added to form 2 (m + n) disks, and m + n disk groups and other m + n disk groups are in a mirroring relationship, and each m + n disk group A capacity expansion method in a RAID device that converts a disk group into a RAID group consisting of striped disk groups,
A first step of adding n disks to the m disk group of the 2m disk group and expanding the capacity by the n disks;
The second step of adding the remaining n disks to the remaining m disk groups out of the 2m disks to have a mirroring relationship with the disk group whose capacity has been expanded is provided. A capacity expansion method in the RAID device. A RAID group consisting of 2m (m is a natural number) disk groups, where m disk groups and other m disk groups are in a mirroring relationship, and each m disk group is striped. A RAID device comprising:
The RAID device includes capacity expansion means for expanding the disk capacity,
When adding 2n (n is a natural number) disk group to the 2m disk group, the capacity expansion means first adds n disks to the m disk group as the primary disk group. Expand the disk capacity of the primary disk group,
A RAID apparatus, wherein the remaining n disk groups are added to the other m disks, which are secondary disk groups, so as to have a mirroring relationship with the primary disk group. 3. The RAID device according to claim 2, wherein the capacity expanding means prevents the primary disk group and the secondary disk group from expanding simultaneously for the same logical block address. 4. The RAID device according to claim 3, wherein the secondary disk group is expanded so as to follow the expansion of the primary disk group. 4. The RAID according to claim 3, wherein the primary disk group is expanded from the smaller or larger logical block address, and the secondary disk group is expanded from the larger or smaller logical block address opposite to the above. apparatus.

Images