Patent Application
20230401010
The present invention relates to a storage system and to a data layout conversion method in a storage system.
As described in Japanese Patent Laid-open No. 2022-55102, a storage program running in a software-defined storage (SDS), which is a storage device, has information regarding a data layout in a drive that is used at the time of storing user data in a self-recognized drive.
In a case where the storage program uses a new function, a drive data layout may require a format change. Meanwhile, the SDS needs to continuously provide an input/output (I/O) function unless an explicit stop instruction is issued by a user. Therefore, even in a case where a function is added such that the drive data layout requires a format change, the I/O function needs to continuously be provided to the user.
The drive data layout is determined at the time when the storage program recognizes a drive in which the user data is to be stored. In a case where the drive data layout subsequently requires a format change, it is demanded that the format change be followed while maintaining the access to the user data already stored in a layout based on an old format.
In view of the above circumstances, the present invention has been made to achieve performance improvement, quality enhancement, and function addition requiring a drive data layout change while maintaining I/O functionality in a storage system.
In order to address the above-described problem, according to an aspect of the present invention, there is provided a storage system including a plurality of nodes. The nodes each include a storage drive and a processor. The processor processes data that is to be stored in the storage drive in response to an I/O request from a host. The processor included in the node creates a redundancy group for forming a redundant configuration of data to be stored in the storage drive, and stores data of the redundancy group in a plurality of the storage drive. When changing a data format of a unit storage area, the processor included in the node discards data stored in the unit storage area where part of the data of the redundancy group is stored, then based on the data of the redundancy group that is stored in an area other than the unit storage area where the data format is to be changed, recovers the discarded data according to a changed data format and re-stores the recovered data in the storage drive, and performs an I/O process on the data to be discarded during the data format change, by achieving recovery based on the data of the redundancy group that is stored in an area other than the unit storage area where the data format is to be changed.
The present invention makes it possible, for example, to achieve performance improvement, quality enhancement, and function addition requiring a data layout change while maintaining I/O functionality in a storage system.
Embodiments of the present invention will now be described with reference to the accompanying drawings. The embodiments described below and the accompanying drawings are merely illustrative, and not restrictive, of the present invention, and are simplified or omitted as needed for clarity of the description. The present invention may be implemented by various other embodiments. Unless otherwise stated, the number of individual components described in conjunction with the embodiments may be either one or more.
In a case where there are a plurality of components having an identical or similar function, they may be described with different additional characters appended to the same reference numeral. Meanwhile, in a case where the plurality of such components need not be distinguished from each other, they may be described with the additional characters omitted.
Further, the components identical or similar to those described in conjunction with an already described embodiment or modification may be omitted from the description in order to mainly describe differences. Furthermore, the components identical or similar to those described in conjunction with an already described embodiment or modification may be designated by the same reference numeral as the corresponding components in order to avoid redundant description.
In the following description of the embodiments, a process performed by executing a program is described in some cases. In such cases, a computer performs the process defined by the program while allowing a processor (e.g., a central processing unit (CPU) or a graphics processing unit (GPU)) to use, for example, a storage resource (e.g., a memory) and an interface device (e.g., a communication port). Therefore, the processor may act as a main element that performs the process by executing the program. Likewise, the main element configured to perform the process by executing the program may be a controller, apparatus, system, calculator, or node having the processor. The main element configured to perform the process by executing the program should be an arithmetic section, and may include a dedicated circuit configured to perform a specific process. Here, the dedicated circuit is, for example, a field-programmable gate array (FPGA), an application-specific integrated circuit (ASIC), or a complex programmable logic device (CPLD).
The program may be installed in a calculator from a program source. The program source may be a non-transitory storage medium that can be read by a program distribution server or the calculator. In a case where the program distribution server is adopted as the program source, the program distribution server may include the processor and the storage resource storing the program for distribution, and the processor included in the program distribution server may distribute the program for distribution to another calculator. Further, in the embodiments, two or more programs may be implemented as one program, and one program may be implemented as two or more programs.
In the description of the following embodiments, various kinds of information are presented in tabular form. However, the various kinds of information may be in a form other than a table.
Some or all of the storage nodes 110, the compute nodes 120, and the control node 130 may be identical computer equipment. Some or all of the back-end network 140, the storage service network 150, and the management network 160 may be identical networks.
The storage nodes 110 each include one or more CPUs 111, a memory 112, and one or more drives 113.
The control node 130 is computer equipment that a system administrator uses in order to manage the storage system 1. The control node 130 is able to get a plurality of storage nodes 110 together and manage them as a group called a cluster. In the example depicted in
Further, the memory 112 includes a mapping table 250 and a chunk state management table 260.
The storage control program 220 and other storage control programs 220 disposed in other storage nodes 110 are managed together as a redundancy group for forming a redundant configuration.
The redundancy group is set in such a manner that one storage control program 220 receives I/Os from the compute nodes 120, and that the remaining storage control programs 220 do not receive the I/Os. A state where I/Os are to be received is hereinafter referred to as being active, and a state where the I/Os are not to be received is hereinafter referred to as being standby. Further, a storage control program 220 set to be active is hereinafter referred to as the active storage control program, and a storage control program 220 set to be standby is hereinafter referred to as the standby storage control program.
The redundancy group is such that, in a case where a fault occurs in the active storage control program, the setting of the active storage control program is changed to standby, and the setting of the standby storage control program is changed to active. Additionally, in a case, for example, where a fault occurs in a storage node 110 where the active storage control program is disposed, the setting of the active storage control program is changed from active to standby, and the setting of the standby storage control program is changed from standby to active.
Consequently, in the redundancy group including the active storage control program, failover occurs from the active storage control program to the standby storage control program. That is, an I/O process that has been performed by a faulty active storage control program is taken over by a storage control program 220 whose setting is changed from standby to active.
The capacity control program 230 assigns a physical storage area of a drive 113 included in a storage node 110 to the redundancy group. When the physical storage area of the drive 113 is assigned to the redundancy group, the capacity control program 230 exchanges data with another capacity control program 230 of another storage node 110 belonging to the redundancy group through the back-end network 140. Via the capacity control program 230, the active storage control program performs I/O to the storage area belonging to the redundancy group, according to an I/O command.
The capacity control program 230 manages the physical storage area provided by a drive 113 included in a storage node 110 by dividing it into physical storage areas of a predetermined size. The resulting physical storage areas of the predetermined size are hereinafter referred to as the physical chunks 310.
In
The capacity control program 230 creates a capacity pool 320 for each redundancy group. Logical storage areas equal in size to the physical chunks 310 are assigned to the capacity pool 320. The logical storage areas are hereinafter referred to as the logical chunks 330. The physical chunks 310 are associated with the logical chunks 330 according to redundancy. When drives 113 usable in the redundancy group are assigned to the capacity pool 320, the capacity control program 230 determines links between the physical chunks 310 and the logical chunks 330.
For example, the capacity control program 230 creates the capacity pool 320 that links to each redundancy group. The logical chunks 330 equal in size to the physical chunks 310 are assigned to the capacity pool 320. The links between the physical chunks 310 and the logical chunks 330 are managed by the mapping table 250.
As depicted in
For example, a “drive 113 identified by D1” is hereinafter referred to as the “drive D1.” Further, for example, a “logical chunk 330 identified by L11” is hereinafter referred to as the “logical chunk L11.” Furthermore, for example, a “physical chunk 310 stored at a location identified by P1 in a drive 113” is hereinafter referred to as the “physical chunk P1.”
For example, information 371 stored in the mapping table 250 indicates that the logical chunk L11 is linked to the physical chunk P1 of the drive D1. Further, information 372 indicates that the logical chunk L11 is also linked to the physical chunk P2 of the drive D4.
One or more virtual logical volumes having the logical chunks 330 (hereinafter referred to as the virtual volumes 340) are defined for the capacity pool 320. The virtual volumes 340 are provided to the compute nodes 120.
A processing flow for writing data from a compute node 120 to a virtual volume 340 will now be described. The compute node 120 transmits, to one of the storage nodes 110 in a cluster, a write request and write data that specify an identifier of the virtual volume 340 at a data write destination and address information identifying the data write destination in the virtual volume 340.
The front-end driver 210 of the storage node 110, which has received the write request and the write data, transfers the write request and the write data to the front-end driver 210 of a storage node 110 where the active storage control program associated with the write target virtual volume 340 specified by the write request is disposed.
The front-end driver 210 having received the transferred write request and write data transmits the write request and the write data to a storage control program 220 that is associated, through the capacity pool 320, with the write target virtual volume 340 specified by the write request.
The active storage control program included in the storage control program 220 having received the write request and the write data assigns, as needed, one of the logical chunks 330 forming the capacity pool 320 associated with the virtual volume 340 to a write destination area in the write target virtual volume 340.
Further, the active storage control program converts the address of the write destination area in the write target virtual volume 340 specified by the write request to a chunk number and an offset location that identify the logical chunk 330 assigned to the write destination area. The active storage control program generates an I/O command by using the information obtained by the conversion, and transmits the generated I/O command and the write data to the capacity control program 230 in the local storage node 110.
Upon receiving the I/O command and the write data, the capacity control program 230 stores the data at the offset location in the drive 113 that corresponds to the physical chunk 310 associated with the logical chunk 330 specified by the I/O command.
In the storage system 1, the write data from the compute node 120 is redundantly stored in a plurality of physical chunks 310 assigned to the corresponding logical chunks 330. The number of physical chunks 310 to be assigned to the logical chunks 330 is determined by the settings of a redundancy method. Mirroring (user data replication), erasure coding, multi-stage erasure coding (MEC), and various other redundancy methods are available. As many physical chunks 310 as needed for implementing a selected redundancy method are associated with one logical chunk 330.
When a plurality of physical chunks 310 are associated with one logical chunk 330, the data is redundantly stored in the plurality of physical chunks 310. In this instance, one of the plurality of physical chunks 310 is set as a master, and the other physical chunks 310 are set as mirrors. In a case where a read request is issued by the compute node 120, the data is read from the physical chunk 310 that is set as the master. Further, in a case where erasure coding or MEC is used for redundancy purposes, a plurality of physical chunks 310 are associated with one logical chunk 330, and master data and redundant data having a predetermined pattern are stored in the plurality of physical chunks 310.
The SDS may occasionally store a warranty code generated from user data in the drive 113 for the purpose of protecting the user data stored in the drive. The user data and the warranty code can separately be stored after being organized into a size equal to or larger than the size of a physical chunk 310 or into a size equal to or smaller than the size of the physical chunk 310 ((a) of
In some other cases, new functions may be provided by an update of the storage system 1, or unused functions may newly be used due to changes in the settings of the storage system 1. In a case where the newly used functions require the storage of metadata used by a program in the SDS on the drive, the data format of the drive is changed in order to start using the new functions. The format of the data of the drive is also changed in a case, for example, where the use of new functions or the extension of functions newly requires a user data write log or entails a change in the storage format (log size or log format).
Likewise, a change in the size of a physical chunk, deallocation of a certain area of the drive for another program (switching between coexistence and separation of system files and user data), allocation, resizing, and deallocation of a memory data save area may occur. In these cases, too, the format of the data of the drive is changed.
Meanwhile, in the storage system 1, the correspondence between the logical chunks 330 and the physical chunks 310 is determined at the time when a drive 113 included in the storage node 110 is assigned to the capacity pool 320. This correspondence is stored in the mapping table 250. Therefore, in a case where the layout of the physical chunks 310 of the drive 113 is to be changed, the mapping table 250 is changed while maintaining access to the user data.
The chunk state management table 260 stores information regarding usage states of the physical chunks 310, which are identified based on drive information 410 regarding drives 113 assigned to the capacity pool 320 and on location information 420 identifying the locations of the physical chunks 310 of the drives 113.
Five different usage states can be set for the physical chunks 310, namely, “unused,” “old,” “new,” “conversion from old in progress,” and “conversion to new in progress.” “Unused” indicates a state where the associated area is not used. “Old” indicates a state where an old format is used. “New” indicates a state where a new format is used. “Conversion from old in progress” indicates a state where a conversion from an old format to a new format is in progress. “Conversion to new in progress” indicates a state of a conversion destination. The number of columns of the drive information 410 correlates with the number of columns of the mapping table 250. In a case where map information is added to the mapping table 250, the map information regarding the associated area is added. In a case where the map information is deleted, the map information regarding the associated area is deleted.
For example, while a map conversion required for a layout conversion of the physical chunks 310 of the drive D6 is in progress, the mapping table 250 and the chunk state management table 260 have information 431 based on an old layout and information 432 based on a new layout.
As depicted in
For the drive D6, a new layout 512 is created based on the new format, and added to the mapping table 250 (
More specifically, as depicted in the mapping table 250 (
Subsequently, as depicted in
When the copy process starts, the copy source area of the chunk state management table 260 (
When the copy process is completed on the data stored in the physical chunk P1 in the old layout 511, the copy source area of the chunk state management table 260 is changed to “unused,” and the copy destination area is changed to “new.”
Subsequently, when all the areas in the old layout 511 of the drive D6 become unused, the old layout 511 based on the old format is deleted from the mapping table 250 and the chunk state management table 260.
Due to a drive data layout conversion process performed on an individual chunk basis, the new layout and the old layout are equal in the data stored in the drives 113 included in the storage node 110, but are different in storage location because of a storage location change. The redundant data is stored in the drives 113 included in a separate storage node 110.
As depicted, for example, in
First of all, in step S101, the storage control program 220 determines whether a storage node 110 having physical chunks 310 using the old format exists in the redundancy group that performs a format conversion. If a relevant storage node 110 exists (“YES” at step S101), the storage control program 220 proceeds to step S102. Meanwhile, if no relevant storage node 110 exists (“NO” at step S101), the storage control program 220 terminates the drive data layout conversion process that is done on an individual chunk basis.
In step S102, the storage control program 220 specifies the capacity pool and transmits the format update instruction to one of the storage nodes 110 having physical chunks 310 using the old format. The storage node 110 (storage control program 220) at a transmission source of the format update instruction may be the same as or different from the storage node 110 (storage control program 220) at a transmission destination of the format update instruction.
Next, in step S103, the storage control program 220 that has received the format update instruction determines whether a drive 113 using the old format exists in the capacity pool specified by the format update instruction. If a relevant drive 113 exists (“YES” at step S103), the storage control program 220 proceeds to step S104. Meanwhile, if no relevant storage node 110 exists (“NO” at step S103), the storage control program 220 returns to step S101.
In step S104, the storage control program 220 references the chunk state management table 260 to check the drive usage state of the relevant drive 113 that uses the old format. Next, in step S105, the storage control program 220 selects one drive 113 that uses the old format, adds a new layout of the selected drive 113, which is based on a new format, to the mapping table 250, and adds a drive usage state based on the new format to the chunk state management table 260.
Here, the new layout added to the mapping table 250 not only keeps the redundancy, but also assigns logical chunks corresponding to physical chunks used in the old layout to physical chunks corresponding to an unused storage area in the old layout.
Next, in step S106, the storage control program 220 determines whether used physical chunks in the old layout exist in the drive 113 associated with “YES” at step S103. If used physical chunks exist, the storage control program 220 confirms the logical chunks corresponding to the used physical chunks in the old layout. The storage control program 220 then confirms the physical chunks corresponding to the relevant logical chunks. Subsequently, the storage control program 220 checks whether the relevant physical chunks are used in the old layout. If the relevant physical chunks are used in the old layout (“YES” at step S106), the storage control program 220 proceeds to step S107. Meanwhile, if the relevant physical chunks are not used in the old layout (“NO” at step S106), the storage control program 220 proceeds to step S112.
In step S107, the storage control program 220 allocates a free space by transferring the data in the physical chunks used in the old layout to another area. In this instance, the storage control program 220 also changes the mapping between the physical chunks and the logical chunks according to the transfer of the data.
Next, in step S108, the storage control program 220 starts the copy process of copying data in the used physical chunks in the old layout to physical chunks that correspond in the new layout to the logical chunks corresponding to the used physical chunks in the old layout. In the copy process, the storage control program 220 writes the data in the new format at a copy destination. Next, in step S109, the storage control program 220 updates the copy source area of the chunk state management table 260 to “conversion from old in progress,” and updates the copy destination area to “conversion to new in progress.”
Next, in step S110, the storage control program 220 completes the copy process of copying data.
Next, in step S111, the storage control program 220 changes the copy source area of the chunk state management table 260 to “unused,” and changes the copy destination area to “new.” Updating the copy source area of the chunk state management table 260 to “unused” is equivalent to discarding the data stored in the copy source area. Upon completion of step S111, the storage control program 222 proceeds to step S106, and determines whether a used physical chunk other than the used physical chunks confirmed in the previously performed step S106 exists.
In step S112, the storage control program 220 deletes the information regarding the relevant drive 113 from the mapping table 250 and chunk state management table 260 based on the old format. Upon completion of step S112, the storage control program 220 proceeds to step S103.
An I/O process performed during a format change will now be described.
When performing an I/O process during a format change, the capacity control program 230 (
The capacity control program 230 selectively uses the old-format control mechanism 231 (
The capacity control program 230 performs the I/O process by using the old-format control mechanism 231 when the physical chunks 310 at the I/O destination are in the “unused” or “old” state, or by using the new-format control mechanism 232 when the physical chunks 310 at the I/O destination are in the “new” state.
Further, during the copy process (steps S108 to S110 (
More specifically, when the state of the physical chunks 310 at the I/O destination is “conversion from old in progress,” the capacity control program 230 performs a write I/O with respect to the physical chunks 310 at a copy source by using the old-format control mechanism 231. Further, similarly, when the state of the physical chunks 310 at the I/O destination is “conversion from old in progress,” the capacity control program 230 performs a write I/O with respect to the physical chunks 310 at the copy destination by using the new-format control mechanism 232.
Meanwhile, when the state of the physical chunks 310 at the I/O destination is “conversion from old in progress,” the capacity control program 230 performs a read I/O with respect to the physical chunks 310 at the copy source by using the old-format control mechanism 231.
As described above, the mapping table 250 contains old-layout information based on the old format and new-layout information based on the new format such that the old-format control mechanism 231 and the new-format control mechanism 232 are selectively used. Therefore, even in a case where an old-format area and a new-format area coexist during a data format conversion, the old-format area and the new-format area are both accessible.
An unused area in the old layout is assigned to a storage area serving as the copy destination in the new layout. In a case where logical chunks 330 corresponding to physical chunks 310 in an unused area in the old layout are assigned to the virtual volume 340 and a write I/O occurs during a format change, the physical chunks 310 corresponding to the logical chunks 330 are changed. That is, the physical chunks 310 corresponding to the logical chunks 330 are changed to separate unused physical chunks 310 in the same storage node 110. This subsequently causes a change in the map information regarding the logical chunks 330 with respect to the physical chunks 310 in the mapping table 250 and the chunk state management table 260.
An I/O process during a format change with respect to data discarded for the format change is performed with respect to copying of the discarded data in a case where the discarded data is to be copied to another physical chunk. Meanwhile, in a case where the discarded data is not to be copied to another physical chunk, the I/O process is performed with respect to data recovered from the discarded data according to the redundant data of the discarded data stored in a physical chunk other than the physical chunk targeted for the format change.
When the format of a drive data layout is changed, the storage control program 220 is unable to access data that is stored in a layout based on the old format. Therefore, the data stored in the layout based on the old format is transferred to another area (e.g., physical chunks 310 or storage drive 113), then changed to a layout based on the new format, and stored anew in the new layout.
Further, during a layout conversion, old/new-layout information (mapping table 250) regarding a data storage destination is managed. Upon receiving an I/O request, the storage control program 220 references the old/new-layout information, and executes the I/O request by performing an I/O process based on the layout of a data storage area at the access destination specified by the I/O request.
Consequently, the first embodiment is able to update the storage system 1 for performance improvement, quality enhancement, and function addition requiring a data layout change while continuously performing an I/O process specified by a host. For example, random read performance improves when the data interchange format (DIF) storage method is changed from a separate method to a continuous storage method.
A second embodiment of the present invention makes a layout change on an individual drive basis by temporarily disconnecting a drive 113 targeted for a format change from the storage system 1, changing the format of the drive 113, and reconnecting the drive 113 to the storage system 1.
Data in used physical chunks 310 in a drive 113 targeted for the conversion is transferred to unused physical chunks 310 in another drive 113 in the same storage node N3 as for the drive 113 targeted for the conversion. In the transfer destination drive 113, either the new format or the old format may be used. Further, one or more transfer destination drives 113 may be involved in the conversion.
As depicted, for example, in
When the copy processes 611 and 622 are completed, the information related to the copy source and copy destination physical chunks 310 and stored in the mapping table 250 (
Before copying, for example, in the mapping table 250 (
Further, in the chunk state management table 260 (
When all pieces of the data stored in the drive D5 are completely copied to the other drive D6 in the same storage node N3, a new layout based on the new format is generated for the drive D5, and the mapping table 250 is changed.
Subsequently, the drive data layout conversion on an individual drive basis is continued until all the drives 113 in the storage node 110 (storage node N3) are in the new layout based on the new format.
When the drive data layout conversion is performed on an individual drive basis, the data storage area is changed to another drive 113 in the same storage node 110 as compared with a state prior to the format conversion. As regards the data whose storage area is changed, data redundancy is kept because the redundant data is stored in the drive 113 in another storage node 110.
An I/O process during the drive data layout conversion is performed in a manner similar to that in the first embodiment.
Further, since a format conversion is performed on the basis of individual drives 113, the chunk state management table 260 may be substituted by a state management table 260B depicted in
Steps S101, S102, and S203 are the same as steps S101, S102, and S103 described in conjunction with the first embodiment, respectively.
In step S204, the storage control program 220 determines whether physical chunks included in a drive 113 that belongs to the capacity pool specified by the format update instruction and that uses the old format are currently used. If the physical chunks are currently used (“YES” at step S204), the storage control program 220 proceeds to step S205. Meanwhile, if the physical chunks are not currently used (“NO” at step S204), the storage control program 220 proceeds to step S207.
In step S205, the storage control program 220 performs a copy process of copying the data in the physical chunks determined in step S204 to be currently used to another drive 113 in the same storage node 110. In the copy process, the storage control program 220 writes the data at the copy destination in a new format. Upon completion of physical chunk copying, the storage control program 220 proceeds to step S206, and changes the mapping table 250 and the chunk state management table 260 according to a configuration formed after copying. In this instance, updating the copy source area in the chunk state management table 260 to “unused” according to the configuration formed after copying is equivalent to discarding the data stored in the copy source area. Upon completion of step S206, the storage control program 220 returns to step S204.
In step S207, the storage control program 220 changes the mapping table 250 and the chunk state management table 260 according to the new format. Upon completion of step S207, the storage control program 220 returns to step S203.
Before a data format change, the storage drive 113 or storage node 110 targeted for the data format change is disconnected from the storage system 1B, and then, upon completion of the data format change, the disconnected storage drive 113 or storage node 110 is connected to the storage system 1B. Therefore, a data format conversion can efficiently be performed in an aggregate, such as the storage drive 113 or the storage node 110.
In a modification of the second embodiment, not the drive but the storage node is temporarily disconnected from the storage system 1, and after a format conversion, reconnected to the storage system 1, thereby performing a layout change. In this instance, the mapping is changed after the data stored in all the drives belonging to a capacity pool that is included in the relevant storage node and that has specified redundancy is copied to a drive in another storage node belonging to the capacity pool.
In a third embodiment of the present invention, the reception of I/Os is temporarily stopped, and then a format conversion is performed on an individual storage node basis to make a layout change on an individual storage node basis.
For example, in a case where an I/O is issued to a logical chunk 330 corresponding to the master physical chunk 310 existing in a drive 113 in the stopped storage node 110, an I/O is performed with respect to a mirror physical chunk 310 existing in a drive 113 in a storage node 110 that is not stopped. When the stopped storage node 110 returns to the cluster, the I/O destination is changed to the master physical chunk 310.
In the above instance, in order to ensure that the data stored in the master physical chunk 310 and the data stored in the mirror physical chunk 310 agree with each other, the data is copied to the master physical chunk 310 in the stopped storage node 110 from the corresponding mirror physical chunk 310. Meanwhile, the data is copied to a mirror physical chunk in the stopped storage node 110 from the master physical chunk 310 in another corresponding storage node 110. This function is referred to as the rebuild function.
Alternatively, a data difference that has occurred between stop and recovery is copied to the physical chunk 310 in the stopped storage node 110. This function is referred to as the difference rebuild function.
The present embodiment causes the storage node 110 to stop the reception of I/Os, and then uses the rebuild function (or the difference rebuild function) to perform a layout conversion on an individual storage node basis.
First of all, the I/O reception of the storage node N3 targeted for the conversion is stopped. In a case where an I/O is issued during the layout conversion with respect to a physical chunk 310 whose master exists in a drive 113 in the storage node N3, the I/O is performed with respect to a mirror physical chunk 310 on the basis of the behavior of the stopped storage node N3.
When the I/O reception of the storage node N3 is stopped, the format used by a drive 113 belonging to the target redundancy group is converted to update the layout of the mapping table 250. When the layout update of the mapping table 250 is completed, the storage node N3 resumes I/Os. On the basis of the behavior of the storage node N3, which has resumed the I/Os, the I/O destination is changed to the master physical chunk 310 in the storage node N3 in the new format.
Further, when the I/Os resume, all pieces of the data in the mirror physical chunk 310 are reflected in the master physical chunk 310 in the storage node N3 in the new layout.
For example, in a case where master data C, data D, and data F are stored in the storage node N3 before the I/Os are stopped (refer to 530 in
In the present embodiment, too, the state management table 260B (
Steps S101 and S102 are the same as those described in conjunction with the first and second embodiments.
In step S303, the storage control program 220 having received the format update instruction stops the reception of I/Os with respect to a capacity pool specified by the format update instruction. Next, in step S304, the storage control program 220 changes the mapping table 250 of the capacity pool specified by the format update instruction. For example, in a case where the storage node 110 includes the drive D5 and the drive D6, the layout of these two drives 113 is changed to a new layout based on the new format.
Next, in step S305, the storage control program 220 updates the chunk state management table 260 of the capacity pool specified by the format update instruction. For example, in a case where the storage node 110 includes the drive D5 and the drive D6, the state information regarding the physical chunks in these two drives 113 is changed from “old” to “new.” In this instance, updating the copy source area of the chunk state management table 260 to “new” according to the configuration formed after copying is equivalent to discarding “old” data stored in the copy source area.
Next, in step S306, the storage control program 220 reports the updates of the mapping table 250 and the chunk state management table 260, to the other storage nodes 110 belonging to the same redundancy group as the capacity pool specified by the format update instruction. Next, in step S307, the storage control program 220 resumes the reception of I/Os, which has been stopped in step S303.
Next, in step S308, the storage control program 220 performs a rebuild process. Specifically, the rebuild process is performed in such a manner that the data in the mirror physical chunks 310 in the other storage nodes 110 is copied to the master physical chunk 310 in the storage node 110 targeted for the format change. In this copy process, the storage control program 220 writes the data to the copy destination in the new format. Further, the master physical chunks 310 in the other storage nodes 110 are copied to the mirror physical chunks 310 in the storage node 110 targeted for the format change. Upon completion of step S308, the storage control program 220 returns to step S101.
When a data format is to be changed, the reception of I/O requests to the storage node 110 or storage drive 113 targeted for the data format change is stopped, and the mapping table 250 is updated according to the changed data format. When the update of the mapping table 250 is completed, the reception of I/O requests is resumed. Further, the data discarded from the storage area targeted for the data format change is recovered by exercising the rebuild function for the data. Consequently, since a format conversion is performed in an aggregate, such as the storage drive 113 or the storage node 110, and recovery is achieved by exercising the rebuild function for originally existing redundant data, a data format conversion can efficiently be performed with a simple configuration.
In a modification of the third embodiment, the reception of I/Os can be stopped on an individual drive basis, and not on an individual node basis, and a layout change can be made on an individual drive basis. In this case, too, data recovery is achieved by exercising the rebuild function or the difference rebuild function after drive recovery.
The present invention is not limited to the foregoing embodiments, and extends to various modifications. For example, the foregoing embodiments have been described in detail to facilitate the understanding of the present invention, and are not necessarily limited to configurations including all the described components. Further, unless inconsistency arises, some of the components in a certain embodiment may be replaced by the components in another embodiment, and some or all of the components in a certain embodiment may be added to some or all of the components in another embodiment. Further, some of the components in each embodiment may be subjected to addition, deletion, replacement, integration, or dispersion. Moreover, the components and processes described in conjunction with the embodiments may be dispersed, integrated, or interchanged as appropriate based on its processing efficiency or implementation efficiency.
| Number | Date | Country | Kind |
|---|---|---|---|
| 2022-095168 | Jun 2022 | JP | national |
