JP2009037564A - Disk management program, disk management method, and disk management apparatus - Google Patents

Abstract

An object of the present invention is to enable efficient file relocation in a global file system.
An area management metadata server machine 30 determines whether or not a file is in a fragmented state when any of the node machines 20 to 50 inquires about metadata accompanying the file access. (Steps S105 to S108) If it is in a fragmented state, information about the file is registered in the fragmented file information management table 73 (Step S109). In addition, in parallel with this operation, the area management metadata server machine 30 registers information in the fragmented file information management table 73 when the operator gives an instruction to add a volume (steps S201 and S202). Files are sequentially moved from the existing disk to the added disk (steps S211, S309, S401 to S403).
[Selection] Figure 13

Description

  The present invention relates to a program, method and apparatus for managing a disk.

  As is well known, the storage area of the disk in the storage machine is finely divided into a plurality of sectors. The plurality of sectors are further grouped into blocks for each predetermined number of units by the file system, and each block is managed with an address. Usually, since the data amount of a file is larger than the storage capacity of one block, when the file system records a file on a disk, the file is written over a plurality of blocks. It is managed by the file system (block management) which block contains a plurality of unit data constituting the file. Each file is classified hierarchically in order to improve user organization, and directory information related to each hierarchy is also managed by the file system (name management).

  Files are written in continuous blocks on a disk immediately after the start of use, but are often written in a discontinuous block on a disk that has been used to some extent. This is because, in a disk that has been used to some extent, file writing, overwriting (ie, increasing or decreasing the number of allocated blocks), and erasing are repeated, resulting in free blocks in some places. A state in which a file is written in discontinuous blocks in this way is called a fragmentation state (fragmentation). If the file on the disk is fragmented in this way, seek of unit data in file access is performed. Takes time.

  Conventionally, a defragmentation function is used to eliminate such a decrease in file access speed. As is well known, the defragmentation function is a function for rearranging fragmented files from discontinuous blocks to continuous blocks. Specifically, the defragmentation function secures a continuous block by erasing the unit data of a fragmented file from a discontinuous block and temporarily replicating it to another block, and then it is once replicated The unit data is transferred to the secured continuous block.

  By the way, the conventional file system completes the name space (file and its hierarchical classification (directory) system) in the disk, and thus lacks reliability and expandability. In recent years, global file systems have been developed to improve the reliability and expandability of file systems.

  The global file system is a file system in which the same metadata and the same disk are shared by a plurality of node machines constituting the cluster system. Metadata mainly consists of area information that defines the relationship between files and disk blocks, and name information that defines a name space, and is managed in a metadata server machine connected to the same network as each node machine. . That is, in the global file system, the metadata server machine is in charge of area management and name management among area management, name management, and file operation control, which are the three main tasks of the file system. On the other hand, each node machine has only a file operation control function and uses metadata as a client.

  In a disk used for a plurality of node machines by such a global file system, the above-mentioned fragmentation often occurs when the free space becomes insufficient. For this reason, when the free space is insufficient, it is necessary to rearrange files by the above-described defragmentation function.

  However, since there is not enough free space, there are few blocks that temporarily replicate unit data when relocating the unit data constituting a fragmented file, and there is a problem that defragmentation cannot be performed efficiently. .

  Patent Document 1 discloses a technique for transferring audio data files fragmented in a disk to a disk of a viewing machine in a short time while rearranging unit data. . However, the technique according to Patent Document 1 relates to data transfer, and does not relocate fragmented files within the same volume.

JP 2006-040227 A (paragraph 0040)

  The present invention has been made in view of the problems of the prior art as described above, and an object of the present invention is to enable efficient file relocation in a global file system.

  The disk management program devised to solve the above problems records area information defining the relationship between files and disk blocks, metadata including name information defining a name space, and files. In a cluster system with a global file system that shares a shared disk with multiple node machines, when a metadata server machine receives a metadata inquiry from one of the node machines, the file becomes fragmented. A determination unit that determines whether or not the file is in a state; a registration unit that registers information about the file in a table in the storage device when the determination unit determines that the file is in a fragmented state; Instruct the operator to add a volume with the input device Receiving accepting means, and when the accepting means accepts an instruction to add a volume, a move for sequentially moving a file whose information is registered in the table from an existing disk to a disk specified by the designation accepted by the accepting means It is characterized by functioning as a means.

  When configured in this way, the metadata server machine determines whether or not the file is in a fragmented state when any of the node machines inquires about metadata accompanying the file access. If so, it operates to register information about the file in the table. Also, in parallel with this operation, the metadata server machine, in response to an instruction from the operator to add a volume, sequentially transfers files whose information is registered in the table from the existing disk to the added disk. Operates to move. According to such an operation, files that are in a fragmented state in the existing disk are collectively written in continuous blocks as they are moved into the added disk. In other words, even if there is not enough free space on the disk, the files can be relocated efficiently.

  In addition, the disk management method devised to solve the above-mentioned problem is that the area information defining the relationship between the file and the disk block, the metadata including the name information defining the name space, and the file In a cluster system that implements a global file system that shares a recorded disk with multiple node machines, if the metadata server machine receives file access from any of the node machines, is the file fragmented? Judgment procedure for determining whether or not, and if the determination procedure determines that the file is in a fragmented state, a registration procedure for registering information about the file in a table in the storage device, and adding a volume along with the specification of an arbitrary disk Accepting instructions from the operator through the input device, and When the volume addition instruction is accepted in the acceptance procedure, the migration procedure for moving the files whose information is registered in the table from the existing disk to the disk specified by the designation accepted in the acceptance procedure is executed. , With features.

  Therefore, the metadata server machine that implements this disk management method functions in the same manner as the metadata server machine on which the disk management program of the present invention operates.

  In addition, the disk management device devised to solve the above-mentioned problem is that the area information defining the relationship between the file and the disk block, the metadata including the name information defining the name space, and the file Whether or not a file is fragmented when file access is accepted from any node machine in order to manage the disk in a cluster system with a global file system that allows multiple node machines to share the recorded disk A determination unit that determines whether the file is in a fragmented state, a registration unit that registers information about the file in the table, and an instruction to add a volume together with designation of an arbitrary disk The reception unit that accepts the volume, and the reception unit receives the volume addition instruction If that, the file information table is registered, the existing disk, the disk that is identified by the designation accepted by the accepting unit, further comprising a moving unit that moves sequentially, are characterized.

  Therefore, this disk management apparatus functions in the same manner as the metadata server machine on which the disk management program of the present invention described above operates.

  As described above, according to the present invention, file relocation can be efficiently performed in the global file system.

  The best mode for carrying out the present invention will be described below with reference to the accompanying drawings.

  First, the hardware configuration and software configuration of the computer network system of this embodiment will be described.

  As shown in the network configuration diagram of FIG. 1, the computer network system of the present embodiment includes a storage machine 10, a name management metadata server machine 20, an area management metadata server machine 30, a defragmentation metadata server machine 40, It consists of a general-purpose server machine 50 and a client machine 60. Among these, the storage machine 10 and each server machine 20-50 are communicably connected via a fiber channel switch (not shown) to constitute a SAN (Storage Area Network). The server machines 20 to 50 and the client machine 60 are communicably connected via a router (not shown) to constitute a LAN [Local Area Network]. Although only one general-purpose server machine 50 is shown in FIG. 1, two or more general-purpose server machines 50 may exist. In addition, although only two storage machines 10 and client machines 60 are shown in FIG. 1, only one storage machine 10 or three or more may exist.

  Further, a global file system that provides a single file system function between different models is applied to the server machines 20 to 50 of the present embodiment. As is well known, the file system mainly performs name management, area management, and file operation control. Among these, the functions for realizing file operation control are as described later. As a global file system client, it is implemented in all server machines 20-50. On the other hand, functions for realizing name management and area management are implemented in the metadata server machines 20 to 40 as servers of a global file system, as will be described later.

  Here, functions for realizing name management and area management of the global file system are generally implemented in one metadata server machine, but in this embodiment, name management and area management are performed. Are mounted on separate machines 20 and 30. The reason why the name management and the area management are implemented in separate machines is to minimize the influence of the file relocation processing by the defragmentation function on the file access processing. This is because, in general file access processing, directory search is frequently performed and the name management function is loaded, while the area management function is relatively unloaded, and file relocation In the process, the area information is frequently updated as the disk block in which the file is recorded is changed, and the area management function is loaded. On the other hand, the name management function is not relatively loaded. It is based on the circumstances.

  Further, in the present embodiment, a defragmentation function generally included in the area management function is assigned to a machine 40 different from the area management metadata server machine 30. This is a process that puts a heavy load on file relocation processing, including data replication associated with file movement and area information update processing. If these are assigned to different server machines, the load on defragmentation is distributed. This is because it can.

  As described above, the server machines 20 to 50 constitute a system (cluster) in which a load applied to the file system is distributed, and thus can also be referred to as node machines.

  The storage machine 10 is a device for recording functions and information provided by the server machines 20 to 50 to the client machine 60. As shown in the configuration diagram of FIG. 2, the storage machine 10 includes a disk unit 10a, a storage controller 10b, and HBA [Host Bus Adapter] 10c is incorporated. The disk unit 10a is a unit that incorporates one or more disks for recording data, a disk rotation mechanism, and a read / write head. The storage controller 10b is a unit that controls reading / writing of data from / to the disk in accordance with commands from the metadata server machines 20-50. The HBA 10c is a communication adapter for exchanging data with the metadata server machines 20 to 50 according to the fiber channel protocol.

  The name management metadata server machine 20 is a metadata server machine in charge of the above-described name management. As shown in the configuration diagram of FIG. 3, a storage unit 20a, a CPU [Central Processing Unit] 20b, a memory unit 20c, The HBA 20d and the LAN adapter 20e are built in. The storage unit 20a is a unit for recording various programs and data. The CPU 20b is a unit that performs processing in accordance with a program recorded in the storage unit 20a. The memory unit 20c is a unit in which the CPU 20b caches programs and data and develops a work area. The HBA 20d is a communication adapter for exchanging data with the storage machine 10 according to the fiber channel protocol. The LAN adapter 20e is a communication adapter for exchanging data with the client machine 60 in accordance with TCP / IP (Transmission Control Protocol / Internet Protocol) and Ethernet protocol (Ethernet is a trademark of Xerox Corporation and Hewlett-Packard Company).

  The name management metadata server machine 20 stores an application 21, an operating system 22, and a file system 23 in a storage unit 20a. The application 21 is software that realizes various functions provided to the client machine 60. Examples of the function realized by the application 21 include a web server, a web application, a file server, and a mail server. The operating system 22 manages data input / output in the HBA 20d and the LAN adapter 20e, manages the storage area of the memory unit 20c, assigns windows to a plurality of tasks, multiplexes screen outputs, and various setting tools This is software for providing such basic functions to many applications including the application 21. The file system 23 is software for realizing the above-described global file system. Among the functions of the global file system, a program 23a that realizes a file operation control function as a client, and a name space (file and its hierarchy) A program 23b for realizing a function of managing a general classification (directory system). The file operation control function 23a is a program that provides the application 21 with operations related to files such as reading, writing, copying, overwriting, erasing, renaming, moving a directory, creating a directory, and deleting a directory. The name management function 23b manages name information (information on the file, information on the directory, information on the directory structure) that defines the name space (file name) in the metadata, and searches for the file name and directory name. It is a program to do.

  The area management metadata server machine 30 is a metadata server machine in charge of the area management described above. As shown in the configuration diagram of FIG. 4, the storage unit 30a, the CPU 30b, the memory unit 30c, the HBA 30d, and the LAN adapter. 30e is built in. Similarly to the name management metadata server machine 20, the area management metadata server machine 30 also stores an application 31, an operating system 32, and a file system 33 in the storage unit 30a. The application 31 and the operating system 32 are software that exhibits the same functions as the name management metadata server machine 20. The file system 33 is software for realizing the global file system described above, and includes a program 33a for realizing a file operation control function as a client among the functions of the global file system.

  The file system 33 also includes a program 33b for realizing the area management function, a fragment information collection program 33c, a volume manager 33d, and a free capacity monitoring program 33e. The area management function 33b is a program that realizes a function of managing the association between unit data constituting a file and the address of a disk block. The area management function 33b assigns an inode number as a unique identification number to a normal file and a file storing directory information, and records information about the file in an inode management table 71 as shown in FIG. To manage. The area management function 33b is configured to record and manage information related to the volume (storage area) of the disk in the storage machine 10 in a volume management table 72 as shown in FIG. These tables 71 and 72 may be stored in any one of the storage machines 10, or may be stored in the storage unit 30 a of the area management metadata server machine 30.

  The inode management table 71 has the same number of records as the file (inode). Each record includes “inode”, “size”, “number of blocks”, “number of extents”, “last access date”, “number of accesses”, and “block” as shown in the schematic table structure diagram of FIG. Each field. The “inode” field is a field in which the inode number of the file is recorded. The “size” field is a field in which the data amount of the file is recorded. The “number of blocks” field is a field in which the number of blocks occupied by the file on the disc is recorded. The “number of extents” field is a field in which the number of extents of the file is recorded. An extent refers to a block of consecutive blocks among the disk blocks in which the file is recorded. The “last access date / time” field is a field in which the date / time of the last access among the accesses to the file is recorded. The “number of accesses” field is a field in which the number of accesses to the file is recorded. The “block” field is a field in which address information specifying a disk block in which the file is recorded is recorded.

  The volume management table 72 has the same number of records as the disks in the storage machine 10. As shown in the schematic table structure diagram of FIG. 6, each record includes “volume”, “volume name”, “size”, “free capacity”, “built-in”, “automatic addition”, and “relocation”. Each field. The “volume” field is a field in which a volume number, which is a unique identification number for specifying the volume (storage area) of the disk, is recorded. The “volume name” field is a field in which the name of the volume is recorded. The “size” field is a field in which the capacity of the volume is recorded. The “free capacity” field is a field in which the capacity of the free area in the volume is recorded. The “embedded” field specifies information that specifies whether or not the volume has been incorporated in the usage target, and specifies whether or not the usage is reserved when the volume has been incorporated in the usage target. Information is a recorded field. In the “embedded” field, “not yet” indicating that the volume has not been incorporated into the usage target, and “but reserved” indicating that the volume has been incorporated into the usage target but has been reserved. “Yes” is recorded, or “Done” indicating that it is incorporated in the object of use and the reserved state of use is released is recorded. In the “automatic addition” field, when the volume is in a reserved state, the free capacity of all the volumes that have been incorporated into the target of use and whose reserved use has been released (hereinafter referred to as “in-use volumes”) is displayed. This is a field in which information indicating whether or not to cancel the reserved use of the volume when a predetermined state is reached is recorded. If a volume to be automatically added (reserved reservation is automatically released) is set, the threshold value of the free area capacity of the volume in use, which is a start condition for the automatic addition, is “automatically added” for the volume. Recorded in the field. Further, the “automatic addition” field of the record of the volume that is not incorporated in the usage target is blank. The “relocation” field is a field in which information indicating whether or not to relocate a file is recorded when the volume is automatically added (reservation is automatically released). Note that the “relocation” field of the record of the volume that is not incorporated in the usage target is blank.

  The area management function 33b in FIG. 4 newly creates a fragmented file information management table 73 as shown in FIG. 7 in the memory unit 30c after activation. The fragmented file information management table 73 is a work table for recording information related to a file determined to be in a fragmented state by processing to be described later. Therefore, immediately after activation, the fragmented file information management table 73 is in a state where no record is recorded. Each record of the fragmented file information management table 73 includes “inode”, “size”, “number of blocks”, “extent number”, and “last access date / time” as shown in the schematic table structure diagram of FIG. And “access count” fields. In each of these fields, a value equivalent to the field having the same name in the i-node management table 71 of FIG. 5 is recorded.

  Further, the fragment information collection program 33c in FIG. 4 specifies whether or not the file is in a fragmented state when the file operation control function 23a, 33a or 43a of any of the server machines 20 to 50 accesses the file. It is a program to do. The contents of processing executed by the CPU 30b according to the fragment information collection program 33c will be described later with reference to FIG.

  The volume manager 33d in FIG. 4 is a program for providing the operator with a tool for incorporating a volume into a usage target and deleting the volume from the usage target. The contents of processing executed by the CPU 30b according to the volume manager 33d will be described later with reference to FIGS.

  Also, the free capacity monitoring program 33e in FIG. 4 is a program for canceling the use suspension of the volume for which the automatic addition setting has been made when the capacity of the free area of the volume in use is in a predetermined state. . The contents of the process executed by the CPU 30b according to the free capacity monitoring program 33e will be described later with reference to FIG.

  Further, the defragmentation metadata server machine 40 in FIG. 1 is a metadata server machine in charge of file relocation processing. As shown in the configuration diagram of FIG. 8, the storage unit 40a, CPU 40b, memory unit 40c, HBA 40d, And the LAN adapter 40e is built in. The defragmentation metadata server machine 40 also stores an application 41, an operating system 42, and a file system 43 in the storage unit 40a in the same manner as the metadata server machines 20 and 30 described above. The application 41 and the operating system 42 are software that exhibits functions equivalent to those of the metadata server machines 20 and 30 described above. The file system 43 is software for realizing the global file system described above, and includes a program 43a that realizes a file operation control function as a client among the functions of the global file system. Further, the file system 43 includes a file relocation program 43b. The file relocation program 43b is registered in the fragmented file information management table 73 in FIG. 7 when the volume is manually incorporated into the use target or when the use suspension of the volume for which the automatic addition setting is made is automatically released. It is a program to move the file that is being moved to the volume. The contents of processing executed by the CPU 40b according to the file rearrangement program 43b will be described later with reference to FIG.

  The general-purpose server machine 50 in FIG. 1 is a machine that provides the client machine 60 and uses the global file system described above as a client. Accordingly, the general-purpose server machine 50 includes a storage unit, a CPU, a memory unit, an HBA, and a LAN adapter (not shown), and stores an application, an operating system, and a file system in the storage unit. ing. The file system includes a file operation control function as described above.

  The client machine 60 is a computer capable of LAN communication to which a function for enjoying the functions provided by the applications 21, 31, 41, such as a web browser, a mailer, and an emulator, is added.

  Next, processing performed in the computer network system of this embodiment will be described.

  The file operation control functions 23a, 33a, 43a included in the file systems of the server machines 20-50 are well known and will be described briefly. The file operation control functions 23a, 33a, and 43a are executed when the file system is activated or immediately after the execution is completed. When the CPU receives a file read instruction from the application after the start of the file operation control process, the CPU inquires the name management function 23b in FIG. 3 to obtain the inode number of the target file, and then uses the inode number. The area management function 33b shown in FIG. 4 is inquired, information on the block in which the target file is recorded is acquired, and the corresponding storage machine 10 to which the block belongs is accessed. When the CPU receives a file write instruction from the application after the start of the file operation control process, the CPU requests the area management function 33b in FIG. 4 to reserve a free area for writing, and then designates the designated area. A process for registering a new file in the directory is requested to the name management function 23b in FIG. Thereafter, the CPU writes the file in the reserved free space, and notifies the region management function 33b in FIG. 4 and the name management function 23b in FIG. 3 that the region has been allocated to the file. Upon receiving this notification, the area management function 33b in FIG. 4 registers information related to the file in the inode management table 71 in FIG.

  The fragment information collection program 33c in FIG. 4 is executed with the activation of the file system or immediately after the execution is completed. As shown in the flowchart of FIG. 9, after starting the fragment information collection process, the CPU 30b performs file access (file read or write) from the file operation control function 23a, 33a, 43a of any file system in the first step S101. ) Wait until there is. If there is a file access, the CPU 30b advances the process to step S102.

  In step S102, the CPU 30b specifies a record including the eye node number notified from the file operation control function 23a, 33a, 43a at the time of file access among the records in the inode management table 71 of FIG. Update the value of the “last access date” field with the current time. Further, the CPU 30b increases the value of the “number of accesses” field of the record by one.

  In the next step S103, the CPU 30b counts the number of blocks in the “block” field of the record of the file in the inode management table 71 of FIG. Subsequently, the CPU 30b determines whether or not the number of blocks obtained by counting matches the value of the “number of blocks” field of the record. If the number of blocks obtained by counting matches the value in the “number of blocks” field of the record, the CPU 30b branches the process from step S103 to S105, assuming that the number of blocks has not increased due to writing. On the other hand, if the number of blocks obtained by counting does not match the value of the “number of blocks” field of the record, the CPU 30b proceeds to step S104, assuming that the number of blocks has increased due to writing.

  In step S104, the CPU 30b updates the value of the “number of blocks” field of the record of the file in the inode management table 71 of FIG. Further, the CPU 30b counts the number of extents in the “block” field of the record, and updates the value of the “extent number” field of the record. Thereafter, the CPU 30b advances the process to step S105.

  In step S105, the CPU 30b reads a value from the “number of extents” field of the record of the file in the inode management table 71 of FIG.

  In the next step S106, the CPU 30b determines whether or not the number of extents read in step S105 exceeds a predetermined threshold value. If the number of extents exceeds the predetermined threshold, the CPU 30b advances the process to step S109. On the other hand, when the number of extents does not exceed the predetermined threshold, the CPU 30b branches the process from step S106 to S107.

  In step S107, the CPU 30b reads the value from the “number of accesses” field of the record of the file in the inode management table 71 of FIG.

  In the next step S108, the CPU 30b determines whether or not the number of accesses read in step S107 exceeds a predetermined threshold value. If the access count exceeds the predetermined threshold, the CPU 30b advances the process to step S109. On the other hand, if the number of accesses does not exceed the predetermined threshold, the CPU 30b branches the process from step S108, and ends the fragment information collection process according to FIG.

  In step S109, the CPU 30b reads the record of the file in the inode management table 71 in FIG. 5, and additionally registers it in the fragmented file information management table 73 in FIG. Thereafter, the CPU 30b ends the fragment information collection process according to FIG.

  According to the fragment information collection program 33c, each time a file is accessed, the region management metadata server machine 30 determines whether or not the file is fragmented (steps S106 and S108). If it is a fragmented file, the file information is registered in the fragmented file information management table 73 of FIG. 7 (step S109).

  The CPU 30b that executes steps S105 to S108 corresponds to the above-described determination means and the above-described determination unit, and steps S105 to S108 correspond to the above-described determination procedure. The CPU 30b executing step S109 corresponds to the above-described registration unit and the above-described registration unit, and step S109 corresponds to the above-described registration procedure.

  In parallel with this fragment information collection processing, the area management metadata server machine 30 is configured to execute the volume manager 33d by receiving an activation instruction from an operator via an input unit (not shown). . As shown in the flowcharts of FIGS. 10 and 11, the CPU 30b displays a volume setting screen on a display unit (not shown) in the first step S201 after the processing is started. As shown in the screen example of FIG. 12, the volume setting screen 74 is a configuration volume in which the names of volumes whose “include” field value is “done” or “possible” in the volume management table 72 of FIG. A column 74a, and an additional candidate volume column 74b in which the names of volumes whose values in the “embedded” field are “not yet” are listed. Further, the volume setting screen 74 removes the volume selected in the addition target volume column 74b from the usage target and the addition button 74c for instructing to incorporate the volume selected in the usage target volume column 74b from the usage target. The delete button 74d for instructing is included. Further, the volume setting screen 74 is used as an additional setting item when a volume is incorporated into a usage target, and the volume is immediately incorporated into the usage target to be used, or the volume is incorporated into the usage target and used. Radio buttons 74e and 74e for determining whether or not to cancel the reserved use of the volume when the capacity of the free area in the used volume is equal to or less than a predetermined value in the reserved state are included. The volume setting screen 74 is a text for inputting a condition for automatically adding a volume when setting to automatically add a volume (cancellation of reserved use) according to the free capacity of the volume in use. Box 74f is included. The volume setting screen 74 also includes a check box 74g for designating that file relocation is performed simultaneously when a volume is added. Further, the volume setting screen 74 includes an execution button 74h to be clicked when registering a volume with the contents set by using the buttons 74c and 74d, the radio button 74e, the text box 74f, and the check box 74g. Yes. The volume setting screen 74 includes a cancel button 74i that should be clicked when canceling these setting actions. When the volume setting screen 74 as shown in FIG. 12 is displayed on a display unit (not shown), the CPU 30b advances the process to step S202 in FIG.

  In step S202, the CPU 30b stands by until any one of the execution button 74h and the cancel button 74i on the volume setting screen 74 is clicked. If any button is clicked, the CPU 30b advances the process to step S203.

  In step S203, the CPU 30b determines whether or not the clicked button is the cancel button 74i. If the clicked button is the cancel button 74i, the CPU 30b ends the processes according to FIGS. On the other hand, if the clicked button is the execute button 74h, the CPU 30b branches the process from step S203 to step S204.

  In step S204, the CPU 30b determines whether the addition of volume (incorporation into the use target) has been selected or the deletion of volume has been selected. If deletion of a volume has been selected, the CPU 30b branches the process from step S204 to step S205.

  In step S205, the CPU 30b performs a process of removing the volume designated when the execution button 74h is clicked from the use target. In this process, the value of the “include” field of the record of the volume in the volume management table 72 of FIG. 6 is switched to “not yet”, and the registration of the volume is deleted. Thereafter, the CPU 30b ends the processes according to FIGS.

  On the other hand, when the addition of volume (incorporation into the use target) is selected in step S204, the CPU 30b advances the process to step S206.

  In step S206, when the execution button 74h is clicked, the CPU 30b determines whether or not the radio button 74e that designates that the volume is immediately added (incorporation into the usage target and transition to the usage state) has been selected. Is determined. If the radio button 74e that designates that the volume is to be added (cancellation of reserved use) when the free space capacity of the volume in use becomes equal to or smaller than the predetermined value is selected, the CPU 30b proceeds to step S206. The process branches from step S207 to step S207.

  In step S207, the CPU 30b performs a process of incorporating the volume designated when the execution button 74h is clicked into the use target. In this process, the value of the “include” field of the record of the volume in the volume management table 72 of FIG. 6 is switched to “possible”, and the volume is registered. Thereafter, the CPU 30b ends the processes according to FIGS.

  On the other hand, when the radio button 74e for designating that the volume is to be added (incorporation into the usage target and transition to the usage state) is selected in step S206, the CPU 30b advances the process to step S208.

  In step S208, the CPU 30b transmits the volume addition information all at once to all the server machines (cluster nodes) 20-50. The volume addition information is information for notifying that the capacity of the volume in use has increased.

  In the next step S209, the CPU 30b stands by until there is a response from all the server machines 20-50. And if there is a response from all the server machines 20-50, CPU30b will advance a process to step S210.

  In step S210, the CPU 30b performs processing for confirming addition of the designated volume (incorporation into the use target and transition to the use state). In this process, the value of the “include” field of the record of the volume in the volume management table 72 of FIG. 6 is switched to “done”, and the volume is registered.

  In the next step S211, the CPU 30b instructs the file system 43 of the defragmentation metadata server machine 40 of FIG. 8 to start the file relocation program 43b, and delivers the volume number of the added volume as an argument. . Thereafter, the CPU 30b ends the processes according to FIGS.

  According to the volume manager 33d, the administrator of the computer network system according to the present embodiment can incorporate and delete a volume from the usage target through the volume setting screen 74 as shown in FIG. In addition, when installing a volume, either immediately incorporate the volume into the usage target and put it in use, or install the volume into the usage target and leave it in a reserved state and use the free space of the volume in use. Using the radio buttons 74e and 74e, it is possible to select whether or not to cancel the reserved use of the volume when the capacity falls below a predetermined value. Furthermore, when the administrator specifies that the volume is to be added (cancellation of reserved use) after the in-use volume is in a predetermined state, the condition of the in-use volume to be added and the file Whether or not to perform rearrangement can be designated using the text box 74f and the check box 74g.

  The CPU 30b that executes steps S201 and S202 corresponds to the above-described reception unit and the above-described reception unit, and steps S201 and S202 correspond to the above-described reception procedure.

  In the area management metadata server machine 30, the free capacity monitoring program 33e is executed when the file system is activated or immediately after the execution is completed. As shown in the flowchart of FIG. 13, after starting the free space monitoring process, in the first step S301, the CPU 30b uses the file operation control functions 23a, 33a, 43a of any of the server machines 20 to 50 for writing. Wait until a free space reservation request is received. When receiving a request for reservation of a free area from any of the server machines 20 to 50, the CPU 30b advances the process to step S302.

  In step S302, the CPU 30b secures a free area from the in-use volume, enters an exclusive state in which no other reservation is accepted, and reads the capacity of all free areas in the in-use volume. Note that the capacity of the free area may be acquired by, for example, the sum of the values of the “free capacity” field of all records that record “done” in the “built-in” field in the volume management table 72 of FIG. .

  In the next step S303, the CPU 30b determines whether or not there is a volume for which automatic addition (automatic release of reserved reservation) is set. Specifically, the CPU 30b can detect a record in which the value of the “embedded” field is “possible” and the value of the “automatically added” field is not “impossible” among the records of the volume management table 72 of FIG. It is determined whether or not. When there is no volume for which automatic addition is set, the CPU 30b branches the process from step S303 and ends the free capacity monitoring according to FIG. On the other hand, if there is a volume for which automatic addition is set, the CPU 30b advances the process to step S304.

  In step S304, the CPU 30b reads a threshold value defined as a condition for a volume for which automatic addition (automatic cancellation of reserved use) is set. That is, the CPU 30b reads the value of the “automatic addition” field of the record detected in step S303.

  In the next step S305, the CPU 30b determines whether or not the free space read in step S302 is less than the threshold read in step S304. If the free space is greater than or equal to the threshold, the CPU 30b branches the process from step S305 and ends the free space monitoring according to FIG. On the other hand, if the free area is less than the threshold, the CPU 30b advances the process to step S306.

  In step S306, the CPU 30b transmits the volume addition information all at once to all the server machines (cluster nodes) 20-50. As described above, the volume addition information is information for notifying that the capacity of the volume in use has increased.

  In the next step S307, the CPU 30b stands by until there is a response from all the server machines 20-50. And if there is a response from all the server machines 20-50, CPU30b will advance a process to step S308.

  In step S308, the CPU 30b performs processing for confirming addition of the designated volume (cancellation of reserved use). In this process, the value of the “include” field of the record of the volume in the volume management table 72 of FIG. 6 is switched to “done”.

  In the next step S309, the CPU 30b instructs the file system 43 of the defragmentation metadata server machine 40 in FIG. 8 to start the file relocation program 43b, and delivers the volume number of the added volume as an argument. . Thereafter, the CPU 30b ends the free space monitoring according to FIG.

  According to the free capacity monitoring program 33e, when the file system of any of the server machines 20 to 50 receives a file write request from the application 21, 31, 41, a volume for which automatic addition setting (reserved use) is made. If such a volume is found (step S303), it is determined whether or not the free capacity of the volume in use is in a predetermined state based on the conditions defined for the volume (step S303). Steps S302, S304, S305) When the free capacity of the volume in use is in a predetermined state, the volume for which automatic addition setting has been made is added to the volume in use (steps S306 to S309).

  An instruction by one of the above two processes, that is, an instruction by step S211 of the process by the volume manager 33d of the area management metadata server machine 30 in FIG. 4 (FIGS. 10 and 11), or free space monitoring The CPU 40b of the defragmentation metadata server machine 40 in FIG. 8 executes the file relocation program 43b based on the instruction in step S309 of the processing by the program (FIG. 13). As shown in the flowchart of FIG. 14, after starting the file rearrangement process, the CPU 40b determines whether or not it is set to perform file rearrangement following volume addition in the first step S401. That is, the CPU 40b reads the value from the “relocation” field of the volume record of the volume number received as an argument together with the activation instruction in the volume management table 72 of FIG. 6, and determines whether the value is “necessary” or “unnecessary”. Is determined. If it is not set to relocate the file, the CPU 40b branches the process from step S401, and ends the file relocation process according to FIG. On the other hand, if it is set to relocate the file, the CPU 40b advances the process to step S402.

  In step S402, the CPU 40b performs a process of moving the files registered in the fragmented file information management table 73 in FIG. 7 one by one to the added volume. Specifically, for one file selected as a processing target from one or more files, the CPU 40b first transfers the unit data from a disk block in which unit data constituting the file is recorded to a memory unit. The process of copying to 40c is executed. Subsequently, the CPU 40b sequentially records the unit data copied to the memory unit 40c from the block with the youngest address to the old block in the added volume. Thereafter, the CPU 40b performs processing for deleting the data of the file from the volume (disk) where the unit data constituting the file to be processed was originally recorded. When the CPU 40b completes the above processing for all the files registered in the fragmented file information management table 73 in FIG. 7, the records related to these files are deleted from the fragmented file information management table 73 in FIG. Thereafter, the process proceeds to step S403.

  In step S403, the CPU 40b requests the area management function 33b in the file system 33 of the area management metadata server machine 30 in FIG. 3 to update the area information on the file rearranged in step S402. In response to this request, the area management function 33b updates information on the corresponding file in the inode management table 71 in FIG. When the CPU 40b requests the area management function 33b to update such area information, the file rearrangement process according to FIG. 14 ends.

  According to the file relocation program 43b, when there is a request from the area management metadata server machine 30 (step S211 or S309), the fragmented file in the existing disk is sequentially transferred to the added disk. It is transferred.

  In this way, when a file that has been fragmented in the existing disk is transferred to the disk to be added, the file is written in a continuous block. For this reason, even if there is not enough free space on the disk, the files can be relocated efficiently.

  The CPU 30b that executes steps S206 to S211, S301 to S309, and S401 to S403 corresponds to the moving means and the moving unit described above. Steps S206 to S211, S301 to S309, and S401 to S403 are This corresponds to the moving procedure described above.

Network configuration diagram of the computer network system of the present embodiment Storage machine configuration diagram Configuration diagram of name management metadata server machine Configuration diagram of the area management metadata server machine Schematic table structure diagram of inode management table Schematic table structure diagram of volume management table Schematic table structure diagram of fragmented file information management table Configuration diagram of defragmentation metadata server machine Diagram showing the flow of processing by the fragment information collection program Diagram showing the flow of processing by the volume manager Diagram showing the flow of processing by the volume manager Figure showing an example of the volume setting screen Diagram showing the flow of processing by the free space monitoring program Diagram showing the flow of processing by the file relocation program

Explanation of symbols

10 storage machine 10a disk unit 10b storage controller 20 name management metadata server machine 20a storage unit 20b CPU
23 file system 23a file operation control function program 23b name management function program 30 area management metadata server machine 30a storage unit 30b CPU
33 file system 33a file operation control function program 33b area management function program 33c fragment information collection program 33d volume manager 33e free capacity monitoring program 40 defragmentation metadata server machine 40a storage unit 40b CPU
43 File system 43a File operation control function program 43b File relocation program 50 General-purpose server machine 60 Client machine 71 Inode management table 72 Volume management table 73 Fragmented file information management table 74 Volume setting screen

Claims (7)

Implemented a global file system that allows multiple node machines to share area information that defines the relationship between files and disk blocks, metadata that includes name information that defines the name space, and disks on which files are recorded A program for managing a disk by a metadata server machine in a cluster system,
The metadata server machine is
A determination unit that determines whether or not the file is in a fragmented state when an inquiry about metadata associated with file access is received from any node machine;
A registration unit that registers information about the file in a table in the storage device when the determination unit determines that the file is in a fragmented state;
A receiving means for receiving an instruction to add a volume together with designation of an arbitrary disk from an operator through an input device; and
When the accepting unit accepts an instruction to add a volume, it functions as a moving unit that sequentially moves a file whose information is registered in the table from an existing disk to a disk specified by the designation accepted by the accepting unit. A disk management program characterized by that. 2. The disk management program according to claim 1, wherein the moving means moves the file after the volume of the existing disk is in a predetermined state after the receiving means receives the volume addition instruction. The accepting means accepts the condition of the free capacity of the volume of the existing disk to which the volume addition is to be performed from the operator through the input device, together with the designation of an arbitrary disk and an instruction to add the volume.
The moving means moves the file after the receiving means receives the volume addition instruction, and after the volume of the existing disk is in a state meeting the free space condition accepted by the receiving means. The disk management program according to claim 2. 4. The disk management program according to claim 1, wherein the determination unit determines that the file is in a fragmented state when the number of extents of the file exceeds a predetermined threshold. 4. The disk management program according to claim 1, wherein the determination unit determines that the file is in a fragmented state when the number of accesses to the file exceeds a predetermined threshold. Implemented a global file system that allows multiple node machines to share area information that defines the relationship between files and disk blocks, metadata that includes name information that defines the name space, and disks on which files are recorded A method for managing a disk by a metadata server machine in a cluster system,
The metadata server machine is
A determination procedure for determining whether or not the file is in a fragmented state when an inquiry about metadata associated with file access is received from any node machine,
A registration procedure for registering information about the file in a table in the storage device when the determination procedure determines that the file is fragmented;
A procedure for accepting an instruction to add a volume together with designation of an arbitrary disk from an operator through an input device; and
When an instruction to add a volume is received in the reception procedure, a movement procedure for sequentially moving a file whose information is registered in the table from an existing disk to a disk specified by the designation received in the reception procedure is executed. A disk management method. Implemented a global file system that allows multiple node machines to share area information that defines the relationship between files and disk blocks, metadata that includes name information that defines the name space, and disks on which files are recorded A device for managing disks in a cluster system,
A determination unit that determines whether or not the file is in a fragmented state upon receiving a metadata inquiry associated with file access from any of the node machines,
A registration unit that registers information about the file in a table when the determination unit determines that the file is in a fragmented state;
A reception unit that receives an instruction to add a volume from an operator together with designation of an arbitrary disk; and
When the reception unit receives an instruction to add a volume, a moving unit that sequentially moves a file whose information is registered in the table from an existing disk to a disk specified by the specification received by the reception unit. A disk management device characterized by the above.