JP2009163529A - System and method for multiplexing data - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To rapidly recover redundancy without stopping writing from a host in the event of a failure of a storage in a data multiplexing system in which the host writes the same data to a plurality of storages. <P>SOLUTION: In the data multiplexing system, the host 1 performs redundant processing of data by writing the same data to storages 2a-2n. A redundancy recovery processing server 4 executes, when a failure occurs in any one of the storages, redundancy recovery processing for copying the same data as data reduced in redundancy from a copy source storage which is any one of the storages free from the failure to a copy destination storage which is a storage to be used for recovering the redundancy. A lock management server 3 performs exclusion in file and directory units so that change operation to files and directories stored in the plurality of storages is not simultaneously performed from the host and the redundancy recovery processing server 4 during the redundancy recovery processing. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a data multiplexing system, and more particularly, to a data multiplexing system in which a host computer on which a user application program runs writes data by writing the same data to a plurality of storages.

  There is a distributed storage system that includes a plurality of storages (disk device, disk array, file server, etc.) and is used as one file system. In such a system, when one file is redundantly stored in multiple storages for troubleshooting, etc., redundancy is achieved by writing the same data from host 1 to multiple storages 2 as shown in FIG. There is a configuration to do.

  For example, in Patent Document 1, a data duplexing device is provided between a host and a plurality of storages, and the data duplexing device multiplexes the data I / O instructions transmitted by the host and issues them to the respective storages. Multiplexing is performed. In Patent Document 1, a data duplexing device is provided in order to reduce the load on the host. However, if the increase in the host load does not become a problem, the host 1 itself can perform the processing performed by the data duplexing device.

  In a multiplexed system where data redundancy is achieved by directly or indirectly writing the same data to multiple storages in this way, when a storage failure occurs, the data is stored in the failed storage. Data redundancy is reduced and reliability is reduced. Therefore, it is necessary to create a new copy of data with reduced redundancy in order to restore reliability.

  An example of processing for recovering the reduced redundancy is described in paragraphs 0065 to 0070 of Patent Document 1. According to the description, if one of the storages (the disk subsystem in Patent Document 1) fails and a recovery target storage (copy destination storage) that is in the same state as at the time of the failure cannot be prepared, normal storage Redundancy is recovered by copying all data stored in (copy source storage) to the copy destination storage. In Patent Document 1, during the redundancy recovery process, access from the host to the copy destination storage is prohibited and access to the copy source storage is permitted.

JP-A-2005-196490

  However, if the host rewrites the data in the copy source storage while copying the data stored in the copy source storage to the copy destination storage, the rewritten part is an area that has not yet been copied. If there is, there is no problem because the data rewritten by the host is subsequently copied to the recovery target storage, but if the area has already been copied, it will not be reflected in the recovery target storage. Contradiction occurs in the storage contents of the storage. For this reason, although details are not described in Patent Document 1, it was recorded how the host rewritten the copy source storage, and the data in the copy source storage was completely copied to the copy destination storage. At this point, it is necessary to refer to the record and copy the portion rewritten by the host again. However, if the host rewrites the copy source storage again during this re-copying, another copy is required, and this state is repeated indefinitely. After all, unless the time when the host does not update the copy source storage appears for a certain period of time, it is difficult to make the contents of the copy source storage and the copy destination storage the same, and it takes a long time to restore redundancy. become.

  The object of the present invention is to quickly increase the redundancy without stopping the writing from the host when the redundancy is lowered in the data multiplexing system in which the host writes the same data to a plurality of storages to make the redundancy. It is to recover.

  In the first data multiplexing system of the present invention, a failure occurs in a plurality of storages, a host that performs data redundancy by writing the same data to the plurality of storages, and any one of the plurality of storages. When the redundancy is reduced due to the occurrence, any one of the plurality of storages having no failure is set as the copy source storage, and the storage used for restoring the redundancy is set as the copy destination storage, and the copy source storage A redundancy recovery processing server for executing redundancy recovery processing for copying the same data as data with reduced redundancy to the copy destination storage, and a change operation to files and directories stored in the plurality of storages. Simultaneously from the host and the redundancy recovery processing server during the redundancy recovery processing And a lock management server for performing exclusion in units of files and directories.

  In the first data multiplexing method of the present invention, a) a host performs data redundancy by writing the same data to a plurality of storages; and b) a redundancy recovery processing server includes the plurality of storages. If the redundancy decreases due to a failure in one of the storage systems, copy the storage that is not faulty among the multiple storages as the copy source storage and the storage that is used to restore the redundancy. Performing a redundancy recovery process for copying the same data as the data with reduced redundancy from the copy source storage to the copy destination storage; c) the host and the redundancy recovery processing server; The change operation to the files and directories stored in the plurality of storages is performed by the host and the And a step of acquiring a lock from the lock management server and performing exclusion on a file and directory basis so that the redundancy recovery processing server does not simultaneously perform the redundancy recovery processing.

  The first host computer of the present invention is a host computer that performs redundancy of data by writing the same data to a plurality of storages, and is redundant due to a failure in one of the plurality of storages. In the case where the degree of storage decreases, any of the plurality of storages having no failure is set as the copy source storage, and the storage used for restoring the redundancy is set as the copy destination storage, and from the copy source storage to the copy destination storage During the redundancy recovery processing in the redundancy recovery processing server for executing redundancy recovery processing for copying the same data as the data with reduced redundancy, the files and directories stored in the plurality of storages are changed. Exclude files and directories in units so that operations are not performed simultaneously Exists when the lock means for acquiring the lock from the lock management server to perform and when accessing the copy destination storage during the redundancy recovery process, it is found that there is no upper directory of the file or directory to be changed. Storage access means for copying the upper directory that has not been copied from the copy source storage to the copy destination storage, and performing a change operation of the file or directory to be changed.

  The second host computer of the present invention is a host computer that performs redundancy of data by writing the same data to a plurality of storages, and is redundant due to a failure in one of the plurality of storages. In the case where the degree of storage decreases, any of the plurality of storages having no failure is set as the copy source storage, and the storage used for restoring the redundancy is set as the copy destination storage, and from the copy source storage to the copy destination storage The redundancy recovery process that copies the same data as the data with reduced redundancy is executed, and the nonexistent upper directory specified in the incomplete processing notification and its subordinate directories and files are copied to the copy source storage. Redundancy recovery processing server that has a function to copy from one to the destination storage During the redundancy recovery process in, a lock is acquired from the lock management server for performing exclusion in units of files and directories so that the change operation to the files and directories stored in the plurality of storages is not performed simultaneously. Lock means, storage access means for discovering that there is no upper directory of the file or directory to be changed when accessing the copy destination storage during the redundancy recovery process, and the discovered upper directory that did not exist And incomplete processing notification means for transmitting the incomplete processing notification designating to the redundancy recovery processing server.

  The first redundancy recovery processing server device according to the present invention has a reduced redundancy due to a failure in one of a plurality of storages that are made redundant by writing the same data from the host. In this case, a redundancy recovery processing server apparatus that executes a process for recovering redundancy, wherein one of the plurality of storages without failure is a copy source storage, and the redundancy is recovered. The storage to be used is the copy destination storage, the copy means for executing redundancy recovery processing for copying the same data as the data with reduced redundancy from the copy source storage to the copy destination storage, and stored in the plurality of storages Change operation to the current file and directory is performed simultaneously from the host and the copy means And locking means for acquiring a lock from the lock management server in order to perform exclusion on a file and directory basis.

  According to the second redundancy recovery processing server device of the present invention, the redundancy is lowered due to a failure in one of a plurality of storages that are made redundant by writing the same data from the host. In this case, a redundancy recovery processing server apparatus that executes a process for recovering redundancy, wherein one of the plurality of storages without failure is a copy source storage, and the redundancy is recovered. The storage to be used is a copy destination storage, a copy means for executing redundancy recovery processing for copying the same data as the data with reduced redundancy from the copy source storage to the copy destination storage, and during the redundancy recovery processing The host that has accessed the copy destination storage uses the upper directory of the file or directory to be changed. Incomplete processing notification storage means for receiving and storing the incomplete processing notification when an incomplete processing notification designating an upper directory that did not exist is transmitted due to the fact that there is no queue. The copy means is characterized in that the non-existing upper directory designated by the incomplete processing notification and its subordinate directories and files are copied from the copy source storage to the copy destination storage.

  According to the present invention, in a data multiplexing system that performs redundant writing from the host to the storage group, when the redundancy decreases, the redundancy can be quickly recovered without stopping the writing from the host. The reason for this is that files and directories are not changed at the same time from the host and the redundancy recovery processing server during the redundancy recovery processing. This is because the copy from the copy source storage to the copy destination storage and the change operation to a plurality of storages including the copy source storage and the copy destination storage by the host are performed in parallel while being excluded in units.

  Next, embodiments of the present invention will be described in detail with reference to the drawings.

"First embodiment"
[Description of configuration]
Referring to FIG. 1, a data multiplexing system according to the first embodiment of the present invention includes a host computer (hereinafter simply referred to as a host) 1 and a plurality of storages 2a to 2n for storing data in a redundant manner. The lock management server 3 and the redundancy recovery processing server 4 are configured.

  The host 1 is a computer on which a user program (not shown) runs. Like a general computer, the host 1 includes hardware such as a CPU and a memory and software such as an operating system. As a related part, the storage access unit 5 is provided. A user program running on the host 1 accesses the files in the storages 2a to 2n via the storage access unit 5.

  The storage access unit 5 is means for accessing the storages 2a to 2n according to instructions from the user program, and is realized by, for example, a CPU and a program constituting the host 1. The storage access unit 5 has a state management unit 6 and a lock unit 7.

  The state management unit 6 manages which of the three states, “normal state”, “failure state”, and “recovery state”, of the storages 2a to 2n. “Normal state” means a state in which all the storages of the storages 2a to 2n are operating normally, that is, a state in which the redundancy of the specified data is n and multiplexed with the redundancy n. . The “failure state” refers to a state in which any of the storages 2a to 2n cannot be accessed and the redundancy is lowered. "Recovering state" means that in order to make it possible to access another storage instead of a storage that has become inaccessible due to a failure, data is copied from the normal storage to the other storage and the redundancy is restored. Say the state.

  A transition diagram of these three states is shown in FIG. In the normal state, when any storage fails, the state transits to the failure state. When the process to recover redundancy is started in the failure state, the state transitions to the recovering state, and when the redundancy recovery process is completed, the faulty storage disappears and the redundancy returns to the original state. If there is still a failed storage, it will transition to the failed state.

  In addition, the state management unit 6 records which storage is inaccessible in the failure state, and in the recovery state, which storage is the inaccessible storage and the recovery target storage (copy destination storage) is Record which storage it is. Further, the state management unit 6 manages the completion of processing started in each state and the closing of opened files and directories.

  The lock unit 7 communicates with the lock management server 3 and performs processing for acquiring and releasing locks in units of files and directories stored in the storage.

  The storage access unit 5 accesses the storages 2a to 2n according to the request from the user program running on the host 1, while acquiring and releasing the lock using the lock unit 7 as necessary, specifically, the file Processing such as creation, file deletion, directory creation, directory deletion, file writing, and file reading is performed. The operation of the storage access unit 5 differs depending on the type of state managed by the state management unit 6 (details will be described later).

  In FIG. 1, only one host 1 exists, but a plurality of hosts may exist.

  The storages 2a to 2n are storage devices such as NAS (Network Attached Storage) or OSD (Object Based Storage Device) for recording files or objects. Each file is distributed and arranged in the storages 2a to 2n. Further, redundancy is achieved by storing the same file in a plurality of storages in the storages 2a to 2n.

  In the present invention, a method of distributing each file in the plurality of storages 2a to 2n and a method of making each file redundant are arbitrary. For example, it may be allocated to storage in units of subordinate directories and files of a directory like an AFS (Andrew File System) volume, or may be performed in units of directories and files. Regarding storage methods for replicas, storage 2a, 2b, 2c, storage 2d, 2e, 2f, etc. are divided into several groups, each group is set, or each storage is divided into multiple areas. A method may be used in which a fixed set is created for each area, and files having the same contents are arranged in each set. In addition, allocation is performed to any combination of storage groups in units of files or directories, the allocation is managed by a meta server (not shown), and the location of the redundant file is obtained by the host 1 querying the meta server during path name resolution. You may make it do. Further, it is possible to assign a storage group of an arbitrary combination in units of files or directories, and to have each host 1 have a table in which the allocation status is recorded. In the following, for convenience of explanation, it is assumed that the same file is stored in the storages 2a to 2n and multiplexed with redundancy n.

  The lock management server 3 acquires locks in units of files and directories so that changes to the files and directories stored in the storages 2a to 2n are not performed simultaneously from the host 1 and the redundancy recovery processing server 4. It is a server that manages users. In the case of redundancy n, there are physically n files and directories, but they are not locked independently, but n files and directories are locked with one lock. In FIG. 1, there is only one lock management server 3, but a plurality of lock management servers 3 may be provided. In the configuration including a plurality, for example, files and directories in charge for each lock management server 3 are shared. By doing so, it is possible to avoid the lock management server 3 from becoming a processing bottleneck.

  When the redundancy is reduced due to a failure in one of the storages 2a to 2n, the redundancy recovery processing server 4 selects any storage without any failure from the storages 2a to 2n as a copy source storage, This is a computer that executes a redundancy recovery process for copying the same data as the data with reduced redundancy from the copy source storage to the copy destination storage, with the storage newly incorporated to recover the redundancy as the copy destination storage.

  The redundancy recovery processing server 4 includes hardware such as a CPU and a memory as well as a general computer as well as software such as an operating system. The copy unit 8 is provided as a part particularly related to the redundancy recovery processing. Have.

  The copy unit 8 is means for copying directories and files from any one of the storages 2a to 2n to another storage, and is realized by, for example, a CPU and a program constituting the redundancy recovery processing server 4. The redundancy recovery processing server 4 has a notification unit 9 and a lock unit 10.

  The notification unit 9 is means for notifying the host 1 of the start of the process of recovering the reduced data redundancy, and notifying the host 1 of the end when the recovery process ends.

  The lock unit 10 communicates with the lock management server 3 and performs processing for acquiring and releasing locks in units of files and directories stored in the storage.

  Although only one redundancy recovery server 4 exists in FIG. 1, there may be a plurality of servers. In the configuration having a plurality of copies, the replicas are created by sharing the targets to be duplicated among the redundancy recovery processing servers 4. In this way, the recovery process can be completed in a short time.

  In addition, the lock management server 3 and the redundancy recovery processing server 4 have been described as independent computers. However, the lock management server 3 and the redundancy recovery processing server 4 can be realized by the same computer, or can be realized as a part of the function of the host 1. .

[Description of operation]
Next, operations related to the storages 2a to 2n in the data multiplexing system according to the present embodiment will be described.

(1) Host 1 operation 1
The operation of the host 1 for the storages 2a to 2n is determined by the state managed by the state management unit 6. Here, the operation in the normal state and the operation in the failure state will be described, and the operation in the recovery state will be described after the operation of the redundancy recovery processing server 4 is described.

(1-1) Normal State The operation in the normal state on the host 1 will be described in the following manner: a file read operation, a file write operation, a file creation operation, etc., and an operation when a failure state is detected.

(1-1-1) Reading of a file The storage access unit 5 of the host 1 is one of a plurality of storages 2a to 2n storing files to be accessed by a user program (hereinafter referred to as target files). Read the corresponding file from the storage.

(1-1-2) File Writing The storage access unit 5 of the host 1 writes to the corresponding file in all of the plurality of storages 2a to 2n storing the target file. As a result, each data of the target file is made redundant with redundancy n.

(1-1-3) File creation, etc. The storage access unit 5 of the host 1 applies file creation, file deletion, directory creation, and directory deletion to all storages that store the target file or directory. Process. As a result, files and directories are made redundant with redundancy n.

(1-1-4) When a failure state is detected The failure of the storages 2a to 2n is detected by an arbitrary method. For example, when the storage access unit 5 of the host 1 issues a processing request to the storages 2a to 2n, the storage that does not return a response is detected as a storage in which a failure has occurred. When there are a plurality of hosts 1, the host 1 that has detected a failure in the storages 2a to 2n may notify the other hosts 1 of the failure detection. A storage monitoring server (not shown) may monitor the status of each storage 2a to 2n and notify the host 1 of the status.

  When a failure is detected in the storages 2a to 2n, the status management unit 6 of the host 1 is notified of the storage in which the failure has occurred. The state management unit 6 transitions from the normal state to the failure state, and records the notified storage as an inaccessible storage. Thereafter, the inaccessible storage is removed from the processing request target in the storage access unit 5.

(1-2) Failure State The operation in the failure state on the host 1 will be described in the following manner: file read operation, file write operation, file creation operation, etc., and operation when a failure state is detected.

(1-2-1) Reading of files The storage access unit 5 of the host 1 excludes storages recorded as inaccessible storages in the state management unit 6 among the plurality of storages 2a to 2n storing the target file. The corresponding file is read from any one of the remaining storage groups.

(1-2-2) File writing The storage access unit 5 of the host 1 excludes the storages recorded as inaccessible storages in the state management unit 6 among the plurality of storages 2a to 2n storing the target file. Write to the corresponding file for all remaining storage.

(1-2-3) File creation, etc. The storage access unit 5 of the host 1 is in the state of a plurality of storages 2a to 2n storing the target file or directory for file creation, file deletion, directory creation, and directory deletion. The corresponding process is performed on all the storages in the remaining storage group excluding the storage recorded as inaccessible storage in the management unit 6.

(1-2-4) Specific method to exclude from processing target The processing in the failure state described above is the path name resolution performed prior to file writing / reading, file creation / deletion, directory creation / deletion. This is achieved by removing inaccessible storage from the scope during processing. Here, the path name resolution process is a process for acquiring information (file handle) for accessing a file or directory having a specified path name. Details of the path name resolution processing will be described later as processing in the recovery state.

(2) Operation of the redundancy recovery processing server 4 (2-1) Overall flow A storage is prepared as a substitute for the failed storage, and the prepared storage receives requests from the host 1 and the redundancy recovery processing server 4. After reception is possible, the redundancy recovery processing server 4 starts processing in response to an instruction from the system administrator. Here, there are two possible storage contents of the alternative storage: the case where the storage contents are the same as when the failure occurred, and the case where it is not. The procedure is the same. Hereinafter, the overall operation flow of the redundancy recovery processing server 4 will be described with reference to FIG.

  The copy unit 8 of the redundancy recovery processing server 4 adds information on the recovery target storage (copy destination storage) to the all hosts 1 through the notification unit 9 and notifies the recovery processing in progress (start of recovery processing). (S100), waiting for responses from all the hosts 1 (S101). The processing of the host 1 that has received the notification during the recovery processing will be described later, but the host 1 returns a response when a predetermined condition is met. When the redundancy recovery processing server 4 receives a response from all the hosts 1, it performs recovery processing (S102). When the recovery process is completed, the notification unit 9 adds information on the storage for which the recovery process has been completed to all the hosts 1 to notify the completion of the recovery process (S103), and ends the process of FIG.

(2-2) Recovery Processing Next, details of the recovery processing in step S102 of FIG. 4 will be described with reference to FIG.

  First, the copy unit 8 selects one unprocessed directory to be stored (copied) from a normal storage (copy source storage) to a recovery target storage (copy destination storage) and sets it as a processing target (S200). In selecting an unprocessed directory in step S200, selection is made in order from the top of the tree structure of the directory (close to the root). Next, the copy unit 8 performs directory recovery processing for the directory to be processed (S201). When the recovery process for one unprocessed directory is completed, it is determined whether or not there is an unprocessed directory to be stored in the storage to be recovered (S202), and if it remains, the process returns to step S200 and described above. If the same process is repeated and there is no remaining, the process of FIG. 5 is terminated (S202).

(2-3) Directory Recovery Processing Next, the directory recovery processing in step S201 in FIG. 5 will be described with reference to FIG.

  The directory recovery process is started by specifying the directory selected in step S200 of FIG. 5 as the target directory.

  First, the copy unit 8 uses the lock unit 10 to specify a target directory, request a lock from the lock management server 3, and wait for a lock acquisition response (S300).

  Next, the copy unit 8 determines the contents of the lower directory of the non-recoverable storage (copy source storage) corresponding to the specified directory of the recovery target storage (copy destination storage) and the specified recovery target storage. The contents of the lower directory of the directory are compared (S301).

  As a result of the comparison, the copy unit 8 creates a lower directory in the target directory of the copy destination storage for the lower directory that exists in the copy source storage and does not exist in the copy destination storage. As a result of the comparison, the copy unit 8 deletes the lower directory that does not exist in the copy source storage but exists in the copy destination storage from the target directory of the copy destination storage (S302).

  Next, the copy unit 8 uses the lock unit 10 to request the lock management server 3 to release the lock acquired in step S300 (S303).

  Next, if there is an unprocessed directory stored in the recovery target storage in a directory below the target directory, the copy unit 8 proceeds to step S305, and if not, proceeds to step S307 (S304).

  In step S305, the copy unit 8 selects one unprocessed directory to be stored in the recovery target storage as a processing target among the directories below the target directory. Next, the copy unit 8 designates the directory selected as the processing target, calls the directory recovery processing of FIG. 6 again (S306), and when finished, proceeds to step S304.

  In step S307, the copy unit 8 designates the target directory and performs file recovery processing. When the file recovery process in step S307 ends, the directory recovery process in FIG. 6 ends, and the process returns to the caller.

  As described above, exclusive creation and deletion of directories can eliminate the host 1 from creating and deleting directories in the target directory during the processing of the redundancy recovery processing server 4 and maintain consistency. it can.

(2-4) File Recovery Processing Next, the file recovery processing in step S307 in FIG. 6 will be described with reference to FIG.

  The file recovery process is started by specifying a directory.

  First, the copy unit 8 uses the lock unit 10 to specify a target directory, request a lock from the lock management server 3, and wait for a lock acquisition response (S400).

  Next, the copy unit 8 determines the contents of the files under the directory of the storage (copy source storage) that is not the recovery target corresponding to the specified directory of the recovery target storage (copy destination storage) and the specified recovery target storage. The contents of the files under the directory are compared (S401). As a result of this comparison, files that exist in the copy source storage but do not exist in the copy destination storage, exist in the copy destination storage, but do not exist in the copy source storage, exist in both storages, but the contents of the files are different. Files are detected.

  As a result of the comparison, if there is no unmatched file, the process proceeds to step S403. If there is at least one unmatched file, the process proceeds to step S404 (S402).

  In step S403, the copy unit 8 uses the lock unit 10 to request the lock management server 3 to release the lock acquired in step S400, and ends the process of FIG.

  In step S404, the copy unit 8 selects one of the files whose contents found in step S401 do not match.

  The copy unit 8 proceeds to step S406 if the file selected in step S404 is not a copy source file, and proceeds to step S407 if not (S405).

  In step S406, the copy unit 8 deletes the corresponding file from the copy destination storage, and proceeds to step S402.

  In step S407, if the selected file is not at the copy destination, the process proceeds to step S408, and if not, the process proceeds to step S409.

  In step S408, the copy unit 8 creates the file selected in step S404 in the copy destination storage.

  In step S409, the copy unit 8 uses the lock unit 10 to designate the file selected in step S404, requests lock to the lock management server 3, and waits for a lock acquisition response.

  Next, the lock unit 8 requests the lock management server 3 to release the lock for the directory acquired in step S400 (S410), and then copies the file selected in step S404 from the copy source storage to the copy destination storage (step S404). S411). Then, using the lock unit 10, the copy unit 8 requests the lock management server 3 to release the lock for the file acquired in step S409, and proceeds to step S400 (S412).

  In this way, by exclusively performing file creation and deletion, it is possible to eliminate the host 1 creating and deleting files in the target directory during the processing of the redundancy recovery processing server 4, and maintain consistency. .

  Also, by exclusively performing file copying, it is possible to eliminate the host 1 changing the target file while the redundancy recovery processing server 4 is processing, and maintain consistency.

  In this embodiment, since the target directory is unlocked before copying the file (S410 in FIG. 7), the host 1 accesses the target directory during file copying and deletes the information of the corresponding file. Then, the file copied to the copy destination storage remains as a file that is not linked from the directory. However, since such files are deleted when their reference count reaches zero, consistency is maintained.

  FIG. 16 shows how directories and files are copied from the copy source storage to the copy destination storage by the directory recovery process and the file recovery process described above. In the figure, (1) to (31) all indicate the state of the copy destination storage, and the initial state is only the root directory as indicated by (1). Further, the copy destination storage is assumed to be in the same state as shown in (31). In the figure, □ indicates a directory, shaded □ indicates a locked directory, and ○ indicates a file.

  As shown in (2) to (4) of FIG. 16, two directories immediately under the root of the copy source storage are created (copied) with the root directory locked at first. Next, one of these two directories is targeted for processing, and one directory directly under the target directory in the copy source storage is created (copied) with the target directory locked ((5) ~ (7)). Next, when the generated directory is the target of processing and the target directory is locked, an attempt is made to copy a directory directly under the target directory of the copy source storage. Lock and finish the process for the target directory ((8) to (10)). Then, since the terminal directory has been reached, three files immediately under the target directory of the copy source storage are created (copied) with the terminal directory locked as a processing target ((11) to (13) ). Then, return the target directory to the previous level and repeat the same operation.

(3) Host 1 operation 2
(3-1) Operation when a notification is received during recovery processing The storage access unit 5 of the host 1 performs recovery processing with information on the recovery processing target storage (copy destination storage) added from the redundancy recovery processing server 4. When the notification is received, execution of the process shown in FIG. 8 is started.

  First, the storage access unit 5 uses the state management unit 6 to change the state to the recovery process in-progress state, and stores the notified information on the recovery process target storage in a memory (not shown) (S500). Next, the storage access unit 5 uses the state management unit 6 to start processing (file writing / reading from / to the storages 2a to 2n, directory creation / deletion, file creation / Wait for completion of (delete) (S501). Further, the storage access unit 5 waits for the state management unit 6 to close the opened file and directory before receiving the notification during the recovery process (S502). When all the processes started before receiving the recovery process start notification are completed and all open files and directories are closed, the response to the notification during the recovery process is restored to the redundancy level. The data is transmitted to the processing server 4 (S503), and the processing in FIG.

(3-2) Operation when receiving notification of completion of recovery processing When the storage access unit 5 of the host 1 receives a recovery processing completion notification from the redundancy recovery processing server 4 with information on the storage for which recovery processing has been completed added Then, the execution of the process shown in FIG. 9 is started.

  First, the storage access unit 5 uses the state management unit 6 to delete storage for which recovery processing has been completed from the recovery target storage (S600). Next, determine whether there is still recovery target storage by referring to the memory records managed by the status management unit 6, and if there is at least one recovery target storage, terminate the processing in FIG. If none remain, the process proceeds to step S602 (S601). In step S602, the storage access unit 5 uses the state management unit 6 to determine whether or not inaccessible storage remains by referring to the memory record. If at least one storage remains, the storage access unit 5 proceeds to step S603, and 1 If none remain, the process moves to step S604. In step S603, the state management unit 6 changes the state to a failure state, and the processing in FIG. In step S604, the state management unit 6 changes the state to the normal state, and the process in FIG. 9 ends.

(3-3) Operation in recovery state (3-3-1) Path name resolution The path name resolution operation in host 1 in the recovery state will be described with reference to FIG.

  Path name resolution is a process of acquiring information (file handle) for accessing a file or directory having a specified path name.

  First, the storage access unit 5 sets the first directory as a search range (S700). Next, the storage access unit 5 acquires information and a file handle of the storage storing the head directory (S701). Next, if the range set in step S700 is the end of path name resolution, the storage access unit 5 proceeds to step S703, otherwise proceeds to step S704 (S702). In step S703, the storage access unit 5 uses the storage and file handle information acquired in step S701 as the return value of the path name resolution process, and the process ends.

  On the other hand, in step S704, the storage access unit 5 extracts the directory name or file name under the directory that is the search range from the path name, and sets it as a search target.

  Next, the storage access unit 5 selects one unrecovered (unprocessed) storage that is not being recovered or accessed from the storage group within the search range (S705). Next, if there is a target storage in step S705, the process proceeds to step S707, and if there is no target storage, the process proceeds to step S711 (S706).

  In step S707, the storage access unit 5 obtains a file handle (lookup) of the search target file or directory from the directory in the storage search range selected in step S705. If the lookup is successful in step S707, the process proceeds to step S710, and if it fails, the process proceeds to step S709 (S708). In step S709, the storage access unit 5 sets the return value of the path name resolution process as an error, and ends the process of FIG. In step S710, the storage access unit 5 stores the file handle acquired in step S707 in a memory (not shown) as a file handle of a search target file or directory in the storage selected in step S705. Then, control goes to a step S705.

  On the other hand, in step S711, the storage access unit 5 selects a storage that is being recovered and has not been searched (unprocessed) among the storage groups in the search range. Next, if there is a target storage in step S711, the process proceeds to step S716, and if not, the process proceeds to step S713.

  In step S713, the storage access unit 5 determines whether or not the end of the path name resolution has been resolved. If it has been resolved, the process proceeds to step S714, and if not resolved, the process proceeds to step S715. To do. In step S714, the stored file handle group is returned, and the process of FIG. 10 is terminated. In step S715, the stored file handle group is set to the next search range, the process proceeds to step S704, and the above-described processing is repeated.

  In step S716, the storage access unit 5 acquires a file handle (lookup) of the file to be searched or the directory from the directory in the search range of the storage selected in step S711 (the storage being recovered). If the lookup is successful in step S716, the process proceeds to step S718. If the lookup is unsuccessful, the process proceeds to step S719 (S717).

  In step S718, the storage access unit 5 stores the file handle acquired in step S716 in a memory (not shown) as a file handle of a search target file or directory in the storage selected in step S711. Then, the process proceeds to step S711.

  The situation where the process proceeds to step S719 occurs when the host 1 tries to access a directory or file that has not yet been copied by the redundancy recovery processing server 4 for the storage being recovered. In this embodiment, at this point, the host 1 itself performs a copy process from the copy source storage to the copy destination storage in the range necessary for the path name resolution this time, instead of the redundancy recovery processing server 4.

  Specifically, first, the storage access unit 5 uses the lock unit 7 to designate a directory in the search range, requests a lock from the lock management server 3, and waits until the lock is acquired (S719). Next, the storage access unit 5 creates (copies) the file or directory targeted in step S704 under the directory in the storage search range selected in step S711 (S720). Next, the file or directory created in step S720 is stored as the file handler of the storage selected in step S711 (S721). Finally, the storage access unit 5 uses the lock unit 7 to request the lock management server 3 to release the lock acquired in step S719 (S722). Then, the process proceeds to step S711.

  FIG. 17 shows an example in which the target directory is created for the recovery target storage with redundancy when the host 1 resolves the path name. In this example, the host 1 creates a new file for the recovery target storage in the state (13) of FIG.

  The host 1 finishes the lookup process up to the right directory of the two directories directly under the root, extracts the directory name below it from the path name (S704 in FIG. 10), and performs the lookup process in step S716. When discovering, there is no target directory. Therefore, the host 1 locks the directory in the search range as shown in (14) of FIG. 17 (S719), and then creates a directory by copying from another normal storage as shown in (15) ( The created directory is stored as a file handler of the recovery target storage (S721), and the search range directory lock is released as shown in (16) (S722). Thereafter, the host 1 creates a new file using the file handler obtained by the path name resolution ((17) to (19) in FIG. 17).

  As described above, in the present embodiment, a directory can be created when a directory that does not exist in the recovery target storage is found during path name resolution. In addition, since the lock is performed at the time of creation, the process of the redundancy recovery processing server 4 is exclusive and consistency is not lost.

  A supplementary explanation will be given regarding the processing of FIG. In step S700, the search is started from the head directory called root. In addition, it is assumed that the root storage storage group and the corresponding file handle are acquired in advance at the timing of mount or the like.

  Also, when specifying a relative path, resolution may be performed from a directory other than root. In this case, it is assumed that the storage storage group of the directory as a viewpoint and the corresponding file handle are acquired in advance at the timing of the working directory resolution or the like even in the case of relative path designation.

  Information stored in the host 1 is used to obtain the file handle. In some cases, the storage 2a to 2n may be inquired. Further, there is a case in which a server that manages storage of files in a storage called a meta server is inquired.

(3-3-2) File creation, etc. In the recovery state, when the storage access unit 5 of the host 1 creates a directory, deletes a directory, creates a file, or creates a directory, lock it before actually creating or deleting it. Using the unit 7, an upper directory in which a file or directory is created or deleted is specified, a lock is requested to the lock management server 3, and it waits until the lock is acquired. Then, after actually creating and deleting, the lock management server 3 is used to request the lock management server 3 to release the acquired lock. Since the file or directory is locked when the file or directory is created or deleted in this way, it is exclusive with the processing of the redundancy recovery processing server 4, and consistency is not lost.

(3-3-3) File Open The storage access unit 5 of the host 1 manages locks every time a file is opened. The file open operation on the host 1 in the recovery state will be described with reference to FIG.

  First, the storage access unit 5 shifts to step S803 when open is requested in the read only mode, and shifts to step S801 when different. In step S801, the lock unit 7 designates an open target, requests lock to the lock management server 3, and waits until the lock is acquired. Next, information necessary for releasing the lock later is stored in the lock management area for each open, and the process proceeds to step S803 (S802).

  In step S803, the storage access unit 5 performs the open process on the storage storing the target file other than the inaccessible storage, and ends the process of FIG.

(3-3-4) File close The file close operation on the host 1 during recovery is described with reference to FIG.

  First, the storage access unit 5 performs a close process on the opened storage (S900). Next, referring to the lock management area for each open, if it has been recorded that it has been locked, the process proceeds to step S902, and if it has not been recorded, the process of FIG. 12 ends (S901). In step S902, the lock unit 7 is used to request the lock management server 3 to release the lock based on the recorded information, and the processing of FIG.

  In this way, by performing lock by opening and releasing the lock by closing, the processing of the redundancy recovery processing server 4 is exclusive and consistency is not lost. Further, the lock is not performed at the time of read only, and the processing of the host 1 is not hindered when the write is not performed.

(3-3-5) File writing When the storage access unit 5 of the host 1 writes to a file in the recovery state, the state management unit 6 among the plurality of storages storing the target file is designated as an inaccessible storage. The file is written to all the remaining storage groups except the recorded storage.

(3-3-6) File Reading When the storage access unit 5 of the host 1 performs reading from the file in the recovery state, the state management unit 6 among the plurality of storages storing the target file and the inaccessible storage and The corresponding file is read from any one of the remaining storage groups excluding the storage recorded as the recovery target storage.

(4) Operation of lock management server 3 The lock management server 3 manages the acquisition status of locks in units of files or directories, and locks are only allowed to be acquired for one request at a time. For example, the lock management server 3 stores the lock request in the server, and requests for the same file or directory determine the next request for acquiring the lock in the order of arrival or the like. Further, the lock management server 3 notifies the lock acquisition request that it has been acquired when it is permitted, and releases the lock in response to the lock release request.

  In this embodiment, the storage access unit 5 of the host 1 issued a lock request when the file was opened. However, the storage access unit 5 performs a lock request before writing to the target file for the first time and releases the lock when the file is closed. good. In this way, if the file is not actually changed (not written), the lock is not performed, so that the possibility of hindering the processing in the host 1 can be further reduced.

“Second Embodiment”
Referring to FIG. 2, the data multiplexing system according to the second embodiment of the present invention is compared with the data multiplexing system according to the first embodiment shown in FIG. The difference is that a storage access unit 5A is provided instead of the unit 5, and the redundancy recovery processing server 4 is provided with a copy unit 8A instead of the copy unit 8.

  The storage access unit 5A of the host 1 is different from the storage access unit 5 in that it newly includes an incomplete processing notification unit 11, and the operation in the recovery state is partially different. .

  The copy unit 8A of the redundancy recovery processing server 4 is different from the copy unit 8 in that it newly includes an incomplete process notification storage unit 12, and a part of the recovery process is different. .

  Hereinafter, the operation of the present embodiment will be described with a focus on differences from the first embodiment.

(3 ') Host 1 action 2
(3-3 ') Host operation in recovery state (3-3-1') Path name resolution The path name resolution operation in host 1 in the recovery state will be described with reference to FIG. As described above, path name resolution is a process of acquiring information (file handle) for accessing a file or directory having a specified path name. In the case of the present embodiment, the storage access unit 5A has an incomplete process flag as an internal state of the process.

  First, the storage access unit 5A sets the head directory as a search range and clears the incomplete processing flag (S1100). Next, obtain the storage information and file handle that stores the first directory (S1101), and if the range set in step S1100 is the end of the path name resolution, move to step S1103. The process proceeds to step S1104 (S1102). In step S1103, the storage and file handle information acquired in step S1101 and the incomplete process flag are returned, and the process ends.

  Of the processes after step S1104, steps S1104 to S1113 and S1115 to S1118 are the same as steps S704 to S713 and S715 to S718 in FIG. In steps S1119 and S1114, the following processing is performed.

  In step S1119, since the lookup in step S1116 failed, the incomplete process flag is set, the directory not found this time is recorded in the memory in association with the incomplete process flag, and the process proceeds to step S1111. In the first embodiment, when a directory that does not exist in the recovery target storage is found during path name resolution, the host 1 creates (copies) the directory. However, in this embodiment, when a directory that does not exist in the recovery target storage is found during path name resolution, it is assumed that the directory is out of the search range, and the lower directory or file is changed to the recovery target storage. Do not do. Instead, in order to cause the redundancy recovery processing server 4 to copy the undiscovered directory later, the incomplete processing flag is set, and the directory not found is recorded in the memory. As a result, the directory creation process and the lock associated therewith are not performed, so that the processing time in the host 1 is shortened as compared with the first embodiment.

  In step S1114, the stored file handle group, the incomplete process flag and the information of the undiscovered directory stored corresponding thereto are used as return values of the path name resolution process, and the process of FIG. 14 is ended.

(3-3-2 ') File creation, etc. In the recovery state, when the storage access unit 5 of the host 1 creates a directory, deletes a directory, creates a file, and deletes a file, before actually creating and deleting it, The lock unit 7 is used to designate an upper directory in which a file or directory is created or deleted, request the lock to the lock management server 3, and wait until the lock is acquired.

  Next, after creating a directory, deleting a directory, creating a file, and deleting a file only for the storage from which the file handle can be acquired, the lock unit 7 is used to release the acquired lock to the lock management server 3. Request. Since the file or directory is locked when the file or directory is created or deleted in this way, it is exclusive with the processing of the redundancy recovery processing server 4, and consistency is not lost.

  After that, if the incomplete processing flag returned by the path name resolution is set, the incomplete processing notifying unit 11 is used, and the higher rank of the undiscovered directories recorded corresponding to the incomplete processing flag A directory is specified, and incomplete processing is notified to the redundancy recovery processing server 4. During path name resolution, no directory is created in the recovery target storage and no change is made to the recovery target storage, but incomplete processing is notified to the redundancy recovery processing server 4 and copy processing is executed again as described later. Thus, the synchronization of the directory state is guaranteed.

(3-3-3 ') File open The file open process of the host 1 in the recovery status in this embodiment is performed only for the storage for which the file handle was obtained. Processing of the host 1 file in the recovery state of the first embodiment, except that the incomplete processing flag returned by path name resolution and the undiscovered directory accompanying it are received and stored in Is the same. Here, the reason why the incomplete processing flag and the undiscovered directory associated therewith are stored is to be used when the file is closed later.

(3-3-4 ′) File Close Next, the file close operation in the host 1 in the recovery state will be described with reference to FIG.

  First, close processing is performed on the opened storage (S1200).

  Next, if it is recorded that the lock has been locked in the lock management area for each open, the process proceeds to step S1202, and if not recorded, the process of FIG. 15 is terminated (S1201).

  In step S1202, the lock management server 3 is requested to release the lock based on the recorded information. By doing so, since locking is performed when a file or directory is created or deleted, it is exclusive from the processing of the redundancy recovery processing server 4, and consistency is not lost.

  Next, the incomplete processing flag stored at the time of opening is referred to. If it is set and passed, the process proceeds to step S1204. If not, the process of FIG. 15 is terminated.

  In step S1204, the incomplete processing notification storage unit 12 is used to specify an upper directory of the undiscovered directories recorded accompanying the incomplete processing flag to the redundancy recovery processing server 4 to perform incomplete processing. And exit. In this way, no directory is created in the recovery target storage and no change is made to the recovery target storage during path name resolution, but incomplete processing is notified to the redundancy recovery processing server 4 and copy processing is executed again. Thus, the synchronization of the directory state is guaranteed.

(3-3-5 ') File writing When the storage access unit 5 of the host 1 writes to a file in the recovery state, the state management unit 6 opens the file among a plurality of storages storing the target file. The corresponding file is written to all the storage groups. Therefore, writing is not performed for the recovery target storage for which path name resolution could not be performed.

(3-3-6 ′) File Reading When the storage access unit 5 of the host 1 reads from a file in the recovery state, the state management unit 6 among the storages storing the target file cannot access the storage. The corresponding file is read from any one of the remaining storage groups excluding the storage recorded as the recovery target storage.

(2 ′) Redundancy Recovery Processing Server 4 Operation (2-1 ′) Overall Flow A storage is prepared as a substitute for the failed storage, and the prepared storage is the host 1 and the redundancy recovery processing server 4 After the request can be received, the processing of the redundancy recovery processing server 4 is started. Here, there are two possible storage contents of the alternative storage: the case where the storage contents are the same as when the failure occurred, and the case where it is not. The procedure is the same. Hereinafter, the overall operation flow of the redundancy recovery processing server 4 will be described with reference to FIG.

  The copy unit 8A of the redundancy recovery processing server 4 adds information on the recovery target storage (copy destination storage) to the all hosts 1 through the notification unit 9, and notifies the recovery processing in progress (S1000). Wait for a response from 1 (S1001). The host 1 that has received the notification during the recovery process returns a response when a predetermined condition is met, as in the first embodiment. When the redundancy recovery processing server 4 receives a response from all the hosts 1, it performs recovery processing (S1002). The operation up to this point is the same as in the first embodiment.

  When the copy unit 8A completes the recovery process of step S1002, the copy unit 8A refers to the incomplete process notification storage unit 12 to determine whether or not the incomplete process notification has come from the host 1, and if so, step S1005 If not, the process proceeds to step S1004 (S1003). In step S1004, the notification unit 9 adds the information of the storage for which the recovery process has been completed to all the hosts 1, notifies the completion of the recovery process, and ends the process of FIG.

  In step S1005, the copy unit 8A selects one unprocessed directory from the directories stored in the incomplete process notification storage unit 12. If there is no unprocessed directory, the process proceeds to step S1003, and if present, the process proceeds to step S1007 (S1006).

  In step S1007, the copy unit 8A designates the directory selected in step S1005, performs directory recovery processing, and proceeds to step S1005. Details of the directory recovery processing performed in step S1007 are the same as those already described with reference to FIG.

  By re-executing the directory recovery process that specified the directory notified by the incomplete processing notification from the host 1 in this way, the files and directories that the host 1 was able to create or delete from the normal storage are stored in the host 1 Can be reflected in the storage during recovery processing that could not be created or deleted, and consistency is not lost.

  Note that a process of waiting for a time defined by the system may be inserted between step S1002 and step S1003.

Block diagram of the first embodiment of the present invention Block diagram of the second embodiment of the present invention State transition diagram showing host state transitions Flow chart showing operation of redundancy recovery processing server Flow diagram showing the operation of recovery processing in the redundancy recovery processing server Flow chart showing the operation of directory recovery processing on the redundancy recovery processing server Flow diagram showing file recovery processing operation on the redundancy recovery processing server Flow chart showing host operation when receiving notification of recovery processing start Flow chart showing host operation when receiving notification of recovery processing completion Flow diagram showing the path name resolution operation during recovery on the host Flow diagram showing the path name resolution operation during recovery on the host Flow diagram showing file open operation during recovery on the host Flow chart showing file close operation during recovery on the host Flow chart showing another operation of recovery processing in the redundancy recovery processing server Flow diagram showing another operation of path name resolution in the recovery state on the host Flow diagram showing another operation of path name resolution in the recovery state on the host Flow chart showing another operation of file close during recovery on the host Diagram showing how directories and files are copied from copy source storage to copy destination storage by directory recovery processing and file recovery processing Diagram showing how directories and files are copied from copy source storage to copy destination storage by directory recovery processing and file recovery processing Diagram showing an example of creating a target directory for the recovery target storage with redundancy when resolving the path name on the host Block diagram of a system where the host multiplexes data by writing the same data to multiple storages

Explanation of symbols

1 ... Host
2a ~ 2n ... Storage
3 ... Lock management server
4 ... Redundancy recovery processing server
5, 5A ... Storage access section
6 State management department
7, 10 ... Lock part
8, 8A ... Copy section
9 ... Notification part
11 ... Incomplete processing notification section
12… Incomplete processing notification storage

Claims (24)

Multiple storage,
A host for performing data redundancy by writing the same data to the plurality of storages;
If the redundancy decreases due to a failure in one of the multiple storages, use one of the multiple storages that has no failure as the copy source storage and restore the redundancy. A redundancy recovery processing server that performs redundancy recovery processing for copying the same data as the data with reduced redundancy from the copy source storage to the copy destination storage,
Exclusion is performed in units of files and directories so that operations for changing to files and directories stored in the plurality of storages are not performed simultaneously from the host and the redundancy recovery processing server during the redundancy recovery processing. And a lock management server for the data multiplexing system.   When the host finds that the upper directory of the file or directory to be changed does not exist when accessing the copy destination storage during the redundancy recovery process, the host copies the upper directory that did not exist from the copy source storage. 2. The data multiplexing system according to claim 1, wherein the data multiplexing system is copied to a destination storage and an operation for changing a file or directory to be changed is performed.   In the path name resolution process, the host performs the discovery that the upper directory of the file or directory to be changed does not exist, and copies the non-existing upper directory from the copy source storage to the copy destination storage. The data multiplexing system according to claim 2. When the host finds that the upper directory of the file or directory to be changed does not exist when accessing the copy destination storage during the redundancy recovery processing, an incomplete processing notification specifying the upper directory that did not exist To the redundancy recovery processing server,
2. The redundancy recovery processing server copies a non-existing upper directory designated by an incomplete processing notification and its subordinate directories and files from a copy source storage to a copy destination storage. Data multiplexing system.   5. The data multiplexing system according to claim 4, wherein the host performs discovery in a path name resolution process that a file or directory to be changed does not exist.   The data multiplexing system according to any one of claims 1 to 5, wherein the redundancy recovery processing server notifies the host of the start and end of the redundancy recovery processing. a) the host performing data redundancy by writing the same data to a plurality of storages;
b) When the redundancy is reduced due to a failure occurring in any one of the plurality of storages, the redundancy recovery processing server designates one of the plurality of storages having no failure as a copy source storage. Performing a redundancy recovery process of copying the same data as the data with reduced redundancy from the copy source storage to the copy destination storage, with the storage used to recover the redundancy as the copy destination storage;
c) The host and the redundancy recovery processing server simultaneously perform a change operation to files and directories stored in the plurality of storages during the redundancy recovery processing between the host and the redundancy recovery processing server. A data multiplexing method comprising: acquiring a lock from a lock management server and performing exclusion on a file and directory basis.   d) When the host finds that there is no upper directory of the file or directory to be changed when accessing the copy destination storage during the redundancy recovery process, the host that does not exist is copied to the copy source storage 8. The data multiplexing method according to claim 7, further comprising a step of performing a change operation of a file or directory to be changed by copying from the copy destination storage to the copy destination storage.   In the path name resolution process, the host performs the discovery that the upper directory of the file or directory to be changed does not exist, and copies the non-existing upper directory from the copy source storage to the copy destination storage. The data multiplexing method according to claim 8. d) When the host finds that there is no upper directory of the file or directory to be changed when accessing the copy destination storage during the redundancy recovery process, the host specified the upper directory that did not exist Sending a processing notification to the redundancy recovery processing server;
e) The redundancy recovery processing server includes a step of copying the non-existing upper directory specified by the incomplete processing notification and the subordinate directories and files from the copy source storage to the copy destination storage. The data multiplexing method according to claim 7.   11. The data multiplexing method according to claim 10, wherein the host performs discovery in a path name resolution process that a file or directory to be changed does not exist.   12. The data multiplexing method according to claim 7, wherein the redundancy recovery processing server notifies the host of the start and end of the redundancy recovery processing. A host computer that makes data redundant by writing the same data to multiple storages,
If the redundancy decreases due to a failure in one of the multiple storages, use one of the multiple storages that has no failure as the copy source storage and restore the redundancy. During the redundancy recovery processing in the redundancy recovery processing server for executing redundancy recovery processing for copying the same data as the data with reduced redundancy from the copy source storage to the copy destination storage. A lock means for acquiring a lock from a lock management server for performing exclusion in units of files and directories so that change operations to the files and directories stored in the plurality of storages are not performed at the same time;
When accessing the copy destination storage during the redundancy recovery process, if it is found that there is no upper directory of the file or directory to be changed, the upper directory that did not exist is copied from the copy source storage to the copy destination storage. And a storage access means for changing the file or directory to be changed.   In the path name resolution process, the storage access means discovers that the upper directory of the file or directory to be changed does not exist and copies the non-existing upper directory from the copy source storage to the copy destination storage. The host computer according to claim 13. A host computer that makes data redundant by writing the same data to multiple storages,
If the redundancy decreases due to a failure in one of the multiple storages, use one of the multiple storages that has no failure as the copy source storage and restore the redundancy. The storage to be copied is the copy destination storage, the redundancy recovery process is executed to copy the same data as the data with reduced redundancy from the copy source storage to the copy destination storage, and the presence specified in the incomplete processing notification During the redundancy recovery processing in the redundancy recovery processing server having a function of copying the upper directory that has not been performed and the directories and files under the upper directory to the copy destination storage, the upper directory is stored in the plurality of storages. Changes to files and directories cannot be performed at the same time. And locking means for acquiring a lock from the lock management server for performing an exclusive with files and directories unit as,
A storage access means for discovering that there is no upper directory of the file or directory to be changed when accessing the copy destination storage during the redundancy recovery process;
A host computer comprising: an incomplete processing notification means for transmitting an incomplete processing notification specifying a discovered higher-order directory to the redundancy recovery processing server.   The host computer according to claim 15, wherein the storage access unit executes discovery in a path name resolution process that a file or directory to be changed does not exist. When the same data is written from the host, the data is made redundant. When a failure occurs in one of the multiple storages, the processing to recover the redundancy is executed. A redundancy recovery processing server device,
Data with reduced redundancy from the copy source storage to the copy destination storage, with any of the plurality of storages having no failure as the copy source storage and the storage used to restore redundancy as the copy destination storage A copy means for executing redundancy recovery processing for copying the same data as
A lock that acquires a lock from a lock management server to perform exclusion in units of files and directories so that operations for changing to files and directories stored in the plurality of storages are not performed simultaneously from the host and the copy unit And a redundancy recovery processing server device. When the same data is written from the host, the data is made redundant. When a failure occurs in one of the multiple storages, the processing to recover the redundancy is executed. A redundancy recovery processing server device,
Data with reduced redundancy from the copy source storage to the copy destination storage, with any of the plurality of storages having no failure as the copy source storage and the storage used to restore redundancy as the copy destination storage A copy means for executing redundancy recovery processing for copying the same data as
When the host that has accessed the copy destination storage during the redundancy recovery process discovers that the upper directory of the file or directory to be changed does not exist, an incomplete processing notification specifying the upper directory that did not exist is issued. Incomplete processing notification storage means for receiving and storing the incomplete processing notification when transmitted,
The redundancy recovery processing server apparatus, wherein the copying means copies a non-existing upper directory specified by an incomplete processing notification and its subordinate directories and files from a copy source storage to a copy destination storage. A host computer that makes data redundant by writing the same data to multiple storages.
If the redundancy decreases due to a failure in one of the multiple storages, use one of the multiple storages that has no failure as the copy source storage and restore the redundancy. During the redundancy recovery processing in the redundancy recovery processing server for executing redundancy recovery processing for copying the same data as the data with reduced redundancy from the copy source storage to the copy destination storage. A lock means for acquiring a lock from a lock management server for performing exclusion in units of files and directories so that change operations to the files and directories stored in the plurality of storages are not performed at the same time;
When accessing the copy destination storage during the redundancy recovery process, if it is found that there is no upper directory of the file or directory to be changed, the upper directory that did not exist is copied from the copy source storage to the copy destination storage. And a program for functioning as a storage access means for changing the file or directory to be changed.   In the path name resolution process, the storage access means discovers that the upper directory of the file or directory to be changed does not exist and copies the non-existing upper directory from the copy source storage to the copy destination storage. The program according to claim 19. A host computer that makes data redundant by writing the same data to multiple storages.
If the redundancy decreases due to a failure in one of the multiple storages, use one of the multiple storages that has no failure as the copy source storage and restore the redundancy. The storage to be copied is the copy destination storage, the redundancy recovery process is executed to copy the same data as the data with reduced redundancy from the copy source storage to the copy destination storage, and the presence specified in the incomplete processing notification During the redundancy recovery processing in the redundancy recovery processing server having a function of copying the upper directory that has not been performed and the directories and files under the upper directory to the copy destination storage, the upper directory is stored in the plurality of storages. Changes to files and directories cannot be performed at the same time. And locking means for acquiring a lock from the lock management server for performing an exclusive with files and directories unit as,
A storage access means for discovering that there is no upper directory of the file or directory to be changed when accessing the copy destination storage during the redundancy recovery process;
A program for functioning as an incomplete processing notification means for transmitting an incomplete processing notification specifying a discovered non-existing upper directory to the redundancy recovery processing server.   The program according to claim 21, wherein the storage access means executes discovery in a path name resolution process that a file or directory to be changed does not exist. When the same data is written from the host, the data is made redundant. When a failure occurs in one of the multiple storages, the processing to recover the redundancy is executed. A computer constituting the redundancy recovery processing server device
Data with reduced redundancy from the copy source storage to the copy destination storage, with any of the plurality of storages having no failure as the copy source storage and the storage used to restore redundancy as the copy destination storage A copy means for executing redundancy recovery processing for copying the same data as
A lock that acquires a lock from a lock management server to perform exclusion in units of files and directories so that operations for changing to files and directories stored in the plurality of storages are not performed simultaneously from the host and the copy unit Program to function as a means. When the same data is written from the host, the data is made redundant. When a failure occurs in one of the multiple storages, the processing to recover the redundancy is executed. A computer constituting the redundancy recovery processing server device
Data with reduced redundancy from the copy source storage to the copy destination storage, with any of the plurality of storages having no failure as the copy source storage and the storage used to restore redundancy as the copy destination storage A copy means for executing redundancy recovery processing for copying the same data as
When the host that has accessed the copy destination storage during the redundancy recovery process discovers that the upper directory of the file or directory to be changed does not exist, an incomplete processing notification specifying the upper directory that did not exist is issued. Functioning as incomplete processing notification storage means for receiving and storing the incomplete processing notification when transmitted, and
The copy unit copies a non-existing upper directory specified by an incomplete processing notification and its subordinate directories and files from a copy source storage to a copy destination storage.

Images