JP2014191754A - Storage system, file server, cache management method and cache management program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a storage system, a file server, a cache management method, and a program which allow disk access load to be suppressed when fail-over occurs.SOLUTION: The storage system includes a plurality of file servers 10-1 to 10-n, and each file server includes a cache management part 11 which, when complying with an access request for a file stored in a file system area shared between the file server and the other file servers, transfers a page cache corresponding to the access request to the other file servers together with identification information of the file and identification information of the file system area. The cache management part 11 stores a page cache, identification information of a file, and identification information of a file system area, which are received from another file server, in association with each other and, when complying with an access request for the stored file, uses the associated page cache to responds to the access request.

Description

  The present invention relates to a storage system in which a plurality of servers share storage, a file server in the storage system, a cache management method, and a cache management program.

  In a cluster-connected NAS (Network Attached Storage) (hereinafter referred to as a cluster NAS), when a file server (hereinafter also referred to as a node) is stopped due to a failure or maintenance, the service is performed on another node due to failover. Will continue. However, in that case, the read page cache (hereinafter simply referred to as “page cache”) cannot be taken over to another node. Therefore, the other node that has taken over the service must read the data directly from the disk. Therefore, at the time of failover, the disk load due to access from other nodes increases rapidly, and there is a possibility that system failure such as service down may occur.

  In recent years, the capacity of DIMMs (Dual Inline Memory Modules) has been reduced and the capacity of NAS read page cache memories has been increased. When failover occurs in such an environment, the disk load due to file access from the node that has taken over the service further increases, and the possibility of a system failure increases.

  There is a method to re-evaluate each communication path to access the copy destination volume and reset the priority of each communication path according to the current status of the storage system when a configuration change or failure occurs in the storage system (For example, refer to Patent Document 1).

Special table 2012-531654 gazette

  The method described in Patent Document 1 increases the possibility that an active path that is currently being used will continue to be used as it is, and can suppress the occurrence of failover from the active path to the standby path. However, the method described in Patent Document 1 does not consider the disk access load when failover occurs.

  Accordingly, an object of the present invention is to provide a storage system, a file server, a cache management method, and a cache management program that can suppress a disk access load when a failover occurs.

  The storage system according to the present invention includes a plurality of file servers. When the file server receives an access request for a file stored in a file system area shared with other file servers, the page cache corresponding to the access request is received. A cache management unit that transfers to the other file server together with the identification information of the file and the identification information of the file system area, and the cache management unit receives the page cache and file identification information received from the other file server, If the access request for the file stored in the file system area is accepted when the other file server is stopped, a response is made using the page cache. Features to do To.

  When a file server according to the present invention receives an access request for a file stored in a file system area shared with another file server, the file cache stores the page cache corresponding to the access request, the identification information of the file, and the file system. A cache management unit that transfers to the other file server together with the area identification information. The cache management unit associates the page cache, the file identification information, and the file system area identification information received from the other file server. When the access request to the file stored in the file system area is received when the other file server is stopped, the page cache is used to respond.

  In the cache management method according to the present invention, when an access request for a file stored in a file system area shared with another file server is received, the page cache corresponding to the access request is stored in the file identification information and the file. Transfers the system area identification information to another file server and retains the page cache, the file identification information, and the file system area identification information received from the other file server, and stops the other file server. In some cases, when an access request to the file stored in the file system area is received, a response is made using the page cache.

  When the cache management program according to the present invention receives an access request for a file stored in a file system area shared with another file server, the cache management program displays the page cache corresponding to the access request as identification information of the file. And the process of transferring to the other file server together with the identification information of the file system area, the page cache and the file identification information received from the other file server, and the identification information of the file system area in association with each other. When an access request to the file stored in the file system area is received when the file server is stopped, a process of responding using the page cache is executed.

  According to the present invention, it is possible to suppress a disk access load when a failover occurs.

1 is a block diagram showing a configuration of a first embodiment of a storage system according to the present invention. It is a block diagram which shows the mode of the storage system of normal times. It is a flowchart which shows operation | movement of the file server at the time of file read. It is a block diagram which shows the mode of a storage system when a failure occurs in a file server. It is a flowchart which shows the file system mounting process at the time of failure occurrence. It is a block diagram which shows the mode of the storage system at the time of failback. It is a flowchart which shows operation | movement of the file server at the time of failback. 1 is a block diagram showing a minimum configuration of a storage system according to the present invention.

Embodiment 1. FIG.
A first embodiment of the present invention will be described below with reference to the drawings.

  FIG. 1 is a block diagram showing a configuration of a first embodiment of a storage system according to the present invention.

  As shown in FIG. 1, the file server 300 and the file server 301 can communicate using an inter-cluster communication function 350. The inter-cluster communication function 350 controls communication between file servers.

  The file servers 300 and 301 and the client 330 are connected via a network 340. The file servers 300 and 301 are connected to the shared disk 320. In FIG. 1, one client 330 is illustrated, but any number of clients may be connected to the network 340.

  The shared disk 320 is a disk device, and includes a file system area (hereinafter simply referred to as a file system) 321 and a file system 322. Any number of file systems may be included in the shared disk.

  The file system 321 is normally mounted on the file server 300. The file system 322 is normally mounted on the file server 301. In the present embodiment, a file 323 is stored on the file system 321. A server IP address 360 shown in FIG. 1 is an IP address used when the client 330 accesses the file system 321 using the file sharing service of the file server.

  The file server 300 includes a VFS (Virtual File System) 200 having the extended cache management unit 100 and a cluster control function 310. The configuration of the file server 301 is the same as that of the file server 300, and thus the description thereof is omitted. In FIG. 1, two file servers are illustrated, but any number of file servers may be provided.

  The extended cache management unit 100 of the VFS 200 has an extended cache management function for sharing a page cache with other file servers. The extended cache management unit 100 includes a cache creation unit 110, a cache transfer unit 120, and a cache copy unit 130.

  The cache creation unit 110 creates a page cache in response to file access from the client 330. The cache creation unit 110 checks whether or not there is a read cache table corresponding to the file system to be accessed at the timing when the page cache is created.

  If there is no read cache table, the cache creation unit 110 creates a read cache table. The read cache table includes an index node (hereinafter referred to as an i-node number) that is file identification information (hereinafter referred to as an i-node number) and information for managing a page cache (hereinafter referred to as an address space (address_space (as)). It is a table for registering an entry including information (pointer) for tracing a structure). The cache creation unit 110 creates a read cache table for each file system. In addition, the cache creation unit 110 stores a device ID for identifying the corresponding file system in the read cache table.

  When there is a file access request, the cache creation unit 110 checks whether there is an entry including the i-node number corresponding to the file in the corresponding read cache table. If there is no entry including the i-node number corresponding to the file, the entry is added.

  The cache transfer unit 120 obtains the device ID stored in the read cache table 160, the i-node number included in the entry of the read cache table 160, and the page cache traced from the pointer of the as structure included in the entry. To the extended cache management unit 101 of the opposing file server 301 using the inter-cluster communication function 350.

  When the cache copy unit 130 receives information (device ID, i-node number, page cache) from the cache transfer unit 121 of the opposing file server 301, the cache copy unit 130 newly creates an as structure. Then, the cache copy unit 130 connects the received page cache for each index under the newly created as structure. Specifically, the page cache pointer is stored in the as structure. The cache copy unit 130 registers an entry including an i-node number and a pointer to a newly created as structure in the read cache table 161. Further, the cache copy unit 130 stores the received device ID in the read cache table 161.

  The extended cache management unit 100, the cache creation unit 110, the cache transfer unit 120, and the cache copy unit 130 in the extended cache management unit 100 are realized by, for example, a computer that operates according to a cache management program. In this case, the CPU reads the cache management program and operates as the extended cache management unit 100, the cache creation unit 110 in the extended cache management unit 100, the cache transfer unit 120, and the cache copy unit 130 according to the program. Further, the extended cache management unit 100, the cache creation unit 110, the cache transfer unit 120, and the cache copy unit 130 in the extended cache management unit 100 may be realized by separate hardware.

  Next, the operation of this embodiment will be described.

  As the cluster configuration of the file server, there are an active-active configuration and an active-standby configuration. The active-active configuration can be regarded as a configuration with two sets of active-standby configurations. Therefore, in this embodiment, a case where the cluster configuration is an active-standby configuration is taken as an example.

  FIG. 2 is a block diagram showing the state of the storage system in a normal state.

  First, an operation when the client 330 reads the file 323 on the file system 321 for the first time using the server IP address 360 when the cluster state is normal will be described.

  FIG. 3 is a flowchart showing the operation of the file server during file reading.

  When receiving a read request from the client 330, the VFS 200 accesses the shared disk 320 using an existing function (step S301). The VFS 200 creates a file management structure 210, a deentry cache (deentry cache) 220, and an i-node cache (i-node cache) 230. The deentry is information that associates a file management structure with an i-node.

  The VFS 200 stores the device ID indicating the file system 321 and the i-node number indicating the file 323 in the i-node cache 230. In this way, management information (file management structure 210, deentry cache 220, i-node cache 230) of the read request target file 323 is created (step S302).

  Here, since the page cache has not been created (No in step S303), the VFS 200 reads data from the disk up to the offset requested to be read (step S304). As a result, the page cache 150 including the as structure 140 and the index corresponding to the offset is created. The VFS 200 stores the pointer to the created page cache 150 in the as structure 140. In addition, a pointer to the as structure 140 is set in the i-node cache 230.

  At the timing when the creation of the page cache is completed, the cache creation unit 110 acquires a device ID, an i-node number, and a pointer to the as structure 140 from the i-node cache 230. Then, the cache creation unit 110 creates a read cache table 160 corresponding to the acquired device ID. The cache creation unit 110 adds an entry 170 including the acquired i-node number and a pointer to the as structure 140 to the created read cache table 160.

  When the counterpart node (file server 301) is operating (Yes in step S305), the cache transfer unit 120 sends the page cache to the extended cache management unit 101 of the file server 301 together with the device ID and inode number. Transfer (step S306).

  After step S306, the VFS 200 transmits the data stored in the page cache to the client 330 as a response to the read request (step S307).

  If the page cache 150 has already been created in step S303 (Yes in step S303), the processes in steps S304 to S306 are not executed. In step S305, when the counterpart node is not operating (No in step S305), the process in step S306 is not executed.

  When the extended cache management unit 101 of the file server 301, specifically, the cache copy unit 131 receives the page cache, device ID, and inode number from the cache transfer unit 120 of the file server 300 (step S308), the as structure As structure 141 corresponding to 140 is newly created. A page cache 151 corresponding to the received page cache (page cache 150) is created. The cache copy unit 131 creates a read cache table 161 corresponding to the received device ID. The cache copy unit 131 registers an entry 171 including the received i-node number and a pointer to the as structure 141 in the read cache table 161.

  In this way, the page cache for the file 323 on the file system 321 is copied from the file server 300 to the file server 301 (step S309).

  Next, an operation when a failure occurs in the file server 300 will be described.

  FIG. 4 is an explanatory diagram showing the state of the storage system when a failure occurs in the file server 300. FIG. 5 is a flowchart showing the file system mounting process when a failure occurs.

  As shown in FIG. 4, when the file server 300 becomes unusable, the processing of the file server 300 is failed over by general cluster control. That is, the processing of the file server 300 is performed by the file server. 301 is taken over. Accordingly, the file system 321 is mounted on the file server 301 side (step S501). A server IP address 360 is assigned to the file server 301 side.

  At this time, after the file system 321 is mounted, the file server 301, specifically, the extended cache management unit 101 of the VFS 201 refers to the read cache table corresponding to the device ID of the file system 321 (step S502).

  Here, since the read cache table 161 corresponding to the device ID exists (Yes in step S503), the extended cache management unit 101 further checks whether or not an entry is registered in the read cache table 161 (step S503). S504).

  Here, since the entry 171 is registered (Yes in step S504), the extended cache management unit 101 makes a disk I / O (Input) as if there was an access request to the file corresponding to the i-node number of the entry 171. / Output) is performed (step S505). As a result, file management information (file management structure 211, deentry cache 221, i-node cache 231) is created. Next, the extended cache management unit 101 sets a pointer to the as structure 141 included in the entry 171 in the i-node cache 231 (step S506).

  If the read cache table 161 does not exist in step S503 (No in step S503), the extended cache management unit 101 ends the process. If no entry is registered in step S504 (No in step S504), the extended cache management unit 101 ends the process.

  After step S506, the extended cache management unit 101 checks whether there are other unprocessed entries in the read cache table 161 (step S507). If there is an unprocessed entry (No in step S507), the extended cache management unit 101 executes the processes in steps S505 to S606 with the unprocessed entry as a processing target (step S508).

  If there is no unprocessed entry (No in step S507), the extended cache management unit 101 completes the file system mounting process.

  Thereafter, for example, when the client 330 makes a read request for the file 323 on the file system 321 using the server IP address 360, the file server 301 returns a response using the page cache 151.

  As described above, the page cache shared in advance between the extended cache management units 100 and 101 is used for file access after failover. Thereby, the number of disk accesses is reduced.

  Finally, the operation of the extended cache management unit when the file server 300 recovers and fails back will be described.

  FIG. 6 is a block diagram showing the state of the storage system at the time of failback. FIG. 7 is a flowchart showing the operation of the file server at the time of failback.

  When the cluster control function 311 of the file server 301 detects that the file server 300 has been recovered and can be failed back, it notifies the extended cache management unit 101 to execute the failback. Specifically, the service target file system 321 and the server IP address 360 are notified of failback (step S701).

  Upon receiving the notification, the extended cache management unit 101 checks the device ID of the file system 321 (step S702). The extended cache management unit 101 checks whether there is a read cache table corresponding to the device ID (step S703).

  Here, since the read cache table 161 corresponding to the device ID exists (Yes in step S703), the extended cache management unit 101 checks whether an entry is registered in the read cache table 161 (step S704). .

  Here, since the entry 171 exists (Yes in step S704), the extended cache management unit 101 instructs the cache transfer unit 121 to execute cache transfer for the entry 171. The cache transfer unit 121 executes cache transfer (step S705).

  The extended cache management unit 101 checks whether there are other unprocessed entries in the read cache table 161 (step S706). If there is an unprocessed entry (No in step S706), the extended cache management unit 101 executes the processes in steps S705 and S706 with the unprocessed entry as a processing target (step S707).

  In step S703, when the read cache table 161 does not exist (No in step S703), the processes in steps S704 to S707 are not executed. If no entry exists in step S704 (No in step S704), the processes in steps S705 to S707 are not executed.

  When the extended cache management unit 101 completes the transfer process for all entries (Yes in step S706), it notifies the cluster control function 311 of the completion of the cache transfer (step S708).

  Based on the information received from the cache transfer unit 121, the extended cache management unit 100 of the file server 300, specifically, the cache copy unit 130, recreates an as structure and a page cache corresponding to each entry (step). S709, S710).

  The cluster control function 311 of the file server 301 performs failback processing using the existing function after the cache transfer is completed. Then, the cluster control function 311 mounts the file system 321 on the file server 300 side and assigns a server IP address 360 to the file server 300 side.

  At this time, when the file system 321 is mounted, the file server 300 proceeds with processing in the same manner as during failover according to the flow shown in FIG. In this way, the cache state in the file server 300 becomes the same as the normal state shown in FIG.

  In this way, the page cache can be shared between the nodes of the cluster NAS.

  As described above, in this embodiment, in the cluster NAS, the read page cache of the file sharing service in a certain node is copied and transferred to another node, and the read page cache is held in both nodes. By sharing the read page cache between the nodes in this way, it is possible to take over the read page cache when the file sharing service moves to another node at the time of failover. That is, the standby node can use the read page cache immediately after failover. Therefore, the disk access load from the standby node to the shared disk at the time of failover can be reduced. Therefore, it is possible to avoid a system failure that occurs immediately after failover due to a sudden increase in the disk access load on the shared disk.

  Further, since the cache hit can be detected on the file server side closer to the client side than the shared disk, the response performance to the client can be improved even when the shared disk cache is hit.

  FIG. 8 is a block diagram showing the minimum configuration of the storage system according to the present invention. As shown in FIG. 8, the storage system includes a plurality of file servers 10-1 to 10-n (corresponding to the file server 300 and the file server 301 shown in FIG. 1), and a file server (file server shown in FIG. 1). Is equivalent to 300.) is a page corresponding to the access request when an access request for a file stored in a file system area shared with another file server (corresponding to the file server 301 shown in FIG. 1) is received. A cache management unit 11 (file server shown in FIG. 1) that transfers the cache to another file server together with the identification information of the file (corresponding to the i-node number) and the identification information of the file system area (corresponding to the device ID). 300 to the extended cache management unit 100 in the VFS 200 The cache management unit 11 stores the page cache, the file identification information, and the file system area identification information received from another file server in association with each other, and when the other file server is stopped, When an access request for the file stored in the file system area is received, a response is made using the page cache.

  According to such a configuration, the read page cache can be shared between nodes. Therefore, it is possible to take over the read page cache when the file sharing service moves to another node at the time of failover. That is, the standby node can use the read page cache immediately after failover. Therefore, the disk access load from the standby node at the time of failover can be reduced.

  In the above embodiment, the following storage system is also disclosed.

(1) The cache management unit 11 creates a read cache table for managing the page cache for each file system area shared with other file servers, and there is an access request to a file stored in the file system area In this case, the cache creation unit 12 (in the extended cache management unit 100 in the VFS 200 in the file server 300 shown in FIG. 1) registers an entry including information that can identify the file and page cache corresponding to the access request in the read cache table. And a cache transfer unit 13 (see FIG. 1) that refers to each entry registered in the read cache table and transfers the page cache corresponding to each entry to another file server. 30 Corresponding to the cache transfer section 120 in the expanded cache management unit 100 in VFS200 in.) And a storage system including a.

  According to such a configuration, the cache can be managed for each file system area and for each file access.

(2) Upon receiving the page cache, file identification information, and file system area identification information from another file server, the cache management unit 11 creates a read cache table corresponding to the file system area, A cache copy unit 14 that registers an entry including information that can identify the page cache in the read cache table (corresponding to the cache copy unit 130 in the extended cache management unit 100 in the VFS 200 in the file server 300 shown in FIG. 1). The cache management unit 11 executes a process of associating the page cache with the file corresponding to the page cache for each entry registered by the cache copy unit 14 when another file server is stopped. Storage system to a usable state the page cache.

  According to such a configuration, it is possible to respond to a file access request using a page cache received from another file server. Therefore, for example, even when the page cache data for reading is volatilized when the NAS is restarted, the disk load due to subsequent file access can be reduced.

  Further, the following file server is also disclosed in the above embodiment.

(3) When an access request for a file stored in a file system area shared with another file server is received, the page cache corresponding to the access request is displayed with the identification information of the file and the identification information of the file system area. The cache management unit 11 also transfers the page cache, the file identification information, and the file system area identification information received from the other file server in association with each other. A file server that responds using the page cache when an access request for the file stored in the file system area is received when another file server is stopped.

  According to such a configuration, the read page cache can be shared between nodes. Therefore, it is possible to take over the read page cache when the file sharing service moves to another node at the time of failover. That is, the standby node can use the read page cache immediately after failover. Therefore, the disk access load from the standby node at the time of failover can be reduced.

(4) The cache management unit 11 creates a read cache table for managing the page cache for each file system area shared with other file servers, and there is an access request to a file stored in the file system area The cache creation unit 12 that registers an entry including information that can identify the file and page cache corresponding to the access request in the read cache table, and refers to each entry registered in the read cache table. And a cache transfer unit 13 that transfers a page cache corresponding to the above to another file server.

  According to such a configuration, the cache can be managed for each file system area and for each file access.

(5) Upon receiving the page cache, file identification information and file system area identification information from another file server, the cache management unit 11 creates a read cache table corresponding to the file system area, The cache management unit 11 includes an entry including information that can identify the page cache in the read cache table. When the other file server is stopped, the cache management unit 11 A file server that executes processing for associating a page cache and a file corresponding to the page cache for each registered entry, and makes the page cache usable.

  According to such a configuration, it is possible to respond to a file access request using a page cache received from another file server. Therefore, for example, even when the page cache data for reading is volatilized when the NAS is restarted, the disk load due to subsequent file access can be reduced.

10-1 to 10-n, 301, 302 File server 11 Cache management unit 100, 101 Extended cache management unit 110, 111 Cache creation unit 120, 121 Cache transfer unit 130, 131 Cache copy unit 200, 201 VFS
310, 311 Cluster control function 320 Shared disk 321, 322 File system area 323 File 330 Client 340 Network 350 Inter-cluster communication function 360 Server IP address

Claims (8)

With multiple file servers,
The file server is
When an access request for a file stored in a file system area shared with another file server is received, the page cache corresponding to the access request is displayed together with the identification information of the file and the identification information of the file system area. Including a cache management unit that transfers to the file server
The cache management unit holds the page cache, file identification information, and file system area identification information received from the other file server in association with each other, and when the other file server is stopped, the file system area When an access request for the file stored in is received, the storage system responds using the page cache. The cache management department
Create a read cache table for managing the page cache for each file system area shared with other file servers, and respond to the access request when there is an access request to a file stored in the file system area A cache creation unit that registers an entry including information that can identify a file and a page cache in the read cache table;
The storage system according to claim 1, further comprising: a cache transfer unit that refers to each entry registered in the read cache table and transfers a page cache corresponding to each entry to the other file server. The cache management department
When the page cache, file identification information, and file system area identification information are received from another file server, a read cache table corresponding to the file system area is created, and the information that can identify the file and the page cache Including a cache copy unit that registers an entry including
The cache management unit executes a process of associating a page cache with a file corresponding to the page cache for each entry registered by the cache copy unit when the other file server is stopped, The storage system according to claim 2, wherein the cache is made available. When an access request for a file stored in a file system area shared with another file server is received, the page cache corresponding to the access request is displayed together with the identification information of the file and the identification information of the file system area. Including a cache management unit that transfers to the file server
The cache management unit holds the page cache, file identification information, and file system area identification information received from the other file server in association with each other, and when the other file server is stopped, the file system area A file server that responds by using the page cache when an access request for the file stored in is received. The cache management department
Create a read cache table for managing the page cache for each file system area shared with other file servers, and respond to the access request when there is an access request to a file stored in the file system area A cache creation unit that registers an entry including information that can identify a file and a page cache in the read cache table;
The file server according to claim 4, further comprising: a cache transfer unit that refers to each entry registered in the read cache table and transfers a page cache corresponding to each entry to the other file server. The cache management department
When the page cache, file identification information, and file system area identification information are received from another file server, a read cache table corresponding to the file system area is created, and the information that can identify the file and the page cache Including a cache copy unit that registers an entry including
The cache management unit executes a process of associating a page cache with a file corresponding to the page cache for each entry registered by the cache copy unit when the other file server is stopped, The file server according to claim 5, wherein the cache is made available. When an access request for a file stored in a file system area shared with another file server is received, the page cache corresponding to the access request is displayed together with the identification information of the file and the identification information of the file system area. To the file server
The page cache, file identification information, and file system area identification information received from the other file server are stored in association with each other, and the file stored in the file system area when the other file server is stopped. A cache management method characterized by responding using the page cache when an access request is received. On the computer,
When an access request for a file stored in a file system area shared with another file server is received, the page cache corresponding to the access request is displayed together with the identification information of the file and the identification information of the file system area. Transfer to the file server of
The page cache, file identification information, and file system area identification information received from the other file server are stored in association with each other, and the file stored in the file system area when the other file server is stopped. A cache management program for executing a response process using the page cache when an access request is received.

Images