JPH05119927A - Duplexed volume processing system - Google Patents

Abstract

PURPOSE:To acquire a final write cylinder as updated information by providing a mechanism to detect the switching of a cylinder and to record the detection as updated information at a device to control a disk device constituting duplexed volumes. CONSTITUTION:Updated information 5 and 6 concerning the respective disk devices of devices X and Y is stored in the non-volatile semiconductor memory of the disk controller, for example. A host 1 manages the volumes while duplexing them by using the disk devices of the devices X and Y, issues an SIO (input/output instruction) to the devices X and Y in the case of updating data, activates the command of a channel or a command chain and writes the data in the respective devices. In that case, the host 1 reports the number of times updating the data and the address of the update starting cylinder to the disk controller 2. The disk controller 2 updates the update information composed of the number of times for update and the update starting and ending cylinders after writing the data in the devices.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION The present invention protects data by performing dual management of data by different disk volumes.
The present invention relates to a dual volume processing system that enhances system reliability, and in particular, enables rapid and reliable recovery of a volume when one disk fails during data update.

[0002]

2. Description of the Related Art In a conventional dual volume processing system, the function of a data management system (IOS) of a host issues an input / output command (SIO) to each volume of two disks to write the same data. Even if a disk fails, the volume of the other disk can continue the work. At the time of writing this data, update information such as the number of updates and the updated cylinder number is recorded so that the last updated cylinder can always be known. The update information is recorded by providing a specific track on the disc or providing a nonvolatile semiconductor memory.

FIG. 5 shows an example of the configuration of such a dual volume processing system. In the figure, 1 is a host, 2 is a disk controller, 3 and 4 are disk devices, and 5 and 6 are disk update information.

When updating data, the host 1 issues SIOs to the disk devices 3 and 4, respectively, and causes the disk control device 2 to write the same data. After updating the data for each disk, the host 1 records the number of updates and update information such as the write cylinder number.

As a result, the update information of the two disks is kept in the same update status. If either of the updates fails, the update information of the two disks is different, so equivalence can be guaranteed by copying the cylinder to the disk with the most or least number of updates.

By the way, this conventional method is effective when the area to be updated by one SIO is limited to one cylinder. For example, the CKD command set in which writing beyond the cylinder boundary cannot be executed. This is true only for SIO that uses a channel command system such as (In the CKD command set, in the case of writing across a cylinder boundary, a positioning instruction to the next cylinder must be issued by issuing CCW (channel command word). Not).

However, in the CKD-EX command set, which is an extended channel command body, WRITE UP
When several DATE DATA (data update write) commands are chained, it is allowed to update the data beyond one cylinder.

For this reason, although the issuer of CCW can manage the last write cylinder in the CKD command set described above, it cannot manage the last write cylinder in the CKD-EX command set. Therefore, the last write cylinder cannot be specified when a failure occurs. If the equivalence of the two disk volumes is lost, the entire disk must be copied or the extent (physically distant storage area) must be restored for recovery.
It was necessary to copy the data in units.

As a result, there has been a problem that the amount of data copied increases and it takes time to restore the data.

[0010]

In the present invention, the host is S
It is an object of the present invention to provide means for enabling the final write cylinder to be acquired as update information in a dual volume processing system in which the final write cylinder cannot be specified when issuing IO.

[0011]

According to the present invention, the above-mentioned problems are solved by providing a disk control device for controlling a disk device forming a duplicated volume with a mechanism for detecting cylinder switching and recording it as update information. Is intended.

FIG. 1 is a principle block diagram of the present invention shown in an exemplary manner. In FIG. 1, reference numeral 1 is a host that performs processing using a duplicated volume.

Reference numeral 2 is a disk control device. Three and four
Is a disk device in which a duplicated volume is constructed. These disk devices will be referred to as device X,
It is represented by device Y.

Reference numerals 5 and 6 are update information of the disk devices of the device X and the device Y, which are stored in, for example, the nonvolatile semiconductor memory of the disk control device. Update information is
It includes the number of updates and each address of the start cylinder and end cylinder at the time of the last update.

Reference numeral 7 denotes a cylinder switching detection mechanism, which is provided in the disk control device 2 and detects when cylinder switching occurs due to continuous data writing to the device X or the device Y.

The update information 5 and 6 are basically set and updated by an instruction from the host 1, but a part is instructed by the cylinder switching detection mechanism 7, and when cylinder switching is detected, the final cylinder address is set. Will be updated.
It should be noted that the host and the disk control device may each be single or plural.

[0017]

In FIG. 1, the host 1 manages the data by duplicating the volume by using the disk device of the device X and the device Y. SIO is issued, a channel command or command chain is activated, and data is written to each device. At that time, the host notifies the disk controller 2 of the number of data updates and the address of the update start cylinder. The disk controller is
When the data writing to the device is completed, the number of updates,
Update information consisting of an update start cylinder and an update end cylinder is updated.

As described above, the channel command system is 1
If only one cylinder is permitted to write for one SIO, the host knows the address of the cylinder to be updated, so the update information always indicates the last updated cylinder of the device. Is to permit writing beyond the cylinder boundary for one SIO issue, the host cannot know the last cylinder but only the writing start cylinder. However, in the present invention, the writing beyond the cylinder boundary is detected by the cylinder switching detection mechanism of the disk controller, and the address of the last cylinder in the update information is updated based on the detection result, so the update information is always the start cylinder of writing. Allows the correct indication of the final cylinder.

For example, for the disks of device X and device Y that compose one duplex volume, 100
Consider a case where the system is out of power when the writing is finished for the device X first and then the writing is performed for the device Y after the second update. At this time, the command chain activated by the SIO issued to the device X is as follows.

DX-LR (CCHH = 02.0D) -WU
D-WUD-WUD-WUD Here, DX is a command instructing seek, LR (CCH
H = 02.0D is a command for instructing a start address (cylinder 2, record 0D), and WUD is a command for instructing data update writing (Write Update Data).
WUD writes in record units, in this case 16
If record / track is set, cylinder switching occurs at the third WUD, and the cylinder address is 2 to 3
Turns into. When the device X is updated, the cylinder switching detection mechanism 7 detects this cylinder switching by reading the index mark appearing next to the last record on the track, and after the update is completed, the number of updates and the update information 5 are updated. Although the start address and the end address of the cylinder are recorded, when the device Y is updated, if the power failure occurs before the end of the cylinder update, the update information 6 is not rewritten.

As a result, the update information 5 of the device X is "1".
Although it is "00th time, cylinders 2 to 3", the update information 6 of the device Y is "99th time, cylinders 4 to 4", etc. and remains the information at the time of the previous update.

When the system is started up after the power failure is restored, the update information 5 and 6 of the devices X and Y in which the duplicated volumes are defined are read and compared. In this case, the fact that the equivalence of the update information of both X and Y systems is broken is detected from the difference in the number of updates, so by analyzing the difference, the 100th time of cylinders 2 to 3 of device X It turns out that it is sufficient to copy the write data of
Equivalence is restored by this copying.

[0023]

FIG. 2 is a block diagram of a duplicated volume processing system according to an embodiment of the present invention. FIG.
2B, a disk control device 2, disk devices 3 and 4 identified by the device “100” and device “200”, update information 5 and 6, and a cylinder switching detection mechanism 7. Has three hosts 1, 1 ', 1 "
And two disk controllers 2, 2'and at least 4
One disk device 3, 3 ', 4, 4', and update information 5,
6 is an example of a system including a cylinder switching detection mechanism 7. The system of FIG. 2 (a) corresponds to the principle configuration shown in FIG. 1, and shows the minimum configuration of the duplicated volume processing system. The system of FIG. 2 (b) has a more complicated configuration than the system of (a), and the device "100",
The disk control device 2 for controlling the disk devices 3, 3'identified by "101" and the devices "200", "20"
A disk controller 2'for controlling disk devices 4, 4'identified by 1 ", and three hosts 1, 1 ',
1 ″ is connected to the disk controllers 2 and 2 ′,
The disk control devices 2 and 2'have cylinder switching detection mechanisms 7 and 7 ', respectively, and manage the update information 5 and 6, and the device "100" and device "200" or the device "101" and device "201". A redundant volume is constructed between the two. An embodiment will be described below centering on the system of FIG.

FIG. 3 is a block diagram of an embodiment of the inside of the disk controller 2 of FIG. 2 (b). The configuration of the disk control device 2'of FIG. 2 (b) is almost the same, so the explanation is omitted. In FIG. 3, 2 is a disk control device, 5 is update information, 7 is a cylinder switching detection mechanism, 8 is a non-volatile memory for storing update information 5, 9 is a command analysis unit for analyzing channel commands, and 10 is update information 5. Is an update information management unit. Update information 5 is created for each host,
It is managed.

Next, the operation of the embodiment shown in FIGS. 2 (b) and 3 will be described using the example of the update information shown in FIGS. 4 (a) and 4 (b). (1) Each host 1, 1 ', 1 "declares the use of a duplicated volume at system startup and requests the host controller's identification number from the disk controllers 2, 2'. (2) Each disk The control devices 2, 2'include the hosts 1, 1 ',
For the request from 1 ″, the identification numbers 00 and 0, respectively
Assign 1,02. (3) The host is an SIO that writes to subsequent devices
In, the CCW is issued by chaining the following two commands at the beginning of the CCW.

SELECT HOST BUFFE
R Informs the disk controller of the host identification number assigned to the host.

WRITE UPDATE INFO
RMATION Transfers the update information created by the host and requests the disk controller to store it. The update information includes the number of updates indicating the number of writes and the start address of the cylinder to be updated by the SIO. (4) When the command analysis unit 9 of FIG. 3 identifies these commands, the update information management unit 10 creates the update information 5 corresponding to the host identification number and stores it in the non-volatile memory 8. (5) Here, the update information 5 in the disk controller 2, 2 ',
6 is as shown in FIG. 4 (a), and the host "00"
The device “100” and the device “200” have been written 45 times, and the host “01” has written the device “100” and the device “200” 60 times. Also the device "100"
And the device “200” have the same start address and end address of the last update cylinder of the duplicated volume. (6) After this, the host "00" has written the 46th time, and the host "01" has written the 61st time, and the system has lost power when the update to the device "100" is completed. I shall. (7) 45th write from host “00”
The writing starts from the cylinder "004" following the last cylinder "003" at the time of the writing for the first time, and the writing is performed to the next cylinder "005" across the cylinder boundary. When this cylinder boundary is crossed, it is detected by the cylinder switching detection mechanism 7 in FIG. 3, and the update information management unit 10 updates the end address of the update cylinder of the update information 5. On the other hand, the 61st write from the host “01” is performed from the same start cylinder “006” to end cylinder “008” as in the 60th write. Since these writes are performed only to the device "100", the update information 5 is updated as shown in FIG. 4B, but the update information 6 is the same as that in FIG. 4A. Therefore, the equivalence between the update information 5 and the update information 6 is lost. (8) When the power failure is restored in this state and the system is started up, the disk controller 2, when the system is IPLed.
The update information 5 and 6 are read from 2'and compared, and the equivalence is checked. From the result, it can be seen that the write statuses of the device “100” and the device “200” are different, so the device “100” to the device “200”, the host “00” to the cylinder “004” to “005”, the host "01" is cylinder "006" to "00"
Up to 8 "are copied respectively. With this copying, the contents of the update information 6 are updated, and the equivalence with the update information 5 is guaranteed.

Since the disk controller can create the number of updates for each device from the write control information itself, and can also acquire and manage the start address of the last updated cylinder, it is possible to manage from the host described in (3). The present invention can be implemented even if the update information transfer process is omitted.

[0029]

According to the present invention, even when the host cannot specify the last updated cylinder by using the extended channel command system such as the CKD-EX command system, the disk controller can surely make the last updated cylinder. Since it is detected and stored as update information, it is possible to restore the duplicated volume with a minimum amount of data copy when a failure occurs.

[Brief description of drawings]

FIG. 1 is a principle configuration diagram of the present invention.

FIG. 2 is a configuration diagram of a duplicated volume processing system according to the embodiment of this invention.

FIG. 3 is a block diagram of an embodiment of a disk control device of the present invention.

FIG. 4 is an explanatory diagram of an example of update information according to the embodiment of the present invention.

FIG. 5 is a configuration diagram of a conventional dual volume processing system.

[Explanation of symbols]

 1 host 2 disk controller 3, 4 disk device 5, 6 update information 7 cylinder switching detection mechanism

Claims (2)

[Claims] 1. A dual volume processing system, wherein the same data is written in respective volumes of two disks, and when a failure occurs in one disk, the volume of the other disk is used to continue the work. A non-volatile memory that stores update information in each of the disk volumes, and the number of updates, the update start cylinder address, and the update end cylinder address in each of the above two disks are updated each time an I / O command is executed. A means for storing the information in the nonvolatile memory as described above and a means for rewriting the address of the update end cylinder in the update information when cylinder switching occurs while writing data to the disk, and If writing to one disk fails, both The disk update information is read and compared, and the data between the update start cylinder and the last cylinder of the disk that has been updated many times is copied to the disk where the above failure occurred, and the equivalence of the duplicated volume is restored. Redundant volume processing system. 2. The dual volume processing system according to claim 1, wherein the update information is provided for each host.