JPS6370974A - Recovery processing system for duplex volume - Google Patents

Abstract

PURPOSE:To eliminate the needs for discriminating a copied area and a non- copied area by executing a preformat processing in which the physical record form of a copy volume is matched to that of a copy volume, permitting a copy processing and a general input output processing to uniformly access to both volumes, and alternately executing them. CONSTITUTION:Before the copy processing 4 starts, the preformat processing 3 is executed in which the physical record form of the copied volume is matched to that of the copy volume. To put it concretely, the content of a volume 1a in an 'ONDUAL' state is copied to a volume 1b in the 'NEUTRAL' state as it is. Since data of the same record form are written in both volumes, updated data by a general input/output processing is written in the same addresses of both volumes in the copied area, and in the non-copied area, same data are already written in both volumes. Thus, the discrimination of the copied area and the non-copied area is eliminated in the recovery processing of a duplex volume.

Description

[Detailed Description of the Invention] [Table of Contents] Overview Industrial Application Fields Conventional Technology Problems to be Solved by the Invention Means for Solving the Problems Actions Examples Effects of the Invention [Summary] Each volume unit In a duplex direct access storage device, the physical record format of the volume to be copied must be changed prior to the copying process that is performed when recovering a single volume from a single-sided state in which it has been separated due to a failure to a duplex state in which both volumes are valid. By performing preformat processing to match the record format of the copy volume, it is possible to uniformly access both volumes during the input/output processing when the copy processing and general input/output processing are executed alternately later. This is what I did.

[Industrial application field]

The present invention relates to a recovery processing method for a duplicated volume in a direct access storage device (DASD), such as a disk device, in which each volume is duplicated.

In current computer systems, especially online systems, most of the data processed by the system is stored semi-permanently or temporarily on direct access storage devices (DASD).
stored in the upper volume.

Therefore, when it becomes impossible to process data on the volume of the direct access storage device (DASD) due to a hardware failure, it will have a serious impact on the computer system.

Therefore, the direct access storage device (DASD)
By duplicating and using the above data in units of volumes, a configuration (for example, a know-down system configuration) is adopted in which the impact of the above hardware failure on the system is significantly reduced.

Specifically, this is a configuration in which two physically independent volumes having the same data are used to logically form one volume.

In such a direct access storage device (DASD) that is duplexed for each volume, when one of the volumes is disconnected due to a failure etc. and is providing online service with only one lung, when returning to a duplexed state where both volumes are valid, the online In a system, it is necessary to perform the recovery process without stopping services to users.

In particular, in online systems such as banking systems, due to the serious impact that a service outage would have on users, the know-down system is preferred, with direct access storage devices (13
ASIN), we have built a redundant system in which not only the terminals installed at each branch but also one main unit are completely duplicated.

A direct access storage device (which is shared between systems in such a redundant system and is duplicated for each volume)
If one system is performing the above recovery process for DASD), the other system cannot know the recovery status, so for example, the recovery process and general input/output process may be alternated. (double-sided light is required for general input/output processing)
An effective recovery processing method for a duplex volume that is shared between systems cannot be recognized as an uncopied area (in general input/output processing, it is necessary to write only to the copying volume). is now in demand.

[Conventional technology]

FIG. 4 is a diagram illustrating a conventional duplex volume recovery processing method.

As previously mentioned, direct access storage devices (hereinafter referred to as
In order to improve the reliability of DASD
A method has been adopted in which data on a volume is duplicated on a volume-by-volume basis.

In the configuration state of the duplex volume, both volumes 2a
, 2b are valid, and there is a single-sided state 1 in which one volume 1b has been separated due to a failure or the like, and service is provided only by the volume 1a of one lung.

In order to restore the poly neem 1a that has entered the single-sided state 1 to the duplex state 2, for example, a management system provided on the main storage device is installed under the control of an operating system (O3) not shown. A new ASD lb is allocated using the table, and all data in the volume la in the single-sided state 1 is copied to the [) ASD lb.

In the conventional method, during this copying process, access from the service program to the volumes la and Ib is suppressed, and input/output processing requests from the service program that arrive at that time are handled by an input/output management program (not shown). The copying process was suspended within the process, and was controlled to be executed after the copying process was completed.

[Problem that the invention seeks to solve]

The conventional duplex volume recovery processing method described above has a problem in that during the copying process, certain services are in a state that is almost equivalent to being interrupted.

Therefore, restoration from single-sided state 1 to duplex state 2 during online service was not performed.

Therefore, as a measure to solve the above problem, the copying process executed in the recovery process from the single-sided state 1 to the duplex state 2,
One possible method is to execute general input/output request processing (access from a service program) to the volume in parallel, but in order to do so, the following points had to be solved.

FIG. 5 is a diagram illustrating problems in executing copy processing and input/output processing in parallel.

fl) Between read ■ and write ■ of copy processing,
- If the update write (■) of the FG input/output process is entered, the above update write (■) will be invalidated by the subsequent write (■) of the copy process, and the update input/output will be treated as if it were inactive, which is a problem for the user. There was a problem in that it appeared to be data destruction.

(2) Also, among -S input/output requests, if the access type is read, it is sufficient to uniformly access the volume on the copy side 1a, but in the case of write, the copy method part (shaded When writing to a volume (indicated by There is a need to do.

Here, '0NDUAL' indicates a volume that has been allocated and is valid, and 'NEIJTRAL' indicates a volume that has been allocated but is invalid (without valid data).

(3) Since read/write for copying is sequential access from the smallest cylinder position, the arm operation time (seek time) is close to 0°, but compared to general input/output requests. When input/output operations occur randomly, the access positions are different from each other, causing disturbances in the arm operations and reducing the overall throughput.

(4) While the copying process is in progress, the response time for general input/output requests is unavoidably affected, but the degree of this impact is
It is necessary to be able to adjust according to the system load situation.

FIG. 6 is a diagram illustrating a conventional recovery processing method for a duplexed volume shared between systems, and FIG. 6(a)
is a diagram schematically showing the recovery process, and FIG. 6(b)
is a diagram explaining the problem.

First, in FIG. 6(a), to explain the operation of recovery processing for a duplexed volume that is shared between systems, job rAJ70a, which is a general input/output process, and recovery job 70b run in parallel operation, which will be described later. System O
In the system 70, it is possible to distinguish between the copy method area of the DASD in the copy processing state and the uncopied area.
Although the problem (2) at the time of read/write access can be solved, in the system 1 system 71 where the recovery job 70b is not running, it is not possible to distinguish between the copy method area and the uncopied area. The problem is that it cannot be done.

In order to solve this problem, as shown in Figure 6(b),
For example, every time copying of one cylinder unit is completed, the position of the copying method area is notified to other systems (system 1 system) 71 etc. through the intersystem communication bus, and the management information in each system is immediately transmitted. It is necessary to do so.

For this reason, for example, each time the copying process for one cylinder is completed, communication between systems is required, and the overhead in the system (system 0 system) 70 where the recovery process is running becomes large. There was a problem in that the throughput at 70 was reduced.

In view of the above-mentioned drawbacks of the conventional technology, the present invention provides a DASD that is duplexed for each volume, especially a DASD that is shared between systems.
In D, it is an object of the present invention to provide a method for resolving the above-mentioned problems while allowing normal access to the user from a service program even during the restoration process when returning from a single-sided state to a duplex state.

, [Means for Solving the Problems] FIG. 1 is a diagram showing an example of the configuration of a duplex volume recovery processing method according to the present invention.

In the present invention, in a direct access storage device in which each volume is duplexed, in order to restore a one-sided state 1 in which one of the duplexed volumes is disconnected due to a failure to a duplexed state 2 in which both volumes are valid, Prior to the copy process 4 executed in the recovery process, a preformat process 3 is performed to match the physical record format of the copied volume to the record format of the copy volume, and then the copy process 4 and general input/output process are performed. When executing these volumes alternately, the input/output processing is configured to uniformly access both volumes.

[Effect]

That is, according to the present invention, in a direct access storage device in which each volume is duplexed, a copy process is performed to restore a single-sided state in which a certain volume is separated due to a failure to a duplexed state in which both volumes are valid. By performing preformat processing to match the physical record format of the copied volume to the record format of the copied volume prior to the subsequent copying process,
In the above input/output processing when the input/output processing of -S is executed alternately, both volumes are uniformly accessed, so the copy method area.

It is no longer necessary to distinguish between uncopied areas, and even for duplex volumes that are shared between systems, serviceability to users is not compromised (recovery operations and general input/output processing can be executed simultaneously, increasing system reliability). It has the effect of improving the degree of

〔Example〕

Embodiments of the present invention will be described in detail below with reference to the drawings.

The above-mentioned FIG. 1 is a diagram showing a configuration example of the duplex volume recovery processing method of the present invention, and FIG. 2 is a diagram showing another configuration example of the duplex volume recovery processing method of the present invention. , FIG. 3 is a diagram explaining the operation of the recovery process according to the present invention, and the preformat process 3. in FIG. The means for alternately executing the recovery processing 40 and the input/output processing 30 shown in FIGS. 2 and 3 are necessary means for carrying out the present invention. Note that the same reference numerals refer to the same objects throughout the plan.

The duplex volume recovery processing method of the present invention will be explained below with reference to FIGS. 1 to 3.

First, FIGS. 2 and 3 show the recovery process when recovering from the single-sided state to the duplex state described above.

A recovery processing method that alternately executes general input/output processing will be described.

First, while the computer system is executing the online service 50, a failure occurs in the duplexed DASD in units of volumes, and when the recovery process 40 occurs to recover from the single-sided state 1 to the duplexed state, access to the DASD is interrupted. For example, the copy processing means 10 for one cylinder is activated via the switching means 60, and recovery processing for the DASD is started. (Figure 3, "nth cylinder copying process"
10) When the copy processing for one cylinder is completed, the copy processing dead period generation means 20 is activated, and the access switching means 60 for the DASD performs the following operations.
Input/output processing 30 originating from online service 50
is selected, and input/output processing for the DAsD is executed. (Refer to Figure 3 "Suspension period" 20) This input/output processing includes read (R) as shown in Figure 3.
), write (-), and update (R/W) processing. For example, a processing request that occurs during the n-th cylinder copy processing period is
There is a waiting period until the copying process 10 is interrupted and a pause period 20 occurs, but since the period of the 8M copying process 10 is limited to, for example, one cylinder, there is little impact on online processing. .

In addition, the execution of the recovery process 40 and the general input/output process 30 is as follows:
As is clear from FIG. 3, since it is serialized, the above problem (1) is solved, and the general input/output processing 30
Since the recovery process 40 is stopped while the copy method area is being executed, it is easy to distinguish between the copy method area and the uncopied area. Accesses that write to both volumes on the copying side and the volume to be copied) and accesses to uncopied areas (accesses that write only to the volumes on the copying side) are performed at least on the system running the recovery process. This can be easily achieved.

In addition, since the copying process is executed in units of one cylinder,
The number of occurrences of disturbances in arm operation is reduced, and the drop in throughput (problem (3) above) can be suppressed to some extent.

Furthermore, by giving the suspension period 20 of the copying process as a parameter, the recovery process 40 is performed.

Load adjustment with the general input/output processing 30 can be performed. (See problem (4) above) However, as described above, for a duplex volume that is shared between systems, the system system (for example, system O system 70) on which the recovery job 70b is running, Although it is possible to distinguish between a copied area and an uncopied area caused by the recovery process, such discrimination is not possible in a system where the recovery process is not running (for example, system 1 system 71).

Therefore, the present invention attempts to solve this problem by performing preformat processing 3 prior to copying processing 4, as shown in FIG.

First, the DAS containing the redundant volume that is shared between systems
During the operation of an online system equipped with D, a failure occurs in the duplexed DASI for each volume, and the above recovery process (see Figures 2 and 3) is performed to recover from single-sided status 1 to duplex status 2. ) 40 and before starting the copy process 4, a preformat process 3 is performed to match the physical record format of the volume to be copied to the record format of the copy volume. (See Figure 3) Specifically, volume 1 in the '0NDUAL' state
Copy the contents of a to volume °1b, which is currently in the 'NEtlTRAL' state.

As a result, the record formats (record addresses, data) of both volumes are the same, so in the subsequent copying process 4, for example, general input data that occurs alternately with the copying process 10 for each cylinder. Write access for output processing can be uniformly performed on both volumes.

In other words, data in the same record format has been written to both volumes (specifically, the same data is stored in each record), so general input cannot be performed on the copied area. Updated data resulting from output processing is only written to the area (record) at the same address in both volumes, and the same data has already been written to both volumes even in uncopied areas, so the same general input/output processing is performed. Copying process 4 will be performed after the updated data has been written to the record at the same address in both volumes, but since the updated data has also been written to the copying side (ONDUAL) la, the copying process 4 will be performed. In the copying process 10 for each cylinder, the side to be copied (NEUTRA
) The same record in the uncopied area of lb is simply overwritten with the data of the copy side (ONDUAL) la, and no contradiction occurs.

In this way, the recovery process for a duplex volume according to the present invention has the feature that it is not necessary to distinguish between a copy method area and an uncopied area.

Therefore, in a computer system in which a DASD, which is duplicated in volume units, is shared between systems, the recovery process is not running (and the system cannot recognize the copy method area and the uncopied area). The DAS from
For general input/output processing to D, it is sufficient to access both duplicated volumes, and recovery processing for the duplicated volume shared between systems can be easily realized.

Of course, it goes without saying that this recovery processing method for duplexed volumes can also be applied to recovery processing for duplexed volumes that are not shared between systems.

In Figure 1, the recovery process shown in route (a) is as follows:
The recovery process 10 and the general input/output process 30 explained in FIG. 2 are executed alternately, and the general input/output process 3
A method that realizes recovery processing for a duplicated volume by performing a write operation with 0 on both volumes for copied areas, but only on the copying volume for uncopied areas. This cannot be applied to recovery processing for duplexed volumes that are shared between systems.

However, the recovery process shown in route (B) shows a method in which copy process 4 is entered after performing preformat process 3 on the non-copy volume, and this is necessary for the recovery process in route (B). Furthermore, since there is no need for processing to distinguish between a copied area and an uncopied area, it can also be applied to recovery processing for a duplexed volume that is shared between systems.

Prior to such recovery processing 40, after performing preformat processing 3 from the copy side volume to the copying side volume, in copy processing 4 in recovery processing 40, 1, for example, copy processing in units of 1 cylinder is performed. In the recovery processing method falling under No. 10, during the preformant processing,
- Since it is not possible to execute input/output processing that performs format writing (change of record format) from input/output processing of C,
This will be suspended until the preformat processing is completed, but such processing is only performed at the beginning of the online system's operation, and rarely occurs during online service, so it is of no practical use. No problem.

Also, when moving from preformat processing 3 to copy processing 4, it is necessary to notify other systems, but since it is only done once, the OHAHEFD for communication between systems is negligible. and is not a problem.

In addition, in the recovery processing including the preformat processing 3 of the present invention, in order to solve the above-mentioned problem (1), the second
The recovery process 40 and general input/output process 30 explained in the figure
Since exclusive control is required with the reserve (BR5V)/release (BRLS) function shown in Figure 3, the reserve (BR5V)/release (BRLS) function is used.

As described above, the present invention changes the physical record format of the replicated volume (NEtlTRAL) to the replicated volume (ONDIJAL) prior to the recovery process performed in the recovery process for a DASD failure that occurs during online service.
Perform preformat processing to match the record format of both volumes to make the contents of each record the same before starting the recovery process. By uniformly performing this on both volumes without distinguishing between areas and uncopied areas, even for duplexed volumes that are shared between systems, the above restoration process from a single-sided state to a duplexed state and general input/output processing can be performed. The feature is that it enables simultaneous execution with

In the above embodiment, the copying process was explained as a method in which copying is performed to l cylinders, but it is not necessary to limit it to one cylinder. For example, it is generally possible to copy to n cylinders as long as it does not affect online services. Needless to say, it is a good thing.

〔Effect of the invention〕

As explained above in detail, the duplex volume recovery processing method of the present invention is capable of converting a one-sided state in which a certain volume is disconnected due to a failure to a state in which both volumes are valid in a direct access storage device that is duplexed for each volume. Prior to the copying process that is performed in the process of restoring the duplex state, preformatting is performed to match the physical record format of the copied volume to the record format of the copy volume. Since both volumes are uniformly accessed during the input/output processing when the output processing and output processing are executed alternately, there is no need to distinguish between copied areas and uncopied areas, and there is no need to distinguish between copied areas and uncopied areas. Even for certain duplex volumes, recovery operations can be performed without compromising service to users.

-i can be executed simultaneously with input/output processing, which has the effect of improving system reliability.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing a configuration example of a duplex volume recovery processing method according to the present invention. FIG. 2 is a diagram showing another configuration example of the duplex volume recovery processing method of the present invention. FIG. 3 is a diagram illustrating the operation of recovery processing according to the present invention. FIG. 4 is a diagram illustrating a conventional duplex volume recovery processing method. FIG. 5 is a diagram illustrating problems in executing copy processing and input/output processing in parallel. FIG. 6 is a diagram illustrating a conventional recovery processing method for a duplex volume shared between systems. It is. In the drawings, 1 is a single-sided state of a direct access storage device (DASD). 1a is an effective volume (ONDUAL). 1b is an invalid volume (NEtiTRAL). 2 is a duplex state of direct access storage devices (DASD). 3 preformat processing. 4 is copy processing. 10 is an n-cylinder copy processing means or an n-cylinder copy process. Reference numeral 20 denotes a copy processing suspension period generation means or suspension period. 30 is input/output processing, 40 is recovery processing 1, and 50 is online service. 60 is a DASD access switching means. 70 is system o system, 71 is system 1 system. ■~■ are the access order to DASD. are shown respectively. Ni'17-Ibi; (tsu (艮艷. ・Tehi's two electric r village water, -mu V) Roi coins, Danri method, θ month, Japanese leap sheep 4 Kuchunegisha Nuyu Rihi shaped sword ``Chanri and Enamida 1'' (2 firings, 4 rows, sword problem Yu, λ theory daily clothes, etc. 5 mouth C) I, the two parties in the yellow old worker who is the fist of water, the chemical body. The troubled ward theory N Σ year 914 σru speech ζ Figure (shi)

Claims (1)

[Claims] In a direct access storage device in which each volume is duplexed, from a single-sided state (1) in which one of the duplexed volumes is disconnected due to a failure, to a duplexed state (2) in which both volumes are valid. To recover, the recovery process (4
Prior to the copying process (4, 10) to n cylinders (n=1, ~) performed in step 0), preformat processing (3) is performed to match the physical record format of the copied volume to the record format of the copy volume. After that, copy processing (4, 10) for the n cylinders (n=1, ~)
1. A recovery processing method for a duplicated volume, characterized in that when the input/output processing and the general input/output processing are executed alternately, both volumes are uniformly accessed in the input/output processing.