JPS6322340B2 - - Google Patents

Description

[Detailed Description of the Invention] (A) Technical Field of the Invention The present invention uses a checkpoint processing method, particularly in a data processing system, to increase the recovery rate of the environment and prevent destruction when restoring the environment in the event of a system abnormality. This invention relates to a checkpoint processing method that provides strong checkpoints.

(B) Prior art and problems Improving reliability in data processing systems requires
This is one of the most important issues, and for this reason, various so-called checkpoint processing methods are being considered to restore the old environment when a system failure occurs. In particular, checkpoint information regarding frequently updated information, such as various management information for virtual input/output (VIO) volumes, data set update control information, etc., is subjected to generation management. Conventionally, checkpoint information subjected to such generation management has been managed as follows.

FIG. 1 shows an example of a conventional method. For example, as shown in Figure 1A, in order to remember the position of each checkpoint record in the checkpoint data set, each record has a pointer P, and when restoring, the information of the pointer P is used to select the next record. Make the location known. In addition, in the method shown in Figure 1B, a management record MR is provided that records the addresses of all records, and this management record
The position of each record R ₁ to _Ro can be known by MR. However, in the method shown in FIG. 1A, if the pointer P of record R2 is destroyed or becomes unreadable, for example, it becomes impossible to restore the subsequent records R3, R4, . . . . In addition, in the method shown in Figure 1 B,
There was a problem in that if the management record MR was destroyed, each record R ₁ to _Ro could no longer be accessed. Furthermore, both methods have problems in that the recovery process for errors during writing is complicated because it affects multiple records, and the input/output time is long.

(C) Object and Structure of the Invention The present invention aims to solve the above-mentioned problems, and by increasing the independence of each record, reduces the recording time for checkpoint data sets, simplifies recovery processing, and improves the recovery rate of the environment. The purpose is to enable the improvement of Therefore, in the checkpoint processing method of the present invention, the system environment of a data processing system in operation is recorded in a checkpoint data set of an external storage medium, and when a system abnormality occurs, the old environment is restored according to the contents of the checkpoint data set. In the checkpoint processing method for restoring the checkpoint information, the checkpoint data set includes a plurality of cells serving as input/output units of checkpoint information, and a cell map area indicating whether each cell of the plurality of cells is valid or invalid. and a cell management unit that manages n generations of the above cells for each type of information subject to checkpoints, and current time information and information types for cells notified by the cell management unit in response to requests from journal requesters. an output processing section that writes out checkpoint information including an information identifier indicating an information identifier; an input processing section that reads all valid cells according to the map information in the cell map area when restoring the environment; The present invention is characterized in that it includes a restoration processing section that refers to the current time information and the information identifier for the cell and restores the old environment according to the latest checkpoint information for each type of information. Embodiments will be described below with reference to the drawings.

(D) Embodiment of the Invention FIG. 2 is an explanatory diagram of the structure of a checkpoint data set according to the present invention, and FIG. 3 shows the configuration of an embodiment of the present invention.

In the figure, 1 is an external storage device, 2 is a checkpoint data set, 3-1 and 3-2 are cell map areas, 4 is a cell, 10 is a journal request unit,
11 is a restoration processing section, 12 is a checkpoint processor, 13 is an output processing section, 14 is a cell management section,
15 is an input processing section, 16 is an input/output execution section, 17 is a channel, and 18 is a controller.

The checkpoint data set 2 is provided in an external storage device 1 such as a magnetic disk device, particularly in a volume of a direct access device (DASD). The storage area of the checkpoint data set 2 is divided in advance into a plurality of cells, which are input/output units for checkpoint information, as shown in FIG. 2, for example. In addition, checkpoint data set 2
Cell map area 3-
1, 3-2 are provided. In the case of this embodiment, the following description will be made assuming that two generations of cell maps are managed. That is, either cell map area 3-1 or 3-2 holds the latest cell map information, and the other holds second generation cell map information. Cell map information is written in cell map area 3-1 and cell map area 3.
-2, alternately. In the cell map, for example, one bit corresponds to one cell 4, and when the bit is "1", it indicates that the corresponding cell 4 is in use and valid, and when it is "0", it indicates that the corresponding cell 4 is in use and valid. ,
This indicates that the corresponding cell 4 is empty and the information therein is invalid. At the beginning and end positions of each cell map area 3-1, 3-2, real-time information managed by the processing device is provided.
TOD (Time Of Day) is written. When restoring the environment, which of the cell map areas 3-1 and 3-2 holds the latest cell map is determined by comparing this real-time information. Ru. Cell map area 3-
The same data is written to the real time information TOD1 at the beginning of 1 and 3-2 and the real time information TOD2 at the last position. If the TOD1 and
If the value is different from TOD2, it means that an input/output failure or a failure in the main body system occurred while writing the cell map, so the information in that cell map area is invalidated, and the other Information in the cell map area is used to restore the environment.

Similarly, real time information TOD1 and TOD2 are written in each cell 4 at the beginning and end positions. When read, TOD1 and TOD
If there is a discrepancy in the contents of 2, the checkpoint information for that cell is discarded. In the case of the present invention, the cells 4 do not have pointer information or the like and are independent from each other. An identifier ID indicating the type of checkpoint information is added to each cell 4 before the checkpoint information DATA. When a system abnormality occurs, valid cells 4 are extracted from the latest cell map in the cell map area 3-1 or 3-2, and for each identifier ID,
Based on the real-time information, the latest checkpoint information will be selected to restore the environment.

Next, an embodiment of the present invention will be described with reference to FIG. The journal requesting unit 10 updates or newly creates the contents of various management tables, control tables, etc. as various events occur.
When the environment changes and it becomes necessary to write information to the checkpoint data set 2, a request is made to the checkpoint processor 12 to write the checkpoint information. At this time, an identifier indicating the information type
ID will also be notified. If you need to synchronize,
It enters a waiting state until it receives a journal completion notification from the checkpoint processor 12.

The restoration processing section 11 operates, for example, when the system is restarted after a system failure, and restores the old environment based only on the contents of the checkpoint data set 2 explained in FIG. 2. First, for the checkpoint processor 12,
Request reading of all valid cells. Once the contents of the cell are read, according to the information identifier ID and the TOD value added before and after each information,
Sort/merge, extract the newest checkpoint information for each type of information, set it in a predetermined area of memory, and restore the old environment. Even if there is an input error in one cell information, due to the structure of checkpoint data set 2 explained in Fig. 2, it will not affect other cells and make them inaccessible. The information that should be restored by the previous generation is simply restored by the cell information from one generation ago.
Therefore, the range of influence of the failure can be localized.

The checkpoint processor 12 manages the checkpoint data set 2 and executes requests for input/output processing of checkpoint information from the journal request section 10 and the restoration processing section 11. When there is a write request from the journal request section 10, the output processing section 13 is activated. The output processing unit 13 calls the cell management unit 14 and requests cell selection in order to determine in which cell information should be written.

The cell management section 14 manages the cell map, cell address, etc. of the checkpoint data set 2 on the memory. The cell map indicating the validity/invalidity of each cell is always kept up-to-date in memory. If a cell needs to be reserved, a cell corresponding to bit value "0" on the cell map is found and that cell is allocated. When a cell is allocated, its bit value in the cell map is set to “1”.
Then, the real time information is written in the cell map area 3-1 or 3-2 shown in FIG. 2, whichever has the older information. That is, the latest cell maps are written alternately into cell map areas 3-1 and 3-2. When certain checkpoint information is managed for n generations, it has at least n cells in which that information is to be recorded. The cell management section 14 holds these cell addresses on the memory so that n cells are used cyclically, so that the checkpoint data set 2 can hold information of n generations. Note that the number of generations to be managed or not is determined based on the identifier ID of information registered in advance. In response to a cell selection request from the output processing unit 13, the address of the newly secured cell or the cell used for cycling is notified. According to the cell address, the output processing unit 13
Create a CCW (Channel Command Word) and add the actual time information to the checkpoint information.
TOD is added to request the input/output execution unit 16 to execute the input/output command. If an input/output failure occurs during execution of this input/output command, the cell management unit 14 is notified, a new cell is secured, and information is written to the cell. This retry process can be executed regardless of other cells.

The input processing section 15 reads checkpoint information of valid cells in response to a request from the restoration processing section 11. First, check the two cells provided in the fixed area of checkpoint data set 2.
Enter information for map areas 3-1 and 3-2,
Obtain the latest cell map by comparing TOD. Next, refer to the cell map to find a valid cell, create a READ CCW, and
The input/output execution unit 16 is requested to perform the process. The input/output execution unit 16 determines the access path of the channel 17 and the controller 18, issues an input/output command,
Request the channel to perform CCW.

Note that it is preferable that the cell management unit 14 determine the positions of the cells so that cells in which the dial information corresponding to the same information is written do not gather on the same track. This way, for example, due to a truck failure,
This further increases the probability that journal information can be restored even if all cells on that track become invalid.

In addition, cell map area 3, which is managed by two generations,
Even if reading cell maps from 1 and 3-2 fails on both sides, restoration is possible as follows. Read all cells, regardless of whether they are valid or invalid, and only those with the same TOD before and after are valid. Next, the information of each cell is classified by identifier ID and TOD. Based on the classification results, the checkpoint information is grouped, and a validity check is performed for each group to see if all the information that should be included in that group is included. The environment is restored using the one that passed the validity check and has the latest TOD, and once the processing for all groups is completed, preparations for restart are completed.

(E) Effects of the Invention As explained above, according to the present invention, a cell storing checkpoint information does not require information pointing to other cells, and can be used for generation determination and normal output determination. Since it has real-time information, even if the cell map area that manages the checkpoint data set itself or a part of the cell is destroyed, other checkpoint information can be safely used for restoration. Therefore, the recovery rate of the environment can be increased. Furthermore, the normal recording time can be shortened, and recovery processing for input/output failures and the like can also be simplified.

[Brief explanation of the drawing]

Figure 1 is an explanatory diagram of an example of the conventional method, Figure 2
The figure is an explanatory diagram of the structure of a checkpoint data set according to the present invention, and FIG. 3 shows the configuration of an embodiment of the present invention. In the figure, 2 is the checkpoint data set, 3
-1 and 3-2 are cell map areas, 4 is a cell, 13 is an output processing section, 14 is a cell management section, 15
represents an input processing section.

Claims (1)

[Claims] 1. In a checkpoint processing method that records the system environment of an operating data processing system in a checkpoint data set on an external storage medium, and restores the old environment according to the contents of the checkpoint data set in the event of a system error, the above The checkpoint data set has a plurality of cells that serve as input/output units of checkpoint information, and a cell map area that indicates whether each cell of the plurality of cells is valid or invalid, and the type of information to be checked. A cell management unit that manages the cell for each nth generation, and checkpoint information including current time information and an information identifier indicating the type of information for the cell notified by the cell management unit in response to a request from a journal requester. an input processing section that reads all valid cells according to the map information in the cell map area when restoring the environment, and the current time information and the information identifier of the valid cells read by the input processing section. A checkpoint processing method characterized by comprising: a restoration processing unit that refers to the latest checkpoint information for each type of information and restores the old environment.