JPH07168794A - Job managing method for computer system - Google Patents

Abstract

PURPOSE:To execute a job successively on a stopped computer and to restart a check point in a distributed system. CONSTITUTION:When a job managing program agent 6000 is stopped, the job managing information 4600 and job execution information 4700 of the job managing program agent 6000 are transmitted to a job managing program master 5000. Also, when the abnormal completion of the job occurs, the check point information of the job managing program agent 6000 is transmitted to the job managing program master 5000. The job managing program master 5000 selects another computer in the same group, and notifies it to the job managing program agent of the computer.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a job management method in a computer system, and more particularly to a job management method in a distributed environment computer system in which a plurality of computers are connected to each other via a network.

[0002]

2. Description of the Related Art Conventionally, as computer systems, there are a centralized system in which a single computer performs processing and a distributed system in which a large number of computers are connected to a network to perform distributed processing. In the centralized system, one computer holds all the management information for managing the system. That is, the management information on all the resources existing in the system is managed by the operating system (OS) on this computer. On the other hand, in a distributed system, each computer, which exists in large numbers in the system, has its own resource. The resources owned by each computer are managed by the OS on each computer.

Conventionally, a centralized system has a warm start function and a checkpoint restart function. The warm start function is a function that, when the system is started (or restarted), it takes over the state of the system at the time of the previous system stop, for example, the input job in the input queue and the output job in the output queue, and starts the processing. . On the other hand, the checkpoint restart function prepares for the re-execution of a job if the job ends abnormally in the process of executing one or more jobs, and the execution status of the job at a certain point This is a function of recording information related to the above and re-executing the process from the state at the time of checkpoint when the job ends abnormally. As a result, it is not necessary to re-execute the interrupted job from the beginning, and computer resources can be effectively used.

In the distributed system, as in the centralized system, each computer is provided with a warm start function or a checkpoint restart function so that each computer can start (restart) the system in the same manner as the centralized system. It is possible to continue the processing of the job that has started) or ended abnormally.

[0005]

However, in a distributed system, there is a plurality of computers in the system, and after the services of a certain computer are terminated, the other computers that are operating are stopped. It is desirable to be able to take over the work of. In addition, when a job that was being executed on one computer ends abnormally, it is desirable that this job can be re-executed on another computer.

In order to meet such a demand, when a computer is stopped due to a service termination or a failure, a job executed on the computer is executed on another computer operating in the system. It is necessary to enable continuous execution (such a function will be referred to as a job continuous execution function hereinafter). However, in the distributed system, for example, as described above, each computer has its own management information regarding files. Therefore, if you try to execute a job that was running on one computer on another computer, the computer that is trying to take over and execute the job will not use it on the previous computer to execute the job. There is a problem that the file cannot be recognized.
Therefore, in the conventional distributed computer system, it is difficult to realize the job continuation execution function between computers.

Accordingly, it is an object of the present invention to provide means for enabling a job continuation execution function and a checkpoint restart function in a distributed system.

[0008]

In order to achieve the above-mentioned object, in the job management method according to the present invention, a plurality of computers are connected to each other via a network, and each computer executes a job job to be executed by each computer. A job management program to be managed runs, and a part of the job management program (hereinafter, referred to as a job management program master) executes a job management program on a computer under the job management program (hereinafter, a job management program agent). In the computer system that manages the management information in (1), computers that execute jobs are grouped by the job management program master, and information regarding the grouping is transmitted to each of the job management program agents. When the job management program agent stops, the job management program agent that stops notifies the job management program master that it will stop, and retains the execution information and the execution information of the job being executed on the computer on which it runs. Send job management information to the job management program master. In the job management program master, from the other computers belonging to the same group as the computer on which the job management program agent to be stopped was running, the jobs that were running on the computer on which the job management program agent to be stopped were running The computer to be executed is selected, and the job execution information and job management information received from the job management program agent to be stopped are sent to the job management program agent on the selected computer. And
The job management program agent on the selected computer executes the job on the computer on which the stopped job management program agent was running.

In such a job management method for a computer system, the job management program agent executes job management on the selected computer before executing the job on the computer on which the agent has stopped. The program agent creates, on the selected computer, a replica of the file used by the job that was running on the computer on which the job management program agent stopped.

Alternatively, the job management program agent on the selected computer stops the job management program agent before the job that was being executed on the computer on which the job management program agent has stopped is executed on the selected computer. Make the files on the computer visible to the selected computer.

According to another aspect of the present invention, a plurality of computers are connected to each other via a network, and a job management program that manages jobs to be executed by each computer runs on each computer, and the job management program is executed. A computer system that manages management information in a job management program (hereinafter, referred to as a job management program agent) on a computer under the job management program (hereinafter, referred to as a job management program master) In the job management method, the job management program master groups computers that execute jobs and sends information about the grouping to each of the job management program agents. At the checkpoint, the job manager agent
The resource status information related to the job and the job management information held by the self (hereinafter referred to as checkpoint information) are transmitted to the job management program master. The job management program master stores the checkpoint information sent from the job management agent. Then, when an abnormal end of a job occurs, the job management program agent running on the computer in which the abnormal end occurred notifies the job management program master of the occurrence of the abnormal end. The job management program master selects the computer that will re-execute the job from the other computers that belong to the same group as the computer that has abnormally terminated, and the checkpoint information for the computer that has abnormally terminated will be displayed on the selected computer. The job management program agent on the selected computer re-executes the job on the computer in which the abnormal termination of the job has occurred.

In such a job management method for a computer system, the job management program agent on the selected computer uses the abnormally terminated job before re-executing the abnormally terminated job on the selected computer. Create a replica of the file you were using on the computer of your choice. Alternatively, the job management program agent on the selected computer is using the abnormally ended job on the computer on which the abnormally ended job was running before re-executing the abnormally ended job on the selected computer. Makes the selected file visible to the selected computer.

[0013]

According to the present invention, by the means described above, the computers on which the job management program master operates are grouped, and the information on the grouping of computers included in the job management program master is transferred to the job management program agent on each computer. As a result, each job management program agent can execute job scheduling while being aware of a group of computers. When the job management program agent stops, the job management program master is notified of this together with the job management information and the resource status information related to the job being executed, so that the status of the computer to be stopped before the job management is stopped. The program master can know. Then, the job management program master selects another computer that belongs to the same group as the stopped computer of the job management program agent based on the information about the grouping of computers that it owns, and immediately before stopping it. The job management program agent's job management information and the job management program agent's resource status information about the jobs being executed on the computer to be stopped are sent to the selected computer. The job that has been executed on the computer in which the job management program agent has stopped can be executed on another computer belonging to.

Further, before executing the job, which was being executed on the computer on which the job management program agent has stopped, on the selected computer, the job management program agent on the selected computer has stopped the job management program agent. A replica of the file used by the job that was running on the computer is created on the selected computer, so you can use the file that is required for the operation of the job that will be inherited from the selected computer. Can be executed.

Alternatively, the job management program agent on the selected computer stops the job management program agent before executing the job on the computer on which the job management program agent has stopped on the selected computer. Since the files on the computer can be recognized by the selected computer, the files required for the job operation can also be recognized by the selected computer, and this computer also needs the files for the job operation. The file can be used and the job can be executed.

[0016]

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

FIG. 1 is a block diagram showing the hardware configuration of a computer system according to this embodiment. Reference numeral 1000 denotes a computer, and a plurality of computers 1000 are respectively connected by a local area network (LAN) 9000 to form a network. In FIG. 1, the computer 100
Although 0 indicates only four computers, the number of computers 1000 is arbitrary. LAN9000 is a gateway 970
Wide Area Network (WAN) 960 through 0
Connected to 0.

FIG. 2 is a block diagram showing the detailed arrangement of the computer 1000. As shown in FIG.
0 is a central processing unit (CPU) 1100, an external storage device 2000, an I / O adapter 2500, a memory 3000,
It is composed of a bus 8000 and a LAN adapter 9500. The CPU 1100, the I / O adapter 2500, the memory 3000, and the LAN adapter 9500 are each connected to the bus 8000, and exchange data via the bus 8000. The external storage device 2000 is connected to the I / O adapter 2500. Generally, a disk device is often used as the external storage device 2000. The external storage device 2000 stores control software 4000 and application software (AP) 3500 for controlling this computer system. The LAN adapter 9500 is connected to the LAN 9000 and controls data transmission with another computer 1000.

FIG. 3 shows the configuration of the control software 4000. The control software 4000 is stored in the external storage devices 2000 of all computers and loaded into the memory 3000 as needed. The CPU 1100 is the control software 400 loaded on the memory 3000.
Various controls are performed according to 0. Control software 40
00 includes an operating system 4100, a distributed file management program 4200, and a job management program 4500. The operating system 4100 has resources (memory 30
00, external storage device 2000, etc.) and the like. The distributed file management program 4200 is a program that uniformly manages files among computers in an environment in which a plurality of computers are connected by a network like this computer system. This distributed file management program 420
0 is, for example, a distributed computing environment (D
There is a Distributed File Service (DFS) in CE). The job management program 4500 is a program that manages jobs (jobs) to be executed by the computer 1000 by the computer user, and performs job scheduling, queuing, monitoring the job execution state, and the like. There are two types of job management programs 4500: a job management program master 5000 and a job management program agent 6000. A job management program (job management program master 5000) on a computer manages the job management program (job management program agent 6000) of the computers under it.

FIG. 4 shows a job management program master 50.
00 and job management program agent 6000
Indicates the main information held by. The job management program agent 6000 has job management information A4600 and job execution information A4700. Job management information A460
0 is the information of the job being executed (or scheduled to be executed) on the computer, which is the queuing information A4610 of the job executed on the computer, the execution status of the job on the computer (waiting for execution, executing) , Execution end, etc.) and monitoring information A 4620. The job execution information A4700 stores execution information related to the job being executed by the computer, which is collected from the OS 4100 and the like. The job management program master 5000 has job management information M4601, queuing information M4611, monitoring information M4621, and job execution information M4701, which are used when the job management program master 5000 operates as a job management program agent. In addition, job management information A4600 and job execution information A4700 notified from the job management program agent 6000.
And an agent job management information storage unit 4800 for storing checkpoint information, an agent job execution information storage unit 4830, and an agent checkpoint information storage unit 4840. These storage units (agent job management information storage unit 4800, agent job execution information storage unit 4830, agent checkpoint information storage unit 4840) exist for each job management program agent 6000.

FIG. 5 shows a job management program master 50.
12 is a flowchart showing an outline of processing when 00 is started. The job management program master 5000 is activated upon the input of the activation command (step 51).
00). Next, the job execution computer group definition file 7010 is read (step 5110).

FIG. 6 shows the contents of the job execution computer group definition file 7010. The job execution computer group of each computer is defined by a pair of the host name of the computer and the group name of the job execution computer group to which the computer belongs. FIG. 6 shows that computers having host names HOST1 to HOST4 belong to the job execution computer group GROUP1 and computers having host names HOST5 and HOST6 belong to the job execution computer group GROUP2.

FIG. 7 is a flow chart showing an outline of processing at the time of starting the job management program agent 6000. Job management program agent 6000
Is activated upon the input of the activation command (step 6100). Next, the job management program master 5000 is requested to send the job execution computer group definition file 7010 (step 611).
0). The job management program master 5000 includes a job management information storage unit 4800 and a job execution information storage unit 4
830 is created (step 6115), and the job execution computer group definition file 7010 is sent to the job management program agent 6000 (step 612).
0). Next, the job management program agent 600
0 stores the received job execution computer group definition file 7010 in its own memory 3000 (or external storage device 2000) (step 6130).

By the processing shown in FIGS. 5 and 7, each job management program agent 6000 can recognize the defined job execution computer group. However, the job management program master 5000 must be activated before the job management program agent 6000.

FIG. 8 is a flow chart showing an outline of processing when the job management program agent 6000 is stopped. The stop processing of the job management program agent 6000 is started upon the input of the stop command (step 6200). If there is a job being executed on the computer, the execution of the job is interrupted, and the execution information of the job is stored in the job execution information A4700 (step 6205). Then, the job management program agent 6000 sends the job management information A4600 and the job execution information A4700 to the job management program master 5000 to stop the job management program agent 6000.
00 (step 6210). At this time, the job management program agent 6000 adds the host name of the computer together with the job management program master 5000.
To notify.

The job management program master 5000 is
The received job management information A4600 and job execution information A4700 are stored in the agent job management information storage unit 48.
00 and agent job execution information storage unit 4830
(Step 6235). Then, based on the information of the job execution computer group definition file 7010 stored in step 6130 in FIG. 7 described above, one computer that belongs to the same job execution computer group as the computer in which the job management program agent 6000 has stopped A selection is made (step 6240). The job management program master 5000 sends the job management information in the agent job management information storage unit 4800 and the job execution information in the agent job execution information storage unit 4830 to the job management program agent 6000 on the computer selected in step 6240. (Step 6260).

The job management program agent 6000 on the computer selected in step 6240 merges the received job management information and job execution information with its own job management information A4600 and job execution information A4700 (step 6270). And step 6
The job management program agent 6000 makes it possible to recognize the file of the job on the computer stopped by the computer selected at 240 (step 6280). As the processing method performed in step 6280, for example, the following two methods can be considered.

First, as the first method, the job management program agent 6000 is stopped using the function of creating a replica of a file provided by the distributed file management program 4200 from the job management information 4600 and the job execution information 4700. There is a method of creating a replica of a job file on a computer on the computer. As a second method, the job management program agent 6 on the computer selected in step 6240
000 to refer to the job management information 4600 and the job execution information 4700, and use the remote mount function of the network file system to mount the file of the job on the computer stopped by the job management program agent 6000 on the computer. There is.

In this way, the job that has been executed on the computer that is going to stop the job management program agent 6000 can be continuously executed on the computers that belong to the same job execution computer group as the computer. .

FIG. 9 is a flow chart showing the outline of the processing when the job execution reaches the checkpoint. The job management program agent 6000 uses the job execution status monitor function of the job management program agent 6000 to recognize that the job reaches the checkpoint (step 7100). When it is recognized that the checkpoint has been reached, the job management information A4600 and the job execution information A4700 are sent to the job management program master 5000 as checkpoint information to notify that the checkpoint has been reached (step 7110). The job management program master 5000 stores the checkpoint information sent from the job management program agent 6000 in the agent checkpoint information storage unit 4840 (step 7120).

FIG. 10 is a flow chart showing an outline of processing when an abnormal end of a job occurs. The job management program agent 6000 recognizes that the abnormal end of the job has occurred based on the monitor information A4620 of the job management information A4600 (step 720).
0). Then, the job management program master 5000
Is notified of the abnormal end of the job together with the host name of the computer (step 7210). The job management program master 5000 is a computer belonging to the same job execution computer group as the computer in which the abnormal end of the job has occurred, based on the information of the job execution computer group definition file 7010 acquired in step 6130 in FIG. 7 described above. One (step 72)
30). The job management program master 5000 transmits the information stored in the agent checkpoint information storage unit 4840 to the job management program agent 6000 on the computer selected in step 7230 (step 7240).

The job management program agent 6000 on the computer selected in step 7230 merges the sent agent checkpoint information with its own job management information 4600 and job execution information 4700 (step 7250). Then, the file of the job on the computer in which the abnormal end of the job has occurred can be recognized (step 7260). In step 7260,
By the same method as step 6280 described above in the processing when the job management program agent is stopped,
Allows the computer that inherits the job to recognize the job file on the computer where the job ended abnormally.

By the above processing, it becomes possible to re-execute a job on a computer in which an abnormal job termination has occurred on a computer belonging to the same job execution computer group as this computer.

According to the above-described embodiment, the job executed on the computer stopped due to the end of the service or the occurrence of a failure in the distributed computer system is taken over by another computer in the system. Can be run with. Further, the computers in the system are divided into groups, and the jobs in the stopped computers are taken over by the computers in the same group, so that the processing can be taken over without making a great change in the execution environment or the like.

[0035]

As described above, according to the present invention,
In a distributed system, a job on a stopped computer can be taken over and executed on another computer in the system, and continuous execution of jobs among computers and checkpoint restart can be performed.

[Brief description of drawings]

FIG. 1 is a hardware configuration diagram of a computer system that implements a job management method for a computer system according to the present invention.

2 is a block diagram showing a detailed configuration of a computer 1000. FIG.

FIG. 3 is a configuration diagram showing a logical configuration of control software.

FIG. 4 is a configuration diagram showing main information held by a job management program master 5000 and a job management program agent 6000.

FIG. 5 is a flowchart showing an outline of processing when the job management program master 5000 is activated.

FIG. 6 is a job execution computer group definition file 701.
It is a file block diagram which shows the content of 0.

FIG. 7 is a flowchart showing an outline of processing when the job management program agent 6000 is activated.

FIG. 8 is a flowchart showing an outline of processing when the job management program agent 6000 is stopped.

FIG. 9 is a flowchart showing an outline of processing when the execution of a job reaches a checkpoint.

FIG. 10 is a flowchart illustrating an outline of processing when a job ends abnormally.

[Explanation of symbols]

1000: computer, 1100: central processing unit (CP
U), 2000: external storage device, 2500: I / O adapter, 3000: memory, 3500: application program (AP), 4000: control software, 4
100: operating system (OS), 420
0: distributed file management program, 4500: job management program, 4600: job management information A, 470
0: job execution information A, 4601: job management information M,
4701: job execution information M, 4800: agent job management information storage unit, 4830: agent job execution information storage unit, 4840: agent checkpoint information storage unit, 5000: job management program master, 6000: job management program agent, 7
010: Job execution computer group definition file, 80
00: bus, 9000: local area network (LAN), 9500: LAN adapter, 9600: wide area network (WAN), 9700: gateway.

Front page continuation (72) Inventor Masaaki Oya 5030 Totsuka-cho, Totsuka-ku, Yokohama, Kanagawa Prefecture Hitachi Ltd. Software Development Division

Claims (6)

[Claims] 1. A plurality of computers are connected to each other via a network, and a job management program for managing a job job to be executed by each computer runs on each computer, and a part of the job management program in the job management program is managed. A program (hereinafter referred to as a job management program master) is managed by a job management program on a computer under the
In the job management method of a computer system for managing management information in a job management program agent), computers that execute jobs are grouped by the job management program master, and information about the grouping is grouped in each of the job management program agents. A step of transmitting, a step of stopping the job management program agent and the job management program master, and a case where the job management program agent stops, the job management program agent to be stopped,
Notifying the job management program master that it will stop, and sending to the job management program master the execution information of the job being executed on the computer on which it runs and the job management information held by itself. The job management program master is being executed on the computer on which the job management program agent to be stopped was running from another computer belonging to the same group as the computer on which the job management program agent to be stopped was running. Selecting a computer to execute the selected job, and transmitting the job execution information and job management information received from the job management program agent to be stopped to the job management program agent on the selected computer; and the selected computer The job management program above By Gram agent, job management method of a computer system, characterized in that a step of the job management program agent executes the job on a computer that was running the said stop. 2. The job management method for a computer system according to claim 1, wherein the job being executed on the computer on which the job management program agent to be stopped is running is the computer on which the job management program agent is stopped. A replica of the file used by the job running on the computer stopped by the job management program agent before the job management program agent on the selected computer executed it on another computer belonging to the same group. A method for managing a job in a computer system, the method including: creating a job on a computer on which the user runs. 3. The computer system job management method according to claim 1, wherein the job executed on the computer in which the job management program agent has stopped is executed before the execution on the selected computer. A job management method for a computer system, comprising the step of allowing a job management program agent on the computer to recognize a file on the computer stopped by the job management program agent from a computer on which the computer runs. 4. A plurality of computers are connected to each other via a network, and a job management program that manages jobs to be executed by each computer runs on each computer, and a part of the job management programs in the job management program. In a job management method of a computer system, (hereinafter, referred to as a job management program master) managing management information in a job management program (hereinafter referred to as a job management program agent) on a computer under the Grouping computers that execute jobs and transmitting information about the grouping to each of the job management program agents; and the job management program agent, at a checkpoint, information on the status of resources related to the jobs. And transmitting job management information (hereinafter, referred to as checkpoint information) held by itself to the job management program master, the job management program master stores the checkpoint information sent from the job management agent. A step of storing, when a job abnormal termination occurs, a job management program agent running on the computer in which the abnormal termination has occurred notifies the job management program master of the occurrence of the abnormal termination; The management program master selects a computer that re-executes the job from other computers belonging to the same group as the computer in which the abnormal termination has occurred, and the checkpoint information of the computer in which the abnormal termination has occurred is the selected computer. Top Job Manager Program Agent And a step of causing the selected job management program agent on the computer to re-execute the job on the computer in which the abnormal termination of the job has occurred. 5. The computer system job management method according to claim 4, wherein a job management program agent operating on the selected computer before the abnormally terminated job is re-executed on the selected computer. A method of managing a job in a computer system, further comprising: creating a replica of a file used by a job operating on the computer in which the abnormal termination has occurred on the selected computer. 6. The job management method for a computer system according to claim 4, wherein a job management program agent operating on the selected computer before re-execution of the abnormally terminated job on the selected computer. A method for managing a job in a computer system, further comprising the step of allowing a file used by the abnormally ended job on the abnormally ended computer to be recognized by the selected computer.