JP2735514B2 - Process status management method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a process state management method, and more particularly to a responsive, simple, and low CPU load state management method for monitoring a plurality of resident processes operating in parallel.

[0002]

2. Description of the Related Art In a recent information processing system such as a general-purpose computer, a frequently used common process is constantly operated as a resident process in order to efficiently process a plurality of user programs in parallel. Is adopted. In this case, if the running resident process terminates abnormally for some reason, it must be restarted and run immediately.

In general, the operating state of a process is managed by the OS of the system, and has a function of grasping the state of all running processes and abnormally terminating the process when an abnormality is detected. However, when an abnormal termination occurs in a certain process, another process that understands the contents of the process and executes control must be executed in order to restart the process or perform necessary processing for related processes. is necessary. Representative conventional methods for performing such control are disclosed in Japanese Patent Application Laid-Open Nos.
-206437.

A first method described in Japanese Patent Application Laid-Open No. 5-342008 discloses a method of providing a parent process that starts a plurality of related processes as child processes, and using a parent-child relationship to execute a child process abnormally terminated. Information (process ID, etc.) is received from the OS to perform necessary control. To do this, the parent process must be running, so start another monitoring process as a child process to monitor the parent process, and confirm that the parent process is running. System calls to the OS are used periodically.

A second method described in Japanese Patent Application Laid-Open No. 1-206437 directly manages the states of a plurality of child processes to be monitored by a parent process by providing a process management table. And process transfer between modules is performed via the dispatcher, and the process management table is updated each time, and the process management module of the parent process periodically checks the process management table by a timer interrupt and identifies abnormal child processes. It is configured to detect.

[0006]

In each of the above-mentioned conventional process state management systems, the parent process is always running to monitor the abnormality of the activated child process, and consumes CPU time. That is, JP-A-5-342058
In the first method described in Japanese Patent Application Laid-Open Publication No. H10-207, a parent process needs to be running in order to receive and recognize the process ID of a child process that has terminated abnormally, which imposes a load on the CPU and further monitors the parent process. Monitoring process periodically
A load is being applied to the PU. Also, Japanese Patent Application Laid-Open No. 1-206437
The second method described in the above publication has a drawback that not only does the CPU load a process for updating and checking the process management table, but also a time difference occurs until an abnormality is detected due to a periodic check by a timer interrupt.

It is an object of the present invention to efficiently manage a plurality of resident processes without imposing a heavy load on a CPU with a simple configuration, and to immediately restart even if an abnormally terminated process occurs, thereby improving reliability. The object of the present invention is to provide a process state management method capable of improving the process status.

[0008]

According to a first aspect of the present invention, there is provided a process state management system for managing operation states of a plurality of resident processes operating in parallel,
A management process as a parent process that starts the plurality of resident processes as child processes, and a counting semaphore having an initial setting value equal to the number of resident processes started by the management process are provided. A process start notifying means for locking the counting semaphore at the time of activation and designating a notification destination at the time of abnormal termination as an unlock signal of the counting semaphore, wherein the management process sequentially starts the plurality of resident processes, A process starting unit that attempts to lock the counting semaphore and puts the management process in a lock waiting state, and stores a process ID of the started resident process into a process ID in a memory.
Process ID management means for registering and managing in a management table, process monitoring means for automatically recognizing the occurrence of abnormal termination of the resident process when lock of the management process in the lock waiting state succeeds, and the process monitoring means A process for sequentially transmitting a signal for confirmation to the process registered in the process ID management table and starting the process which has been abnormally terminated by utilizing the fact that the signal transmission to the process which has abnormally terminated cannot be performed normally. And a faulty process extracting means for detecting the faulty process.

According to a second aspect of the present invention, there is provided a process status management system according to the first aspect, wherein a command is issued instead of the faulty process extraction means to obtain information on a running process from an OS. I
An operating process status reference unit for detecting a process that has terminated abnormally by checking the correspondence with the process ID on the D management table is provided.

According to a third aspect of the present invention, in the process state management method according to the first or second aspect, the process executed by the process activation notifying means provided in the plurality of resident processes is executed at the time of the activation. The method is characterized in that the process starting means of the process is configured to execute instead of immediately before starting the resident process.

[0011]

Next, embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of the first embodiment of the present invention.

As shown in FIG. 1, the process state management method according to the present embodiment comprises a plurality of resident processes P1 to Pn operating in parallel, and a management process for monitoring and controlling the child processes P1 to Pn. Parent process 1a
And a shared memory 2 for registering the process ID management table 21, and a counting type semaphore 3 whose initial value of the semaphore value is the same as the number n of child processes. The parent process 1a and the child processes P1 to Pn are stored in the control unit 4 of the system in the shared memory 2 and the semaphore 3
Are included in the storage unit 5. The parent process 1a attempts to lock the semaphore 3 after sequentially activating the child processes P1 to Pn, and places the process monitoring unit 13 in a lock waiting state. A process ID management means 12 for registering and managing the process ID in the process ID management table 21, a process monitoring means 13 for automatically recognizing the occurrence of abnormal termination of a child process when locking is successful, and a process monitoring means 13 activated by the process monitoring means 13. Process ID management table 21
And a failure process extraction unit 14 for sequentially detecting a process which has been abnormally terminated by transmitting a signal for confirmation to the processes registered in the subprocesses P1 to Pn.
Each is provided with process start notifying means PP1 to PPn for locking the semaphore 3 at the time of startup and designating a notification destination at the time of abnormal termination as a lock release signal of the semaphore 3.

The parent process 1a is provided with a process
To start the child processes P1 to Pn (see FIG. 1), and register the process IDs of the child processes P1 to Pn started by the process ID management means 12 in the process ID management table 21 on the shared memory 2 (see FIG. 1). reference). The activated child processes P1 to Pn sequentially request the semaphore 3 for the semaphore lock by the process activation notifying means PP1 to PPn one by one (see FIG. 1). When the semaphore 3 succeeds in locking, the semaphore value is reduced by one. Then, the child processes P1 to Pn
Are all successfully semaphore locked, the semaphore value is zero. Next, the process monitoring means 13 of the parent process 1a attempts semaphore lock on the semaphore 3 (see FIG. 1), but the semaphore value is already 0, so a semaphore lock wait state occurs. This state is a normal state in which all started resident processes are operating normally.

Next, a case where one of the active child processes P1 to Pn (i = 1 to n) abnormally ends will be described. When the child process Pi ends abnormally,
The information is sent to the semaphore 3 of the notification destination specified by the process activation notifying means PPi at the time of startup, and the semaphore lock performed on the semaphore 3 is invalidated, and the semaphore value of the semaphore 3 is changed from 0 to 1. Change. Therefore, the parent process 1a in the semaphore lock waiting state succeeds in the semaphore lock, and the process monitoring unit 13 of the parent process 1a recognizes that one of the child processes P1 to Pn has terminated abnormally. The process monitoring unit 13 that has recognized the abnormal termination of the child process notifies the failed process extraction unit 14 (see FIG. 1), and the failed process extraction unit 14 starts the process of detecting the process ID of the abnormally terminated child process. (See FIG. 1).

The details of the processing from detection of a child process that has terminated abnormally to restart will be described below with reference to FIG. The faulty process extraction means 14 of the parent process 1a requests the process ID management means 12 to identify which child process among the child processes P1 to Pn has abnormally terminated. The process IDs of the child processes P1 to Pn are acquired from the sub-process 21 (see FIG. 3), and signals for confirming that the child processes P1 to Pn are sequentially operating are transmitted (see FIG. 3). Child processes P1, P2, and P not abnormally terminated
n, the signal transmission to the abnormally terminated child process P3 fails and no confirmation is obtained.
3 is detected (see FIG. 3). The failed process extraction unit 14 that has detected the abnormally terminated child process requests the process activation unit 11 to restart the child process P3, and the process activation unit 11 restarts the child process P3 and restarts the child process P3 ′. (See FIG. 3). The process activation notifying means of the child process P3 'performs semaphore lock on the semaphore 3 (see FIG. 3), and the semaphore 3 updates the semaphore value from 1 to 0.
The process initiating means 11 notifies the process ID managing means 12 of the process ID of the child process P3 '.
The D management means 12 updates the process ID of the child process P3 in the process ID management table 21 of the shared memory 2 to the process ID of the child process P3 '(see FIG. 3). The process starting means 11 which has finished restarting the child process,
It notifies the process monitoring means 13 of the completion of the restart. The process monitoring means 13 requests the semaphore 3 to perform a semaphore lock, but since the semaphore value has already been set to 0, the process monitor returns to the semaphore lock wait state (see FIG. 3), that is, returns to the normal state and resumes the process monitoring.

As described in detail above, the process state management method of the present embodiment employs a plurality of resident child processes P1.
To a common semaphore 3 whose initial setting value is the number n of child processes,
Pn is shared, and the process monitoring unit 13 is set to a semaphore lock wait state to monitor a child process. Therefore, the parent process 1a is in a waiting state for monitoring and does not impose an extra load on the CPU.
When an abnormal termination occurs in any of the child processes, the operation can be immediately performed and the abnormality can be detected by taking over the CPU time of the abnormally terminated child process. When the occurrence of the abnormal end is detected, the faulty process extracting unit 14 refers to the process ID management table 21 and registers the registered n.
It is possible to identify the abnormally terminated child process by transmitting a confirmation signal only to the child processes P1 to Pn, and to immediately restart the child process. That is, a load is not applied to the CPU for monitoring, and if an abnormal end occurs, the CPU can be operated without delay to perform necessary processing.

FIG. 2 is a block diagram showing the configuration of the second embodiment of the present invention. The difference between the second embodiment shown in FIG. 2 and the first embodiment shown in FIG. 1 is that the parent process 1b is different from the first embodiment in that the operation process It is provided with the state reference means 15. Other configurations are the same as those of the first embodiment, and are indicated by the same reference numerals as in FIG. A process in which the parent process 1b sequentially activates the child processes P1 to Pn and enters a semaphore lock waiting state, which is a normal state;
And abnormal termination occurs in the child process and the process monitoring means 1
The process until 3 detects this is the same as in the first embodiment. Since there is a difference in a detection method for specifying a child process that has terminated abnormally when an abnormal termination has occurred, this point will be described below.

When one of the child processes P1 to Pn terminates abnormally, the semaphore lock that has been performed on the semaphore 3 is invalidated, the semaphore value changes from 0 to 1, and the semaphore lock waits. The process monitoring unit 13 in the state detects the abnormal end and notifies the operating process state reference unit 15 (see FIG. 2). The running process status reference means 15 acquires the process IDs of the child processes P1 to Pn from the process ID management means 12 (see FIG. 2), and sends a command to send the process ID of the currently running child process from the system OS. (See FIG. 2), and the abnormally terminated child process is detected by taking the difference between the two. The process after the detected child process is restarted by the process starting unit 11 is the first process.
This is the same as the embodiment. As a method of detecting the abnormally terminated child process of the operating process status referring means 15, in addition to directly obtaining the process ID from the OS and directly comparing it, an O / O response indicating whether or not the process is active is indicated by O.
It may be obtained from S.

As described above, according to the second embodiment, when the process monitoring means 13 detects that one of the child processes has terminated abnormally, and identifies the abnormally terminated child process, the operating process state reference means 15 Uses operating information from the OS. On the other hand, in the first embodiment, the faulty process extraction unit 14 sends a signal to each child process to confirm that it is operating sequentially, and terminates abnormally without relying on the OS. Identify child processes.

In the above description of the first and second embodiments, each child process started by the process start unit of the parent process locks the semaphore by the process start notifying unit immediately after starting, and succeeds in locking. Although the original process is started after each process, the process of locking the semaphore may be performed by the process starting means of the parent process that starts the child process.
Further, the process ID management table is created on the shared memory of the storage unit, but is not limited to the shared memory.

[0022]

As described above, according to the process state management method of the present invention, when a management process manages a plurality of resident processes, the semaphore is used to monitor the resident processes in a semaphore lock wait state. And no CPU time is consumed for monitoring.
Further, when an abnormal termination occurs in the resident process, the operation is automatically started, and the abnormally terminated resident process can be specified and restarted. In other words, the process state can be monitored without imposing a load on the CPU, and the abnormal termination can be dealt with immediately, thereby effectively managing the resident process.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of a first exemplary embodiment of the present invention.

FIG. 2 is a block diagram illustrating a configuration of a second exemplary embodiment of the present invention.

FIG. 3 is an explanatory diagram of a process from detection of an abnormal end process to restart in the first embodiment of the present invention.

[Explanation of symbols]

 1a, 1b Parent process 2 Shared memory 3 Semaphore 4 Control unit 5 Storage unit 11 Process start-up unit 12 Process ID management unit 13 Process monitoring unit 14 Faulty process extraction unit 15 Active process state reference unit 21 Process ID management table P1, P2, P3 ... Pn child process P3 ′ Child process after restart PP1, PP2.

Continuation of the front page (72) Inventor Kaori Nishikawa 1-18-6 Shinkiba, Koto-ku, Tokyo Japan Electric Software Co., Ltd. (56) References JP-A-7-200347 (JP, A) JP-A-7-200 141229 (JP, A) JP-A-5-334121 (JP, A) JP-A-5-46411 (JP, A) JP-A-63-65535 (JP, A)

Claims (3)

(57) [Claims] 1. A process state management method for managing the operation states of a plurality of resident processes operating in parallel, comprising: a management process as a parent process for starting the plurality of resident processes as a child process; A counting semaphore having a value equal to the number of resident processes started by the management process is provided, and the plurality of resident processes respectively lock the counting semaphore at startup and unlock the counting semaphore at the time of abnormal termination. A process start notifying unit for designating a signal as a signal, wherein the management process attempts to lock the counting semaphore after sequentially starting the plurality of resident processes, and sets the management process to a lock waiting state; and , The process ID of the started resident process in the process
Process ID management means for registering and managing the management process in a D management table; process monitoring means for automatically recognizing the occurrence of abnormal termination of the resident process when the lock of the management process in the lock waiting state has succeeded; A process that is started by the means and sends a signal for confirmation sequentially to the processes registered in the process ID management table, and a process that ends abnormally using the fact that signal transmission cannot be performed normally to a process that ends abnormally And a faulty process extracting means for detecting a faulty process. 2. In place of the faulty process extracting means,
Operating process status reference means for detecting a process that has terminated abnormally by issuing a command to acquire information on the running process from the OS and checking the correspondence with the process ID on the process ID management table. The process state management method according to claim 1, wherein 3. The process startup notifying means of the plurality of resident processes is configured to execute at the time of startup, instead of the process startup means of the management process immediately before the startup of the resident process. 3. The process state management method according to claim 1, wherein