JPS62143141A - Detecting system for down of processor - Google Patents

Abstract

PURPOSE:To detect the down of a processor without subjecting heavy load to a CPU at the time of normal operation and without making a bus busy by providing the system with an auto-repute information control device. CONSTITUTION:Respective processors 11-14 are connected to the auto-repute information control device 2 and a common storage device 3, and when a fault is generated in a self-processor, the processor can execute automatic rerising on the basis of the auto-repute function. When auto-repute start information or auto-repute end information is sent from the processors 12-14, the device 2 sends the processor number and the information to the master processor 12. When the information is sent from the master processor 11, the information is sent to the master-substitutive processor 12. The processor 11 or 12 receiving the auto-repute start information starts time monitoring, and if receiving the auto-repute end information within a fixed time, regards the processor concerned as a downed processor and informs the processor number of the downed processor to all the other processors through inter-processor communication.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a processor down detection method, and more particularly to a Kel processor down detection method for a loosely coupled multiprocessor system.

(Prior Art) Conventionally, in order to detect when a processor is down in a loosely coupled multiprocessor system, a watchdog timer is provided in a device shared by each processor or in a specific processor, and the watchdog timer is periodically transmitted from each processor. A notification needs to be issued so that it can be reset.

(Problems to be Solved by the Invention) The conventional processor down detection method described above has the drawback of placing a heavy burden on the CPU because notifications are issued periodically. Another drawback is that the bus becomes busy periodically due to the issuance of the above notifications.

The object of the present invention is to have a master processor, a master replacement processor, and a plurality of slave processors, and an auto-reput function (if a failure occurs during operation of the computer system and it becomes impossible to continue, the system automatically restarts the 7-system system). ), each processor notifies an autoreput notification control device of the start and end of autoreput. Receives notifications of the start and end of autoreput, and if the above notification comes from a source other than the master processor, sends the processor identification number and the above notification to the master processor, and if the above notification comes from the master processor, sends the above notification to the master processor. The above drawback is removed by sending the notification to the master alternative processor, and if the processor that receives the notification of the start of autoreput does not receive the notification of the end of autoreput within a certain period of time, it sends the notification of the start of IJ put. It is an object of the present invention to provide a processor down detection method configured so that a processor that has failed can be regarded as having gone down.

(Means for Solving the Problems) A processor down detection method according to the present invention is configured to include a master processor, a master replacement processor, a plurality of slave processors, and an autoreput notification control device.

The master processor, master replacement processor, and multiple slave processors are each connected to a shared storage device and are loosely coupled, and can notify the start/end of autoreput, and can also notify the start/end of autoreput in the event of a failure. It has an IJ boot function for automatic restart.

The autoreput notification control device receives a notification of the start/end of autoreput, and if the notification arrives from a master replacement processor or a slave processor, sends the notification and processor identification number to the master processor,
When the notification arrives from the master processor, it is sent to the master replacement processor to control processing.

(Example) Next, the present invention will be explained with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of the processor down detection method according to the present invention. In FIG. 1, 11 is a master processor, 12 is a master replacement processor, 13 is a slave processor, 14 is a slave processor, 2 is an autoreput notification control device, and 3 is a shared storage device.

In FIG. 1, processors 11-14 are connected to autoreput notification control device 2 and shared storage device #3, respectively. When a failure occurs within the processors 11-14, the processors 11-14 can be automatically restarted using an autoreput function.

At this time, each processor sends a notification of the start of autoreputation to the autoleague notification control device at the start of autoreputation, and sends a notification of the end of autoreputation to the autoreputation notification control device 2 at the end of autoreputation. When a notification of the start of autoreput or a notification of the end of autoreput is sent from the processors 12 to 14, the autoreput notification control device 2 sends the processor number and the above notification to the master processor 11. On the other hand, when the above notification is sent from the master processor 11, it is sent to the master substitute processor 12.

The master processor IL or the master substitute processor 12, which has received the notification of the start of autoreputation, starts time monitoring and waits for the notification of the end of autoreputation. If the notification of the end of autoreput does not arrive within a certain period of time, the processor that sent the notification of the start of autoreput will fail during autoreput, and will be unable to continue IJ boot and go down. shall be deemed to have been done.

The master processor 11 or master substitute processor 12 that has detected that a processor is down notifies all other processors of the number of the down processor through inter-processor communication. If the processor that has gone down is the master processor 11, the master replacement processor 12 takes over as the master processor and executes subsequent processes such as monitoring autoreput time in place of the master processor 11.

Here, the processors 11-14 sequentially write the operating status of their own processors to the shared storage device 3, so if a certain processor goes down, a previously designated processor takes over the processing of the down processor. This handover is performed when the number of the down processor is notified.

(Effects of the Invention) As explained above, according to the present invention, the master processor,
In a loosely coupled multi-grocer 7 system that has a master replacement processor and multiple slave processors and has an autoregate function, each processor sends autoreput start and end notifications to the autoreput notification control device, and performs autoreput notification control. The device receives notifications of the start and end of autoreput from each processor, and if the above notifications arrive from a source other than the master processor, sends the processor number and the above notifications to the master processor, and if the above notifications arrive from the master processor. By sending the data to the master alternative processor, the CPU
This has the effect of being able to detect processor failure without placing a heavy burden on the processor or making the bus busy.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an embodiment of the processor down detection method according to the present invention. 11...Master processor 12...Master replacement processor 13.14・Conflict・Slave processor 2 repair...O) IJ put notification control device 3...Shared storage device

Claims (1)

[Claims] Each master is connected to a shared storage device, is loosely coupled, can notify the start/end of autoreput, and has an autoreput function to automatically restart itself in the event of a failure. processor,
A master replacement processor and a plurality of slave processors receive a notification of the start/end of the autoreput, and if the notification comes from the master replacement processor or the slave processor, send the notification and the processor identification number to the master processor. and an autoreput notification control device for sending the notification to the master replacement processor to control processing when the notification arrives from the master processor. Processor down detection method.