JPS58144263A - Fault processing system for dispersion processing system - Google Patents

Abstract

PURPOSE:To reduce fault correcting time, by executing automatic diagnosis, automatic memory dumping and automatic IPL under the control of a console service processor in a dispersion processing system having a host and plural subhosts. CONSTITUTION:In addition to a console service processor 22 and a service file, a program file 23, a diagnosis program file 24 and a memory dumping file 25 are connected to a subhost 2. When a fault is generated in normal operation, the console service processor 22 detects the fault, loads the diagnosis program file 24, the memory dumping file 25 and the program file 23 and then, after the automatic diagnosis for hardware and automatic memory dumping and automatic initial program loading for software, reports these results to a host 1. Thus the fault can be discriminated in the host 1 side and the repairing and recovering time is reduced, improving the maintenance efficiency and system reliability.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a distributed processing system having a host and a plurality of subhosts.
This invention relates to a failure handling method suitable for improving maintenance efficiency and system reliability.

Prior Art In distributed processing systems that have a host and multiple sub-hosts, various unmanned operation methods have been attempted based on the assumption that sub-host operators and maintenance personnel are absent. However, since the subhost lacks automatic hardware diagnosis, automatic software memory dump, and automatic initial program load (IPL) functions, maintenance personnel are dispatched to the subhost when a failure is detected on the host. The current situation is as follows.

(1) Since it is unknown at the time of dispatch whether it is a hardware failure or a software failure, it is necessary to dispatch maintenance personnel for both hardware and software, and furthermore, the maintenance personnel arrive at the location where the sub-host is installed. After that, troubleshooting will start, so troubleshooting will take time.

(2) Even in the event of a hardware failure, the faulty part is not pointed out when maintenance personnel are dispatched, so after the maintenance staff arrives at the sub-host installation location, they can point out the faulty part and then procure maintenance parts. , the failure recovery time becomes longer.

(3) In the case of software failure, the system is more likely to operate than re-IPLK, but since there is no automatic re-IPL or automatic memory dump function for later software failure analysis, system reliability or software There are problems in improving quality.

Purpose of the Invention The purpose of the present invention is to solve the drawbacks of the prior art as described above, and to shorten the time for troubleshooting, improve maintenance efficiency, and improve system reliability in distributed processing systems. There is a plan.

The key feature of Jin's invention is that the subhost has a file memory such as a magnetic disk device that stores a diagnostic program (MD), a memory dump program, and regular programs, and automatically runs under the control of a console service processor. By running diagnostics, automatic memory dump, and automatic IPL, fault recovery time is shortened, maintenance efficiency is improved,
This makes it possible to improve system reliability.

DESCRIPTION OF THE EMBODIMENTS FIG. 1 is a block diagram of a distributed processing system having a host and a plurality of sub-hosts. A sub-host processing device 2 (hereinafter abbreviated as sub-host) is connected to each sub-host processing device 2 via a terminal line 12, and a terminal device group 8 is connected to each sub-host 8 via a terminal line 12.

FIG. 2 is a block diagram of one embodiment of the present invention, and for convenience only shows the connection between the host and one sub-host. In FIG. 8, in addition to a console service processor 22 and business files, a program file 28, a diagnostic program (MD) file 24, and a memory dump file 25 are connected to the subhost 2. Here, a memory dump is an output of the contents (programs, data, etc.) in the main memory to an external storage device, and is very effective in analyzing software failures.

The memory dump program executes a memory dump after being loaded into the main memory from the file 25. In addition,
21 indicates the hardcore section in Subpos) Z. FIG. 8 is a flowchart 1 for explaining the operation of this embodiment.

Operation of the sub-host 2 is started by loading the program stored in the program file 28 into the sub-host 2 upon activation from the console service processor 22 or the host 1. During normal operation, the terminal device 118
After the message arrives at subhost 2, certain processing is executed within the subhost, and then the message is sent to host 1).
The message is transferred from subhost 2 to host 1 for processing. After the processing is completed, the host 1 sends a response message to the dermatology + 18 via the sub-host 2.

Next, an explanation will be given of the operation when some kind of failure occurs during normal operation. When a machine check, which is a failure of the sub-host) 2, occurs during normal operation, the console service processor 22, after detecting the machine check, loads a diagnostic program from the MD7 aisle z4. The loaded diagnostic program ')' executes the diagnosis of the sub-host 2 by running the entire system 81 and transfers the diagnosis result to the host.

After receiving the diagnosis result, in step 1), the name of the sub-host where the failure occurred and the diagnosis result are output to the maintenance personnel.

The maintenance personnel who received the notification will
carry out repairs.

Next, the operation in the case where a program failure such as a program loop occurs during normal operation will be described. If a program failure, such as a program loop, occurs during normal operation, the console service processor 82 detects the program loop and then stores it in the memory stamp file 25.
Load the memory dump program. The loaded memory dump program executes memory dump to the memory dump file 25. After the console service processor 2z detects the completion of the memory dump, it reloads the program from the program file 2B and attempts to restart the subhost I's independent operation. Here, subhost 2
If the system restart is successful, the host) 1
Using the communication line 11 between the host 1 and the sub-host 8, the memory dump contents are transferred from the memory dump file z5 to the host 1, and the cause of the program failure is analyzed.

On the other hand, if the system restart of sub-host 2 is unsuccessful, a failure is detected by a health check from host 1, and failure repair processing is started.

That is, in a distributed processing system, a failure in a subhost is detected on the host side by transmitting diagnostic data from the host to the subhost and checking the response from the subhost to the diagnostic data. This is a health check.

Next, an explanation will be given of the operation when a 0 error, which is a failure of a peripheral device, occurs during normal operation.

If an IO failure occurs during normal operation, the operating system within the sub-host 8 determines whether the sub-host is operable or not after detecting the IO failure. If the sub-host is operational, it sends a message to host l to the effect that an IO failure has occurred, and then continues normal operation. If the sub-host 2 cannot be operated due to the IO failure, the failure is detected by a health check from the host 1, and failure repair processing is started.

Effects of the Invention As is clear from the above explanation, according to the present invention, the following effects can be obtained.

(1) Even if a failure occurs in a sub-host located in a remote location, processing device failures, program failures, and peripheral device failures can be identified on the host side, and the initial actions of dispatched maintenance personnel will be accurate.

(2) In the event of a processing device failure, the maintenance parts available are accurate, so maintenance efficiency and economic efficiency are improved.

(3) In the case of a processing device failure, the repair time is shortened.

(4) In the case of program failure, a memory dump is automatically obtained, reducing recovery time.

(5) In the case of a program failure, automatic IPL is executed, so even if a timing-related program failure occurs, operation can be continued without stopping the system, improving system reliability.

(6) Even in the case of (5) above, it is possible to send the memory dump contents to the host and analyze the cause of the program failure.
This is useful for improving program quality.

(7) If the sub-host is able to operate even in the case of a failure in the Sapphire device, appropriate maintenance is possible because the host is notified of the failure.

[Brief explanation of drawings]

FIG. 1 is a block diagram of a distributed processing system targeted by the present invention, FIG. 2' is a block diagram of an embodiment of the present invention, and FIG.
This figure is the ninth flowchart explaining the operation of FIG. 2. l...' host, 2... subhost, 8... terminal device, 11... communication line, 12... terminal line, 21
...hardcore m, 22...console service processor, z8...program file, 24...
Diagnosis program file, 25...Memory dump file. 1st Figure 3 35 5 5
3Figure 2

Claims (1)

[Claims] (1) In a distributed processing system having a host processing device, a plurality of sub-host processing devices connected to the host processing device, and nine terminal devices connected to the sub-host processing device, the sub-host processing device includes a console service processor. A file memory is provided to store a diagnostic program and a memory dump program, and under the control of the console service processor, when a hardware failure occurs in the sub-host processing device, the diagnostic program is automatically executed and the results are processed by the host. In the event of a program failure, the memory dump is automatically acquired and an initial program load is executed to recover the system, and in the event of a terminal device failure, the terminal device failure is communicated to the host processing device. A failure handling method for distributed processing systems.