JP2000242617A - Fault processing method for multiprocessor communication mechanism - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prevent a fault from being reported frequently by allowing other prosessor to continuously perform processing even in a fault processing and reporting no fault detection even if 2nd and succeeding faults are detected until a report on the completion of the fault processing is received from a processor as a management source. SOLUTION: When it is decided that a fault occurs to an inter-processor communication mechanism or another processor, the processor as the management source of the inter-processor communication mechanism or the processor whose fault is decided, requests a fault recovering processing, and sets a timer at the same time (5-2). The management-source processor 1-1 having received the fault detection report switches the in-use system to a standby system by using a device managing means in itself to perform the recovering processing (5-3). The processor having sent the fault detection information does not send another fault detection report even if detecting a fault of the same contents with the report, while the fault recovery monitor timer is in operation.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a multiprocessor distributed processing system having an interprocessor communication mechanism, and when a communication mechanism between processors fails, a redundant failure notification is sent to a processor managing the communication mechanism between processors. The present invention relates to a failure processing method for a multiprocessor communication mechanism to be suppressed.

FIG. 3 is a diagram schematically showing an inter-processor communication mechanism. A multi-processor as shown in FIG. 3 in which a plurality of processors (1-1 to 1-n) are connected by an inter-processor communication mechanism (2-1). In a processor distributed processing system, when an inter-processor communication mechanism fails, a phenomenon occurs in which each processor autonomously detects the failure and repeatedly sends a failure notification to a processor that manages the inter-processor communication mechanism. Sometimes. In such a case, an extra load is applied to the processor that manages the inter-processor communication mechanism.

[0003]

2. Description of the Related Art FIG. 4 shows an example of a system configuration of a conventional distributed processing multiprocessor.

In FIG. 4, all processors (1-1 to 1)
1-n) enables communication between processors by connecting to the inter-processor communication mechanism (2-1), and each processor has inter-processor failure communication means (A-1 to A-
n) and the device management means (B-1 to B-
n) is deployed, and for the device management of the inter-processor communication mechanism (2-1) which is a shared device of all processors, the representative processor (1-1) is provided with the device management means (C-1) of the inter-processor communication mechanism. An example of a deployed system configuration is shown.

[0005] FIG. 5 shows a sequence of a fault processing of a conventional distributed processing multiprocessor system.

Each processor performs fault detection by normal processing and fault detection by periodic tests to maintain the system semi-permanently. When a failure is detected, the following processing is performed. (3-1) In the processor that has detected the failure, the fault is separated by a continuity test or the like of the inter-processor communication mechanism.

When it is determined that the failure has occurred in its own processor as the detection source, failure recovery processing such as switching the active system to the standby system is performed using the device management means B in the own processor. (3-2) When the failure location is determined to be the inter-processor communication mechanism or another processor, the inter-processor failure communication means A is used for the processor that is the management source of the inter-processor communication mechanism or the processor determined to be faulty. Perform failure detection notification and request failure recovery processing. (3-3) The management-source processor that has received the failure detection notification performs a failure recovery process such as switching the active system to the standby system using the device management unit B in the processor itself. (3-4) (3-5) When the processor 1-1 that manages the inter-processor communication mechanism detects a failure having the same content as the processor that is recovering the failure in the inter-processor communication mechanism during the recovery, the processor A failure detection notification is sent to the processor 1-1 that manages the inter-communication mechanism, and a failure recovery process is requested. However, the next failure recovery processing is not performed until the failure recovery of (3-2) is completed.

[0008]

In a distributed processing multiprocessor system having a function of autonomously detecting a fault in each processor, when a fault occurs in an inter-processor communication mechanism, which is a processor sharing device, each of the processors becomes independent. May detect the same failure. for that reason,
The same failure notification is frequently transmitted to the processor 1-1 which is the management source of the inter-processor communication mechanism. for that reason,
In the processor 1-1 which is the management source of the inter-processor communication mechanism, there is a problem that the failure recovery processing is temporarily unable to be performed, and the processing is delayed.

SUMMARY OF THE INVENTION It is an object of the present invention to prevent a notification of a failure from occurring frequently to a managing processor and eliminate a delay in failure processing.

[0010]

According to the present invention, in a failure processing method for an inter-processor communication mechanism of a distributed multi-processor type comprising a plurality of processors, a failure recovery request is issued from a processor which detects a failure in the inter-processor communication mechanism. Is transmitted to the managing processor of the inter-processor communication mechanism, and the receiving managing processor closes the own processor to execute the failure processing, and during the execution of the failure processing, the other processors continue the processing, Until the failure processing completion notification is received from the processor of the management source, or within a fixed time (maximum recovery time), a failure detection notification is issued even if a second or later failure is detected so that the same failure is not detected. A means is provided to prevent such actions.

A fault recovery monitoring timer is set before sending a fault detection notification to a processor that manages the inter-processor communication mechanism, and the same fault is detected for the second and subsequent times until the timer is released. Even so, the failure detection notification is suppressed so as not to be performed. By providing such means, it is possible to reduce the notification of the same failure, thereby reducing the processing load on the processor that manages the inter-processor communication mechanism.

[0012]

FIG. 1 shows an example of a system configuration in a distributed processing multiprocessor according to the present invention.

In the figure, reference numeral 1-1 denotes a processor, 2-1 denotes an inter-processor communication mechanism, A-1 denotes an inter-processor communication means, B-1 denotes a device management means in its own processor, and C-1 denotes an inter-processor communication. D-1 represents a failure recovery completion notification process, and E-1 represents a failure recovery monitoring timer process.

In FIG. 1, all processors (1-1 to 1)
-N) enables communication between processors by connecting to the inter-processor communication mechanism (2-1), and each processor has inter-processor failure communication means (A-1 to A-
n) and the device management means (B-1 to B-
n), and the device management means of the inter-processor communication mechanism (2-1), which is a shared device of all processors,
A system in which a device management means (C-1) of an inter-processor communication mechanism is provided in a representative processor (1-1) is shown as an example.

Compared with FIG. 4, in FIG. 1, the inter-processor failure communication means (A-
A failure recovery completion notification process (D-1) is added to 1), and a failure recovery monitoring timer process (E) is provided to the inter-processor failure communication means (A-2 to An) of the processors (1-2 to 1-n). -2
To En) are added.

FIG. 2 shows a failure processing sequence of the distributed processing multiprocessor system for explaining the present invention.

Each processor performs fault detection by normal processing and fault detection by periodic tests to maintain the system semi-permanently. When a failure is detected, the following processing is performed. (5-1) In the processor which has detected the failure, the fault is separated by a continuity test or the like of the inter-processor communication mechanism.

When it is determined that the failure has occurred in its own processor as the detection source, failure recovery processing such as switching to the standby system is performed using the device management means B in the own processor. (5-2) When the failure location is determined to be the inter-processor communication mechanism or another processor, the inter-processor failure communication means A is used for the processor that is the management source of the inter-processor communication mechanism or the processor determined to be faulty. Perform failure detection notification and request failure recovery processing. At the same time, a timer is set in the failure recovery monitoring timer process E. (5-3) The processor 1-1 that has received the failure detection notification performs failure recovery processing such as switching the active system to the standby system using the device management unit B-1 in the own processor. (5-4) While the failure recovery monitoring timer is running, the processor that has issued the failure detection notification suppresses transmission of the failure detection notification even if the processor detects the same failure that has been notified.

The trigger for canceling the failure recovery monitoring timer is as follows.
The case where the notification of the processing restart is received from the processor 1-1 of the management source that has received the failure detection notification and the case where the failure processing of the processor 1-1 of the management source that has received the failure detection notification has not been completed within the maximum failure recovery time. Case (timeout). (5-5) (5-6) (5-7) If the processor that has received the failure detection notification detects the same failure in a processor that has not received the failure detection notification during recovery from the failure, the management source A failure detection notification is transmitted to the processor of the processor, and a failure recovery monitoring timer is set. However, the processor 1-1 of the management source that has received the failure detection notification does not execute the processing during the recovery from the failure in (5-2). (5-8) After receiving the failure detection notification, the management processor 1-1 sends a failure recovery completion notification to each processor after the failure recovery processing is completed.

The processor 1 that has received the failure recovery completion notification
Steps 2-2 to 1-n cancel the failure recovery monitoring timer and restart the normal processing.

By adopting the above-described processing, even if the failure detection of the same inter-processor communication mechanism occurs frequently in each processor, the processor 1-1 as the management source of the inter-processor communication mechanism may be used.
, It is possible to transmit only the minimum necessary failure detection notification.

[0022]

As described above, according to the present invention,
If an error occurs during failure processing in the management-source processor, processors other than the management-source processor can perform processing. Therefore, even if an error occurs during failure processing in the representative processor, the system does not stop. In addition, there is no interruption of processing other than the processor that executes the failure processing, and even if failure detection of the same inter-processor communication mechanism occurs frequently in each processor, only a minimum necessary failure notification is transmitted to the management processor, The processing load on the managing processor is reduced. Further, the effect that the delay of the processing of the processor that does not execute the failure processing is reduced is obtained.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating an example of a system configuration in a distributed processing multiprocessor of the present invention.

FIG. 2 is a diagram showing a failure process in the distributed processing multiprocessor of the present invention.

FIG. 3 is a diagram schematically illustrating an inter-processor communication mechanism.

FIG. 4 is a diagram illustrating an example of a system configuration in a conventional distributed processing multiprocessor.

FIG. 5 is a diagram showing a failure process in a conventional distributed processing multiprocessor.

[Explanation of symbols]

1-1, 1-2-1-n Processor 2-1 Communication mechanism between processors A-1, A-2 to An Fault communication means between processors B-1, B-2 to Bn Internal processor Device management means C-1 Device management means of the interprocessor communication mechanism D-1 Failure recovery completion notification processing E-2 to En Failure recovery monitoring timer processing

 ────────────────────────────────────────────────── ─── Continued on the front page (72) Inventor Masahiko Noguchi 3-19-2 Nishi-Shinjuku, Shinjuku-ku, Tokyo Nippon Telegraph and Telephone Corporation F-term (reference) 5B034 BB02 5B045 BB28 BB49 JJ09 JJ13 JJ48 5B048 AA05 AA18 CC14 5B089 GA02 JA35 JB17 KA06 KA12 KB04 KC28 KC43 MC18 MD02 ME02 ME04

Claims (1)

[Claims] 1. A failure handling method for an inter-processor communication mechanism in a distributed multiprocessor mode comprising a plurality of processors, comprising: a processor that detects a failure in the inter-processor communication mechanism; The managing processor that has transmitted and received the original processor closes to its own processor and executes the failure processing, and the other processor continues the processing while the failure processing is being executed.
A multiprocessor communication mechanism, wherein a failure detection notification is not performed even if a second or subsequent failure is detected until a failure processing completion notification is received from a management processor or within a predetermined time. Failure handling method.