JPS63288337A - Fault processing system for hot stand-by system - Google Patents

Abstract

PURPOSE:To obtain a supervisory device of a low cost by adding a function by which an existing system host and a stand-by system host can execute communication through a system control mechanism, and executing periodically a report of a fact being normal, from the existing system host to the stand-by system host through a general interface adaptor. CONSTITUTION:For instance, it is supposed that a normal notice 14 from an existing system host 1 is stopped through a communication path 25 via a GSA 15 (general interface adaptor). In this case, it is also considered that the existing system host 1 is broken down, but it is also considered that the notice is stopped due to a fault of said path 25, or a failure of an inter-host communication processing of an existing system host control program 11 and a stand-by system host control program 31, and in case of the latter, the existing host 1 is normal, therefore, switching to a stand-by system host 3 is not required. Accordingly, in order to demarcate whether the existing system 1 is broken down or not, a patrol message is transmitted to the existing system host 1 through a system control mechanism 23.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to a hot standby method in an information processing system, and in particular, to a hot standby method in an information processing system, and in particular to a hot standby system that deals with failures when a host-to-host interface mechanism that connects an active host and a standby post breaks down. Regarding processing method.

[Conventional technology]

Conventionally, in this kind of true standby system.

The active host and standby host are connected by a host-to-host interface mechanism, and when the standby host detects that the active host is down via the two mechanisms, the standby host connects the active host via the system control mechanism. After stopping,
The standby host took over the tasks that were being processed by the active host.

[Problem that the invention seeks to solve]

However, normal responses from the active host stop not only when the active host actually goes down, but also when the host-to-host interface mechanism fails (reliability issue) or when the host-to-host interface mechanism is controlled. This can also occur due to software defects (quality problems), so the conventional hot standby system described above also has the disadvantage of stopping the active host during business processing in such cases. There is also a method of duplicating the inter-host interface mechanism.

In this case, although the reliability of the host-to-host interface mechanism is improved, the disadvantage is that the cost is doubled.

[Means for solving problems]

The present invention includes an active host and a standby host.

In a hot standby system where the active host and standby host are connected by a general-purpose interface adapter, and further connected by a system control mechanism so that the standby host can control the stoppage of the active host, the system control mechanism is By adding a function that allows the active host and the standby host to communicate via the general-purpose interface adapter, the active host periodically reports that it is normal to the standby host via the general-purpose interface adapter, and the standby host 7 If a normality report cannot be received from the active host, the standby host queries the active host via the system control mechanism and at the same time monitors whether there is a response from the active host. When this occurs, the active host is normal and business processing continues.

When there is no response, the standby host stops the active host via the system control mechanism, and then the standby host automatically takes over the business that was being processed by the active host. do.

〔Example〕

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

FIG. 1 is a block diagram showing the configuration of a hot stun sieve system according to an embodiment of the present invention. Active host 1
and the standby host 3 are connected via an inter-host communication station 25 and a system control mechanism 23, and the active host 1.
Each standby host 3 has a control program 11.31, and a database 12 dedicated to the active host 1 is also connected to the standby host 3.

Figure 2 shows the active host control program 1 shown in Figure 1.
1 and the standby host control program 31.

The operation of this embodiment will be described below with reference to FIGS. 1 and 2.

First, it is assumed that the active host 1 is performing business processing using the control program 11 and database 12. It is also assumed that the standby host 3 uses the control program 31 to quickly pull in the database 12 when the active host 1 goes down and is in a state where it can continue business operations, that is, in a hot standby state.

At this time, the internal states of the active host 1 and standby host 3 are as follows. That is, the active host 1 performs processing as shown in process 14 and process 14-1.

Based on the control program 11, the active host normality notification 14 is sent to the standby host 3 at regular intervals via the communication path 25 that connects the general purpose interface adapter (GSA) 5 shown in FIG. 1 and the GSA 35. is sent repeatedly. On the other hand, on standby host 3.

In processing 34 and processing 36, it is checked whether or not the normal notification 14 from the active host 1 via the GSA communication path has been received, and if it has been received, the active host 1 will continue to be down. Processing 34 and processing 36 are repeated to detect.

Here, communication via GSA No. 4? Suppose that the normal notification 14 from the active host 1 via the host 25 is discontinued.

In this case, it is conceivable that the active host 1 is down, but there is also a failure of the communication via GSA/4' bus 25.

It is also possible that the interruption occurred due to a malfunction in the inter-host communication processing of the active host control program 11 and the standby host control program 31; in these latter three cases, the active host 1 is normally executing business processing, so There is no need to switch to the standby host 3.

Therefore, it is determined whether the active host 1 has gone down or not through subsequent processing. That is, processing 3 on the standby host 3 side
If it is determined that normal notification has ceased due to 6, process 37
A') A roll message is sent to the active host 1 via the system control mechanism 23.

Here, if the active host 1 is not down, the patrol message is received in steps 17 and 18, and a response to the patrol message reception is sent to the standby host 3 via the system control mechanism 23. As a result,
By receiving and receiving responses to the patrol messages through the process 38 on the standby host 3 side, it can be determined that the active host 1 is about to go down, and subsequent downtime of the active host 1 is detected without system switching. The process then returns to step 34.

On the other hand, if the active host 1 is down, it is checked in process 38 that no response to the patrol message reception is received, and the process advances to process 39. As a result, the power to the active host 1 is cut off via the system control mechanism 23, and after the active host 1 is completely stopped, system switching is performed in process 32, and the system is switched from the point in time when the processing on the active host 1 is interrupted. Business can be resumed.

〔Effect of the invention〕

As explained above, the present invention is directed to a mechanism provided solely for the purpose of stop control between hosts.

Added a function that allows hosts to communicate with each other, and when the active host loses normal notification via the host-to-host general-purpose interface mechanism, it is possible to check if the active host is down by communicating between the hosts using the above additional function. Because you can
This has the effect of eliminating unnecessary system switching. Furthermore, compared to the functions provided by conventional system control mechanisms, the expansion of functions at the protocol level that enables communication between hosts is limited, so the cost is lower than duplicating the general-purpose interface mechanism between hosts. It also has the advantage of being cheaper.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of a hot standby system according to an embodiment of the present invention, and FIG. 2 is a part of the processing flow related to failure handling of the active host control program and standby host control program shown in FIG. It is a flowchart which shows. 1... Active host, 3... Standby host), 11゜31... Host control program, ]2... Active host database, 15.35... GSA (general purpose interface adapter), 25...Communication path via GSA. 23...System control mechanism.

Claims (1)

[Claims] 1. Includes an active host and a standby host, connects the active host and the standby host with a general-purpose interface adapter, and further connects the active host and the standby host to send/receive messages between the hosts and stop them. In a hot standby system connected to a controllable system control mechanism, the active host periodically reports normality to the standby host via the general-purpose interface adapter, and the standby host If the host is no longer able to receive normal reports from the active host,
The standby host makes a health inquiry to the active host via the system control mechanism, and at the same time monitors whether or not there is a response from the active host, and when there is a response, the active host executes business processing. When there is no response, the standby host stops the active host via the system control mechanism, and then the standby host automatically resumes the work being processed by the active host. A failure handling method in a hot standby system characterized by takeover.