JPH0122654B2 - - Google Patents

Description

[Detailed description of the invention]

[Industrial Application Field] The present invention is designed to identify the cause of an error that occurs during input/output processing in a data processing system that includes a plurality of input/output devices and their control devices (hereinafter referred to as input/output devices). This invention relates to a method for accurately detecting whether a device is present. [Prior Art] Input/output devices include multiple layers of devices such as channel devices, input/output control devices, input/output adapters, and input/output devices, and multiple access paths are provided for one input/output device. The connection is complicated. Therefore, if an error occurs while performing input/output processing on a certain input/output device, it is not easy to determine which device is actually causing the error. However, it is necessary to discover a malfunctioning device as early as possible and either disconnect it from the system or repair it. Conventionally, it has been proposed to use statistical methods to identify faulty devices in such complex systems. For example, in Japanese Patent Application Laid-Open No. 49-66042, an error counter for each device is updated with an update amount according to the degree of possibility related to the error for each type of error.
The device corresponding to the counter showing the highest value is considered to be the cause of the failure. [Problem to be solved by the invention] However, when using such a statistical method,
There is a problem that accuracy is poor unless the parameter is large to some extent and distributed evenly. That is, if the actual input/output processing is biased toward a specific input/output device, or if the amount of input/output processing within a certain period of time is small, accurate detection cannot be performed. In order to increase the population and expect an even distribution, it is necessary to continue counting for quite a long time, which causes a problem in that identification of faulty devices is delayed. [Means for Solving the Problem] Therefore, in the present invention, two thresholds are set for the error counter value, and when the counter exceeds the first lower threshold, diagnostic input/output processing is generated,
As a result, when a counter exceeds a second higher threshold, the faulty device is identified from the value of each counter. [Effect] In other words, when the first threshold is exceeded, the number of input/output processes may be small or may be biased toward a specific device, but any abnormality will not be detected. Ru. Then, by generating diagnostic input/output processing and forcing all input/output devices to perform input/output processing using all paths, the above-mentioned problem of number of times and imbalance can be solved. Now, assume that there are devices A, B, and C, each having error counters α, β, and γ, and that there are two routes for accessing device C, A→C and B→C. A
→If an error occurs in route C, the error counters corresponding to A and C are counted up. Similarly, when an error occurs on the route B→C, the error counters of B and C are counted up. Assuming that the requests (loads) from A to C and B to C are equal, the following can be said. however,
Assuming that multiple devices do not experience failures at the same time, and considering that it takes only seconds to detect a failure, and that there are no simultaneous failures as a practical matter,

[Table] −
←

Claims (1)

[Scope of Claims] 1. In a data processing system that includes at least a plurality of input/output devices and one or more input/output control devices that control them, an error counter is provided for each of the devices, and a certain input/output If an error occurs during input/output processing to a device, the error counters of each device involved in the input/output processing are updated, and if the value of any error counter exceeds the first threshold, all input/output Input/output processing is generated for the output device via all paths, and if an error occurs at that time, the error counter is updated in the same way as above, and the value of one of the error counters is set to the above first value. A fault detection method for an input/output device, characterized in that, when a second threshold larger than a threshold is exceeded, a faulty device is identified based on at least an error counter that exceeds the second threshold. 2. Claim 1, characterized in that the update amount of the error counter differs depending on the type of error.
Fault detection method for input/output devices described in section. 3. The input/output device according to claim 1 or 2, characterized in that when any of the above error counters exceeds a second threshold, a faulty device is determined based on the values of a plurality of error counters. Failure detection method for system equipment. 4. Claim 1, characterized in that the error counter is provided for each device, and if the device is connected to a host device through two or more paths, it is provided for each path. or third
A fault detection method for an input/output device as described in any of the above.