JPH01134637A - Supervising system for information in stall processing system - Google Patents

Abstract

PURPOSE:To automatically detect a stall by executing the hardware processing with a service processor when a system abroad processing cannot be executed by an operating system at the time of generating a watch dog timer run-out. CONSTITUTION:When an operating system OS is activated, the OS sets, for example, 15sec to a watch dog timer 3a in a system control device 3 and executes an instruction. When the processing is completed, the OS activates the processing, for example, after 10sec. Presently, the activation is not executed by the stall of the OS, and after 15sec pass, the run-out of the timer 3a is generated. Then, the device 3, when an indicator 3b is set, informs a service processor 4 of the run-out, and when the indicator 3b is reset, this is set and the run-out is informed to the OS. Next, the system is stopped and the hardware information of respective devices is collected. Thus, without using a special device, the stall detection can be executed automatically.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to error handling in an information processing system, and particularly to a stall monitoring method including system stall detection and post-detection processing in an information processing system.

[Conventional technology]

In general, conventional stall monitoring in an information processing system that includes an arithmetic processing unit, an input/output control device, a system control device, a main storage device, and a service processor that can collect information on each of these hardware is performed using a system monitoring device, etc. This was done either automatically using special equipment, or by monitoring the situation by an operator.

[Problem that the invention seeks to solve]

As mentioned above, conventional stall monitoring methods either connect a special device such as a system monitoring device or are monitored by an operator; the former requires an expensive special device called a system monitoring device. The disadvantages are that system prices have increased, and since the latter is solely responsible for monitoring by operators, it takes time to notice abnormalities, and information necessary to analyze the cause may not be collected due to operational errors.

An object of the present invention is to provide a stall monitoring system that can automatically perform stall monitoring without using special equipment such as a system monitoring device, and can automatically and reliably collect information necessary for cause analysis. The goal is to provide a method.

[Means for solving problems]

In order to achieve the above object, the present invention provides an information processing system including an arithmetic processing unit, a human output control device, a system control device, a main storage device, and a service processor capable of collecting information on each of these hardwares. , a watchdog timer whose stop and value can be controlled from the operating system; an indicator that is reset at system initialization; and a watchdog timer that sets the indicator if it has not been reset when the watchdog timer runs out. notification means for notifying the operating system of the occurrence of a runout of the watchdog timer, and notifying the service processor of the runout of the watched knob timer if the indicator is set; The watchdog timer is configured to set a predetermined time to start, and thereafter restart the watchdog timer every time within the predetermined time, and receives a notification of runout of the watchdog timer from the notification means. The service processor is configured to stop the watchdog timer after performing a system abort process, and the service processor collects hardware information of each device when receiving a notification of runout of the watchdog timer from the notification means. has.

[Effect]

If the operating system stalls for some reason after it starts operating, the watchdog timer runs out because it cannot be restarted. At the first watchdog timer runout after system initialization, the indicator has been reset, so the notification means notifies the operating system, the operating system performs system abort processing, and then stops the watchdog timer. . If the operating system operates normally, the abort process described above is normally performed and the watchdog timer is stopped. However, if the operating system stalls again, the watchdog timer is not stopped and the watchdog timer runs out again. Since the indicator is set when the watchdog timer runs out for the second time, the notification means notifies the service processor of this fact, and the service processor collects hardware information of the arithmetic processing unit and the like.

〔Example〕

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

FIG. 1 is a block diagram showing an example of an information processing system to which the present invention is applied, in which arithmetic processing units (hereinafter referred to as EPUs) 1. Main memory unit (hereinafter referred to as MMU)2. System control unit (hereinafter referred to as SCU) 3, service processor (
(hereinafter referred to as SvP) A magnetic disk device 41 and a console 42 connected to the 4°5VP4. Consists of an input/output control device (hereinafter referred to as 10P) 5, a magnetic disk device 51 connected to the 10P5, and a console 52, and 5CU
3 is EPtJ4. MMU2,5VP4. Connected to l0P5. Further, in this embodiment, a watchdog timer (hereinafter referred to as WDT) 3a, an indicator 3b, and a notification means 3c are provided in the 5CUa.

Note that the indoor cake 3b may be provided at another location.

FIG. 2 shows a configuration example of the WDT 3a provided in 5CUa, where 31 is a WDT register, 32 is a subtracter, 311 is a WDT setting signal line, 312 is a WDT runout notification signal line, 313 is a read signal line, and 321 is a WDT The activation signal line 322 is a write signal line.

FIG. 3 is a flowchart showing an example of processing performed when starting up the operating system (hereinafter referred to as O8) that controls the basic control of the information processing device shown in FIG. 1, and FIG. FIG. 5 is a flowchart showing an example of the OS exception processing that is started when a WDT runout exception occurs; FIG. 6 is a flowchart showing an example of processing within 5CUa that is executed when a WDT runout occurs; FIG. 7 is a flowchart of an example of processing within SVP d that is executed when the occurrence of WDT runout is notified from 5CtJ3.

The operation of this embodiment will be explained below with reference to each figure.

When startup of the information processing system shown in FIG. 1 is started by an instruction from the console 52, for example, control is passed to processing 61 shown in FIG. In process 61, for example, O8 is loaded from the magnetic disk device 51 to MMU2 via l0P5, 5CU3 and O3 is started up, and then the process proceeds to process 62. In process 62, the WDT value setting/starting process shown in FIG. 4 is started, thereby completing the O3 starting process. Also, at this time, the indicator 3b is reset.

Next, control passes to process 63 of O3 in FIG. 4, which was activated in process 62. In process 63, O3 is WDT3 in 5CU3
A WDT value setting/starting command is executed to set a value of, for example, 15 seconds and start the WDT 3a. When this command is executed by the EPUI, the contents are transmitted to the 5CU3 through inter-processor communication, and the 5CU3 sets a value indicating 15 seconds in the WDT register 31 via the WDT setting signal line 311 in FIG. Thereafter, the value of the WDT register 31, which sets the logic value of the WDT activation signal line 321 to "1", is input to the subtracter 32 via the continuous output signal line 313. When the logical value of the WDT start signal line 321 is "1", the subtracter 32 subtracts the contents of the WDT register 31 read out via the read signal line 313 by "1" every second, and outputs the contents to the write signal line 322. The value is returned to the WDT register 31 via.

When the value of the WDT register 31 becomes rOJ, the logical value of the WDT runout notification signal line 312 is set to "1",
5CU3 recognizes the occurrence of runout of WDT3a.

When the process 63 in FIG. 4 is completed, the OS next proceeds to process 64. In process 64, in order to set/start the WDT value again within a time shorter than the 15 seconds set in the WDT 3a, a request is made to, for example, an interval timer or the like to start the WDT value setting/starting process shown in FIG. 4 after, for example, 10 seconds. The process ends.

While O3 is operating normally, the WDT value is set and the WDT 3a is activated every 10 seconds, and runout of the WDT 3a does not occur.

Suppose now that O8 is stalled due to a bug. Under normal conditions, WDT value setting/startup was performed every 10 seconds, but due to O8 stall, WDT value setting/startup was performed as shown in Figure 4.
The startup process is not started, and a runout of the WDT 3a occurs after 15 seconds have passed.

As a result, the logical value of the WDT runout notification signal line 312 becomes "11," and the 5CU3 recognizes the runout of the WDT 3a.Thereby, the 5CU3 starts the process shown in FIG. 6 by the notification means 3C.

First, the 5CU3 determines whether or not the indicator 3b is set in a determination process 81. If the indicator 3b is set, the process proceeds to process 82, and if the indicator 3b is reset, the process proceeds to process 83.Here, The indicator 3b is reset at the time of system initialization, and its state is not changed thereafter except when it is set in process 83. Therefore, after system startup, 1
When the WDT runout occurs for the second time, it is in the reset state, and as a result of the judgment process 81, the process proceeds to process 83. In process 83, the indicator 3b is set to indicate that the first WDT runout has occurred, and the process moves to process 84. In process 84, the occurrence of WDT runout is reported to O3 as an exception, and the process ends.

When a WDT runout exception is reported to O8, the OS exception handling shown in FIG. 5 is started. First, in process 71, in order to continue monitoring with the WDT 3a, WDT settings/
Start the startup process. As a result, the WDT setting/activation process shown in FIG. 4 starts running again every 10 seconds. Next, in process 72, O8 collects a memory dump necessary for finding the cause of the stall in the magnetic disk device 51. After that, in process 73, system abort processing is performed, and in process 74, W
A stop command of the DT 3a is executed to stop the operation of the WD 73a, and the system is stopped in step 75.

It is sufficient that the O8 operates normally after the WDT run completion exception is notified to the O8, but in some cases the system may stall again. If O5 stalls again in this way, WDT runout occurs again as described above, and the judgment process 81 of FIG. 6 of 5CU3 is activated. Since this is the second WDT runout and the indicator 3b has been set, the process proceeds to process 82 based on the result of determination process 81. In process 82,
5CtJ3 notifies 5VP4 of the occurrence of WDT runout, and ends the process.

When the 5VP4 is notified of the occurrence of WDT runout, the process shown in FIG. In process 92, EPIJI, MM
The hardware information of each device U2, 5CU3, and tOP5 is collected and stored in the magnetic disk device 41, and the process ends.

〔Effect of the invention〕

As explained above, the present invention starts from O8.

A WDT that can be stopped and its value set, and if this is the first runout of this WDT, it will notify the O8 to that effect,
If it is the second time, provide a notification means to notify the SVP,
If system abort processing cannot be performed in O5 when a WDT runout occurs, hardware information is automatically collected in SvP, so the system can automatically stall without using special equipment such as a system monitoring device. This has the effect of not only being able to detect a stall but also automatically and reliably collecting information when detecting a stall.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing an example of an information processing system to which the present invention can be applied, Fig. 2 is a diagram showing an example of the configuration of the WDT, Fig. 3 is a flowchart of an example of processing at the time of starting the OS, and Fig. 4 is a block diagram showing an example of the configuration of the WDT.
Figure 5 is a flowchart of an example of T value setting/startup processing; Figure 5 is a flowchart of an example of OS exception handling that is activated when a WDT runout exception is notified; Figure 6 is a flowchart of an example of SCU processing that is executed when a WDT runout occurs; Figure 7 shows the Sv activated upon notification of WDT runout occurrence.
It is a flowchart of the processing example of P. In the figure, ■... Arithmetic processing unit (EPU) 2... Main memory unit (MMU) 3... System control unit (SCU) 3a... Watchdog timer (WDT) 3b...
Indicator 3c...Notification means 4...Service processor (SVP) 5...Input/output control device (IOP) 31...WDT register 32...Subtractor 41.51...Magnetic disk device 42. 52...Console

Claims (1)

[Claims] An information processing system including an arithmetic processing unit, an input/output control device, a system control device, a main storage device, and a service processor capable of collecting information on each of these hardwares, A watchdog timer whose settings can be controlled from an operating system, an indicator that is reset at system initialization, and when the watchdog timer runs out, if the indicator is not reset, the indicator is set and the operating system is set to watch. and notification means for notifying the occurrence of runout of the dog timer and, if the indicator is set, notifying the service processor of the runout of the watchdog timer, the operating system notifying the watchdog timer at the start of its operation. The watchdog timer is configured to be activated by setting a predetermined time, and thereafter restart the watchdog timer every time within the predetermined time, and to perform system abort processing when receiving a notification of runout of the watchdog timer from the notification means. The service processor is configured to stop the watchdog timer after the watchdog timer has run out, and the service processor collects hardware information of each device when receiving a notification of the runout of the watchdog timer from the notification means. Processing system stall monitoring method.