JPH02199557A - Method of acquiring memory dump - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to a method for obtaining a dump based on an extended architecture in a large-scale computer, and in particular, to a method for obtaining a dump based on an extended architecture in a large-scale computer, especially when it is necessary to minimize system downtime such as an online real-time system. The present invention relates to a memory dump acquisition method suitable for a certain computer system, which acquires a memory dump of a large capacity storage device in a short time.

(Prior art) A memory dump is a method that saves the state at the time of a system failure by storing the contents of the computer's memory on a storage medium, and is used for researching and debugging the cause of the failure.System Failure There are differences depending on the level, but the operating system (OS), which is the basic control program, and the database/data communication program (
If the computer cannot continue operating due to a program failure such as DB/DC, a separate dump acquisition program (self-loading dump program) is loaded and the memory contents are output onto the storage medium.

(Therefore, the contents of the memory area loaded with the self-loading dump program are different from the contents at the time of the failure. This is unavoidable.) In that case, HIT,
, AC Program Product VO33/ES System Support Manual Volume 2 Chapter 1 Pages 17-46 (manual code 8091-3-057, December 1980), in the conventional self-loading dump, all the memory The contents of the device are unconditionally captured on a dump storage medium.

A obtained when one job ends abnormally
S obtained when a BEND dump, debug dump, or O8 error occurs but the system does not go down.
In vC dump and the like, partial acquisition is performed in which only the part related to the abnormality is dumped.

[Problem to be solved by the invention]

In the extended architecture of large computers, 31-bit addressing has made it possible to implement a maximum storage capacity of 2 GB. Even using today's fastest storage media, it takes 10 to 15 seconds to dump this large storage capacity.
Requires acquisition time of more than one minute. On the other hand, the time during which the computer can be stopped is required to be short due to the constraints of its intended use, and in particular, in online real-time systems that require prompt response, the computer may be stopped during the above-mentioned dump acquisition. It has become possible for things to happen even in unforgivable situations. The present invention aims to shorten the dump acquisition time in order to solve the problem of the extended system stop time for dump acquisition.

[Means to solve the problem]

In order to achieve the above objective, the present invention considers a method in which, even in the case of self-loading dumps when the system is down, only a portion of the storage device is acquired while the system is stopped, and the remaining portion is acquired after the system is restarted. . Conventionally, this partial dump acquisition has been carried out by SvC dump, etc., but this is a method of dumping the area where the main information is stored depending on how the O8 storage device is used. However, this is based on how the storage devices are used in a running system, and is based on the premise that the system is running.

The present invention takes advantage of the O8's ability to logically isolate a specified range of storage and operate without using any part in order to perform a memory dump on a downed system. That is, only a specific portion of the storage device is dumped while the system is stopped after the system goes down, and other portions are obtained after the system is restarted. At this time, in order to avoid damaging the contents of the memory area that has not been dumped yet, the system is restarted with the unobtained area separated (offline state). These unobtained areas are sequentially dumped after the system is started up, and the areas for which the acquisition has been completed are brought into a usable state (online state) and connected. This makes it possible to shorten the system downtime related to dump acquisition until the system is restarted. Finally, the entire area of the storage device can be dumped, and the usage state of the storage device after restarting can also be returned to the state before the system was stopped. (However, since the available storage area is small when restarting the system, -
Performance deteriorates over time. ) In the present invention, since restarting the system is performed only in the area for which dump acquisition using self-loading dump has been completed, the distinction between the area for which dump has been acquired and the area for which dump has not been acquired is correct for restarting the system. It needs to be passed on. Currently, the self-loading dump acquisition operation and the system restart operation are completely independent, so if things continue as they are, the operator will have to specify twice the identification information of whether the dump has been acquired or not. This causes erroneous operation and complicates the operation. Therefore, in the present invention, a self-loading dump program is incorporated at the beginning of the restart O8 program as part of the O8 program, and the identification information of dump acquired/unacquired is set to 1.
This malfunction can be prevented by simply specifying the number of times and using it in common for both parties.

[Effect]

In the present invention, regardless of how the O8 uses the storage devices, the system can be restarted as long as the dump of the minimum amount of storage devices necessary for the next startup operation is completed. Since the storage area required to start up such a system is generally concentrated in low-order addresses, it is possible to efficiently obtain a partial dump of this area.

On the other hand, storage areas that have not yet been dumped and should be separated when the value is raised can be stored in a relatively high-order address section, making it possible to efficiently obtain these dumps and connect them to the system.

〔Example〕

An embodiment of the present invention will be described with reference to FIG. 1 and subsequent figures.

The present invention shortens the system downtime by dividing the dump obtained during system stoppage due to system downtime, and makes it possible to obtain a dump of all storage devices by obtaining the remaining dump after the system restarts. It has characteristics.

FIG. 1 shows the relationship between dump acquisition during system stoppage, next system restart, and dump acquisition after system restart in the present invention. In FIG. 1, a dump is immediately obtained for a 64-Mbyte storage device (111) on the lower address side of the 2-Gbyte real storage device fi (110) installed. Remaining storage device (112
), no dump is taken at this point. When the system is restarted, the storage device on which the dump has already been taken (
The system is restarted only in step 131), and the storage device (132) for which no dump has been obtained is not used for restarting the system and its contents are saved. Then, the dump program is executed in the restarted system to complete dump acquisition for all areas.

FIG. 2 shows the configuration of the self-loading dump program. The self-loading dump control section (201) controls the following processing sections. The output control unit (204) controls the dump acquisition medium (recognizes device addresses, processes input/output start/completion interrupts), and controls the real storage device dump unit (204).
2) acquires a dump of the contents of the real storage device, and the auxiliary storage device dump unit (203) acquires a dump of the contents of the auxiliary storage device.

As shown in FIG. 3, first, the self-loading dump program is transferred from the storage medium to IPL (Initial
Program Load, 301) L. Transfers control to the self-loading dump control unit (302). When the output control section is executed, the dump program inputs the device address of the dump acquisition medium (303), and when the real storage device dump section is executed, it inputs the real address of the area from which the dump should be acquired (304). Then, the contents of the real storage device within the specified range are output to the dump acquisition medium at the specified address (305).

Previously, there was no range of areas to be dumped, and the fourth
As shown in the figure, the entire area of the real storage device was dumped (405). At the same time, the contents of the auxiliary storage device in the virtual storage system are dumped (406). In the present invention, it is not possible to dump the contents of the auxiliary storage device at point 4.

In order to obtain a dump of the contents of the auxiliary storage device, information related to virtual memory such as page tables and segment tables must be obtained as a dump, but since the dump is a partial acquisition in the present invention, this information is This is because the dump is not necessarily acquired at the same time. Also, for the purpose of shortening the dump acquisition time during system down, it is better to dump the auxiliary storage device after the system is restarted.

When starting up the system, in order to manage the real storage device, PFT E (PageFr
ame Table Entry). This P
The state and attributes of the relevant real page frame are described in the FTE.

The existence of this PFTE means the existence of a real page frame. In the present invention, there is a real storage device that cannot be used when the system is restarted because a dump has not been obtained. A bit for displaying this is newly established in the PFTE of each real page frame. This bit is the initial offline bit (PF
TEIOFF).

The meaning is as follows.

PFTE! 10FF='1': The relevant real page frame cannot be used at system startup because a dump has not been obtained. Further, the state of the real page frame is offline.

= g()): The page frame □ is usable and online at the time of system startup.

FIG. 5 shows the processing procedure when restarting the system according to the present invention. In the present invention, when the system is restarted, there are parts of the real storage device that can be used and parts that cannot be used. However, for any real page frame in any state, a PFTE is created when the system is restarted (50i, 502).

Then, input the range of real storage devices that can be used to start the system (11). After that (503), turn off the initial offline bit in the PFTE that corresponds to the real page frame in the real storage device in the available range. and displays that it is a usable real storage device (504). In addition, the initial offline bit in the PFTE belonging to the real page frame in the real storage device in the unusable range is turned on to indicate that the real storage device is unusable (505), and these Make the real page frame offline.

As shown in FIG. 6, in the past, this initial offline bit was not taken into account, so it was cleared to zero by the initial reset when creating the PF'TE (603).

FIG. 7 shows the dump acquisition procedure performed after restarting the system. Basically, it searches for real pages with the initial offline bit set to on in each real page frame (701 and 702), obtains a dump of this page (703), and returns the page to a usable state (online state) with the initial offline bit set to off. , 70
Repeat step 4) until the last page (706),
Follow these steps. □ In which case, it is better to bring each page online in batches by a certain amount, rather than bringing each page online (704, 705). This is because it takes overhead to refresh the pagination state.

Currently, self-loading dump acquisition and system startup are performed by completely independent operations. Therefore, the designation of the dump acquisition area theft range in the cell loading dump and the designation of the range of real storage devices that cannot be used at the time of system startup are also performed in 1i11-n. However, this causes erroneous operation and is complicated. Therefore, erroneous operations can be prevented by linking the self-loading dump with the system startup, making the specification of the dump acquisition range offline and the specification of the range of the universal storage device common.

In other words, first, when the required dump acquisition is completed using self-loading dump, the system startup operation is executed as a continuous process of the program, and at that time, the value specified as the dump acquisition range is used as is for the real storage device that should be taken offline. As a result, the area that has not yet been dumped is correctly dumped.

〔Effect of the invention〕

According to the present invention, a memory dump can be acquired in a short time. In this case, the amount of dump that must be obtained while the system is stopped is only the amount of memory required for the next system startup, which is about 1/4 to 1/16 of a large capacity storage device of 1 to 2 GB. be. The system downtime associated with dump acquisition is proportionally shortened.

In addition, by linking self-loading dumps with system re-booting, it is possible to standardize the specification of the range in which the dump should be obtained and the specification of the range in which the contents of the storage device should be saved by going offline after rebooting. This prevents dumps from being omitted due to incorrectly specifying both.

[Brief explanation of the drawing]

FIG. 1 is an overview of the behavior of the self-loading dump program and partial dump acquisition in the present invention, FIG. 2 is a configuration diagram of the self-loading dump program, FIG. 3 is a flowchart of dump acquisition of the self-loading dump in the present invention, and FIG. Figure 5 is a flowchart of the conventional self-loading dump acquisition process, Figure 5 is a flowchart of the system restart process in the present invention, Figure 6 is a flowchart of the conventional system restart process, and Figure 7 is the main It is a flowchart of dump acquisition after restarting the system in the invention. 110... Real storage device, 111... Lower 64 Mbyte area, 112... Remaining area, 131... Storage device for which dump has been obtained, 132... Storage device for which dump has not been obtained. Figure 3 ~ Figure

Claims (1)

[Claims] 1. When writing the contents of a computer's storage device to another secondary storage device and performing a memory dump, only a portion of the storage device is dumped, and the startup operation when restarting the system is performed by acquiring this dump. The remaining part that has not been dumped is stored in the offline state of the operating system (physically implemented but logically unused).
By leaving it as , the contents of the storage device before startup will be saved, a dump of this part will be obtained after the system restarts, and the area from which the dump has been obtained will be brought online (logically available). . A method for obtaining a memory dump, which is characterized in that all storage devices can finally be made usable, dumps of all storage devices can be obtained, and system stop time until the system can be restarted can be shortened. 2. The process of dumping a portion of the storage device mentioned above and the process for system startup are linked as a series of steps to specify the range to obtain a partial dump and to specify the storage that should be taken offline during the system startup process. Memory dump acquisition according to claim 1, characterized in that by standardizing the specification of the range of devices, it is possible to prevent the occurrence of mismatches due to illegal specification of both devices, and the occurrence of portions where the dump is not obtained. Method.