Patent Application
20070174722
1. Field of the Invention
The present invention relates to a technology for detecting an error made by a magnetic disk device at an early stage.
2. Description of the Related Art
Magnetic disk devices are widely used to store data. However, input/output errors can occur in the magnetic disk devices due to a physical damage, vibration, or faulty parts.
When an error occurs, an attempt is made to recovery the data from the magnetic disk device by, for example, reissuing an input/output command to the magnetic disk device. However, because it is common to process the input/output commands based on timeout/retry (reissuance), the recovery process takes a considerable time. An operating system of a computer that employs the error magnetic disk device, or application programs operating on the operating system, are forced to wait until the time-taking recovery process is completed.
Various techniques are know that take care of this issue. For example, Japanese Patent Application Laid-Open No. H06-051914 discloses counting the number of failures occurring in a magnetic disk device and closing the magnetic disk device if the number exceeds a predetermined value. Japanese Patent Application Laid-Open No. 2003-233540 discloses a technology for grasping a retry state concerning disk I/O with a predetermined processing unit.
However, the conventional techniques can take care of only non-fatal (recoverable) errors. In other words, the conventional techniques are of no use when a fatal (unrecoverable) error has occurred.
In recent years, the need to store a large amount of data has increased. To store a large amount of data, storage devices, such as disk array apparatus, having a large number of magnetic disk devices are employed. When an error occurs in one magnetic disk device, that magnetic disk device is closed, and instead a substitute magnetic disk device is used. However, unless a faulty magnetic disk device is detected in short time, the data recovery process takes longer time.
Therefore, there is need for a technology to detect an error in a magnetic disk device fast.
It is an object of the present invention to at least solve the problems in the conventional technology.
According to an aspect of the present invention, an input/output control method for a magnetic disk device that includes a data command queue of data input/output commands and a control command queue of control commands, includes issuing an inquiry command to the control command queue; receiving state information indicative of a state of the magnetic disk device in response to the inquiry command; judging whether the magnetic disk device is in a normal state based on the state information; issuing a data input/output command to the data command queue when it is judged at the judging that the magnetic disk device is in the normal state; and outputting error information indicative of a faulty magnetic disk device when it is judged at the judging that the magnetic disk device is not in the normal state.
According to another aspect of the present invention, an input/output control method for a magnetic disk device that includes a data command queue of data input/output commands and a control command queue of control commands, including data command issuing including issuing a data input/output command to the data input/output command queue; inquiry command issuing including issuing an inquiry command at a predetermined interval from issuance of the data input/output command at the data command issuing; receiving state information indicative of a state of the magnetic disk device in response to each of the inquiry command issued at the inquiry command issuing; judging whether the magnetic disk device is in a normal state based on each of the state information received at the receiving; and outputting error information indicative of a faulty magnetic disk device when it is judged at the judging that the magnetic disk device is not in the normal state.
According to still another aspect of the present invention, a disk control apparatus for performing input/output control on a magnetic disk device that includes a data command queue of data input/output commands and a control command queue of control commands, includes a receiving unit that issues an inquiry command to the control command queue and receives state information indicative of a state of the magnetic disk device in response to the inquiry command; a judging unit that judging whether the magnetic disk device is in a normal state based on the state information received by the receiving unit; an issuing unit that issues a data input/output command to the data command queue when it is judged by the judging unit that the magnetic disk device is in the normal state; and an outputting unit that outputting error information indicative of a faulty magnetic disk device when it is judged by the judging unit that the magnetic disk device is not in the normal state.
According to still another aspect of the present invention, a computer-readable recording medium stores therein a computer program that implements the above method on a computer.
The above and other objects, features, advantages and technical and industrial significance of this invention will be better understood by reading the following detailed description of presently preferred embodiments of the invention, when considered in connection with the accompanying drawings.
Exemplary embodiments of the present invention are explained in detail below with reference to the accompanying drawings.
A case has been shown in
As shown in
The control command queue and the I/O command queue are queues generally provided in almost any existing magnetic disk device. The control command queue and the I/O command queue are realized by firmware operating on the magnetic disk device or by an electronic circuit.
The I/O command queue is a queue that is used by the magnetic disk device to receive a write request (an INPUT command) and a readout request (an OUTPUT command). The magnetic disk device executes an input/output operation for a medium (a disk) by sequentially processing I/O commands stored in the I/O command queue and outputs a response corresponding to each of the I/O commands.
If there are a large number of I/O commands, it takes time to output the response. Conventionally, when there is an unexpected delay in the response, it is judged that a R/W request has failed, and the R/W request is reissued.
However, a fatal error in the magnetic disk device can be detected only after repeatedly issuing an I/O command. During that time, the operating system and the application are forced to wait.
On the other hand, the control command queue is a queue of control commands defined by standards such as AT attachment (ATA) and Small Computer System Interface (SCSI). A control command stored in the control command queue is executed in a shorter time than receiving individual I/O command separately. The control command includes a command for acquiring a device state of the magnetic disk device.
On the contrary, according to the present embodiment, the error detecting module is constituted to issue a command for acquiring a device state of the magnetic disk device to the control command queue (see (1) in
As a result, the operating system and the application can recognize whether there is a fatal error in the magnetic disk device at an early stage so that the operating system and the application need not wait.
In the input/output control method according to the present invention, a device state of the magnetic disk device is acquired using control commands defined by the standards such as ATA and SCSI. Thus, even when a general magnetic disk device is used, it is possible to detect abnormality of the magnetic disk device at an early stage.
A disk control apparatus that implements the input/output control method is explained with reference to
However, the present invention is not limited to this constitution. It is also possible to provide a function equivalent to that of the disk control apparatus 10 on a computer on which the operating system 5 operates or on the magnetic disk device 20. Such a modification is explained below with reference to
As shown in
The I/O request receiving unit 11 performs processing for receiving requests like a write request and a readout request from the operating system 5 and passing the requests received to the state judging unit 13. The I/O response notifying unit 12 performs processing for notifying the operating system 5 of a device state or an I/O response passed from the state judging unit 13 or the I/O response acquiring unit 15.
In this way, the operating system 5 sends an I/O request to the I/O request receiving-unit 11 serving as an interface section and receives an I/O response from the I/O response notifying unit 12. Thus, the operating system 5 only has to perform usual I/O processing without being conscious of other processing executed in the disk control apparatus 10. A device state of the magnetic disk device 20 is included in the I/O response when the device state is notified to the operating system 5.
The state judging unit 13 is a processing unit that issues, when the I/O request is received from the I/O request receiving unit 11, a device state inquiry command to a control command queue (see
Specifically, when the acquired device state is normal, the state judging unit 13 instructs the I/O requesting unit 14 to issue the I/O request received from the I/O request receiving unit 11 to the magnetic disk device 20. On the other hand, when the acquired device state is abnormal, the state judging unit 13 instructs the I/O response notifying unit 12 to notify the operating system 5 that the magnetic disk device 20 is abnormal.
The state judging unit 13 can also repeat the issuance of the device state inquiry command to the magnetic disk device 20 for a plurality of times at each predetermined interval. As described above, the device state inquiry command issued to the control command queue is processed by the magnetic disk device 20 without delay. However, when a temporary failure (a recoverable failure) occurs in the magnetic disk device 20, the magnetic disk device 20 may fail in response to the device state inquiry command issued once. Therefore, if the device state inquiry command is issued a plurality of times, it is possible to distinguish such a temporary failure and a serious failure (an unrecoverable failure) from each other and notify the operating system 5 of the failures.
The I/O requesting unit 14 is a processing unit that performs processing for issuing the I/O request received by the I/O request receiving unit 11 to an I/O command queue (see
A processing procedure in the disk control apparatus 10 shown in
When the state judging unit 13 receives a response indicating that a device state is normal from the magnetic disk device 20 (“Yes” at step S103), the I/O requesting unit 14 issues an I/O command to the I/O command queue of the magnetic disk device 20 (step S104). On the other hand, when the state judging unit 13 does not receive a response indicating that a device state is normal from the magnetic disk device 20 (“No” at step S103), that is, when the state judging unit 13 receives a response indicating that a device state is abnormal or does not receive a response within a predetermined time, the I/O response notifying unit 12 notifies the OS 5 of error information (step S107). The disk control apparatus 10 ends the processing.
Following the processing at step S104, when the I/O response acquiring unit 15 receives an I/O request response from the magnetic disk device 20 (“Yes” at step S105), the I/O response notifying unit 12 notifies the OS 5 of an I/O response (step S106). The disk control apparatus 10 ends the processing. On the other hand, when the I/O response acquiring unit 15 does not receive an I/O response (“No” at step S105), the I/O response acquiring unit 15 repeats the processing at step S105 until the I/O response acquiring unit 15 receives an I/O response.
As shown in
In the flowchart shown in
When the state judging unit 13 receives a response indicating that a device state is normal from the magnetic disk device 20 (“Yes” at step S203), the I/O requesting unit 14 issues an I/O command to the I/O command queue of the magnetic disk device 20 (step S204). On the other hand, when the state judging unit 13 does not receive a response indicating that a device state is normal from the magnetic disk device 20 (“No” at step S203), that is, when the state judging unit 13 receives a response indicating that a device state is abnormal or does not receive a response within a predetermined time, the I/O response notifying unit 12 notifies the OS 5 of error information (step S209). The disk control apparatus 10 ends the processing.
Following the processing at step S204, when the I/O response acquiring unit 15 receives an I/O request response from the magnetic disk device 20 (“Yes” at step S205), the I/O response notifying unit 12 notifies the OS 5 of an I/O response (step S206). The disk control apparatus 10 ends the processing. On the other hand, when the I/O response acquiring unit 15 does not receive an I/O response (“No” at step S205), the state judging unit 13 issues a state check command to the magnetic disk device 20 (step S207) and judges whether a response indicating that a device state is normal is received (step S208).
When the state judging unit 13 receives a response indicating that a device state is normal (“Yes” at step S208), the disk control apparatus 10 proceeds to step S205 and waits for reception of an I/O response. On the other hand, when the state judging unit 13 does not receive a response indicating that a device state is normal from the magnetic disk device 20 (“No” at step S208), that is, when the state judging unit 13 receives a response indicating that a device state is abnormal or does not receive a response within a predetermined time, the I/O response notifying unit 12 notifies the OS 5 of error information (step S209). The disk control apparatus 10 ends the processing.
The disk control apparatus 10 in
Differences between the disk control apparatus 10a and the disk control apparatus 10 shown in
The processing state information 16 is information that is updated by the state judging unit 13 and represents the progress of a state check command issued to the magnetic disk device 20. For example, such processing state information 16 is information including items like the number of times of issuance of a state check command and an elapsed time after the issuance.
The I/O request receiving unit 11 shown in
If the disk control apparatus 10a has the device state response function in this way, the OS 5 is capable of acquiring a device state of the magnetic disk device 20 at desired timing. Therefore, the OS 5 or an application (see
A processing procedure of processing state response processing executed in the disk control apparatus 10a shown in
When an I/O request command corresponding to the processing state request is being processed (“Yes” at step S303), the I/O response notifying unit 12 notifies the OS 5 of a processing state extracted from the processing state information 16 (step S304). The disk control apparatus 10a ends the processing. On the other hand, when an I/O request command corresponding to the processing state request has been processed (“No” at step S303), the I/O response notifying unit 12 notifies the OS 5 that the I/O request command has been processed (step S305). The disk control apparatus 10a ends the processing.
In
In the following explanation of the respective processing units included in the disk control apparatus 10 shown in
As shown in
If the processing units such as the state judging unit 13 are provided in the magnetic disk device 20a in this way, it is possible to adopt an independent interface as an interface between the computer 1 and the magnetic disk device 20a. This makes it easy to provide a command with high processing efficiency and a command with high response speed anew and makes it possible to promptly acquire a failure state of the magnetic disk device 20.
In the explanation with reference to
As shown in
When there is a substitute disk device (“Yes” at step S404), the disk control apparatus 10c accesses this disk device (the substitute disk device) (step S405). On the other hand, when the disk control apparatus 10c does not detect an error of the disk device at step S403 (“No” at step S403), the disk control apparatus 10c accesses the disk device (step S405).
Following step S405, the I/O response notifying unit 12 notifies the OS 5 of an I/O response (step S406). The disk control apparatus 10c ends the processing. On the other hand, when it is judged at step S404 that there is no substitute disk device (“No” at step S404), the I/O response notifying unit 12 notifies the OS 5 of error information (step S407). The disk control apparatus 10c ends the processing.
As described above, in the device control apparatus, the state judging unit, to which an I/O request received by the I/O request receiving unit is passed, acquires a device state by issuing a state check command to a control command queue of the magnetic disk device. When the device state acquired is normal, the I/O requesting unit issues an I/O command to the I/O command queue of the magnetic disk device. When the device state acquired is not normal, the I/O response notifying unit notifies a request source of the I/O request of error information of the device state. This makes it possible to detect an error of the magnetic disk device at an early stage.
According to one aspect, a device state of the magnetic disk device is acquired by issuing a device state inquiry command to the control command queue and judged. On condition that the device state of the magnetic disk device is judged as normal, a data input/output command is issued to the data input/output command queue. On condition that the device state of the magnetic disk device is judged as not normal, an error of the magnetic disk device is informed. Thus, even when an unrecoverable failure occurs in the magnetic disk device, it is possible to detect the error of the magnetic disk device at an early stage.
Moreover, according to another aspect, after the data input/output command is issued, a device state of the magnetic disk device is acquired by issuing a device state inquiry command to the control command queue at each predetermined interval. After the data input/output command is issued, an error of the magnetic disk device is informed on condition that the device state is judged as not normal. Thus, even when the magnetic disk device fails after the issuance of the data input/output command, it is possible to detect the error of the magnetic disk device at an early stage.
Furthermore, according to still another aspect, the device state acquired is informed in response to an inquiry from the outside. Thus, an operating system and an application operating on the operating system can acquire error information of the magnetic disk device at desired timing.
Moreover, according to still another aspect, after a data input/output command is issued to the data input/output command queue, a device state of the magnetic disk device is acquired by issuing a device state inquiry command to the control command queue at each predetermined interval and judged. On condition that the device state of the magnetic disk device is judged as not normal, an error of the magnetic disk device is informed. Thus, even when an unrecoverable failure occurs in the magnetic disk device after the issuance of the data input/output command, it is possible to detect the error of the magnetic disk device at an early stage.
Although the invention has been described with respect to a specific embodiment for a complete and clear disclosure, the appended claims are not to be thus limited but are to be construed as embodying all modifications and alternative constructions that may occur to one skilled in the art that fairly fall within the basic teaching herein set forth.
| Number | Date | Country | Kind |
|---|---|---|---|
| 2005-359476 | Dec 2005 | JP | national |
