JP2830846B2 - Method of controlling storage device array group - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a computer system and, more particularly, to a control method for enhancing the ability of a computer system to connect a central processing unit and an array group for redundantly recording data via a plurality of storage directors to detect a failure. .

[0001]

2. Description of the Related Art Disk spindles for performing redundant recording
In a computer system that connects groups, that is, array groups, even when a failure occurs in a part of the array group, recording / reproduction of user data is performed within a range permitted by the redundancy by utilizing a part free from the failure. Is possible.

In this case, the operation state is referred to as degeneration of an array group, and a failed spindle in the array group is referred to as a degenerated spindle.

When an error occurs in the operation of the array group, the array control device must reliably determine whether a failure has occurred in the path from the storage director to the spindles constituting the array group or has failed. For example, if a path failure is erroneously recognized as a spindle failure, the next time a different array group is accessed, it is often erroneously recognized as a spindle failure again. If a spindle failure is mistakenly recognized as a path failure and an I / O command (hereinafter, I / O) abnormal termination is reported to the OS, the I / O to the alternative path reissued by the OS is similarly mistaken and the I / O Since the OS abnormal termination is reported to the OS, there is a problem that the OS cannot access the array group.

In a conventional array controller, a hardware status obtained during processing of a single I / O or a status obtained from a spindle is used to determine whether a path failure or a spindle failure has occurred. .

[0005]

Since these statuses obtained while processing a single I / O are only the result of execution on a single path that has processed the I / O, a failure occurs. There has been a problem that misidentification of a part is likely to occur.

It is an object of the present invention to increase the resolution of path failures and spindle failures and maximize the excellent availability inherent in array disks.

[0007]

According to the present invention, there is provided a method of controlling an array group, comprising: a set of at least one array group including a plurality of disk spindles (hereinafter, spindles) capable of performing redundant recording; a central processing unit; An array control device including a plurality of storage directors and a shared memory for sharing control information among the storage directors, wherein the computer system connects the central processing unit and the array group via the plurality of storage directors. ,
In the shared memory, as a variable for each of the array groups, a degeneration candidate spindle number indicating the spindle that is likely to degenerate, an abnormal termination path number indicating the storage director used when abnormally terminated, and a normal termination The recording director receives an input / output command issued by the OS and operates the subordinate array group, and the storage director first checks the operation result of the array group. It is classified into a normal end and an abnormal end, and when classified as the abnormal end, if there is a possibility of normal end if the array group is degenerated, it is classified as a suspected spindle operation failure, and otherwise, a non-operation is suspected. A spindle operation failure suspected, and a combination of the classification of the operation result and a variable value of the shared memory before the operation; Based on, performs a predetermined processing, and changing each variable values together.

Further, in the array group control method according to the present invention, the storage director receives an I / O from an OS,
When the operation result indicates that the spindle operation is suspected due to the abnormal end, if the invalid value not corresponding to any storage director is stored in the abnormal end path number or the I / O is received. If the degeneration candidate spindle number is different from the suspected spindle determined to have possibly failed as a result of this operation, the normal termination path number is regarded as an invalid value. Abnormally terminate I / O, and
Requesting an I / O retry on an alternative path, and if the abnormal termination path number indicates the other-system storage director and the degeneration candidate spindle number matches the suspected spindle number, the suspected spindle number is When the abnormal termination path number indicates the own system storage director and the normal termination path number indicates the other system storage director, the I / O is abnormally terminated and the OS terminates. One of the suspected spindles is requested if the route is blocked and a retry on the alternate route is requested, and if the abnormal end path number indicates the local storage director and the normal end path number does not indicate the other storage director. After performing these determinations, the degenerate candidate spindle number or the suspected spindle number is assigned to the own system storage director as an abnormal end path number. Characterized by storing the number.

By the above means, the storage director can use the I / O processing results of its own system and the other system, that is, the I / O processing results of a plurality of paths, for determining whether a path failure or a spindle failure has occurred. In addition, it is possible to prevent erroneous degeneration due to erroneous recognition of a failure location as in the related art.

[0010]

FIG. 1 is a block diagram of a computer system using the present invention.

An operating system (OS) 11 is located in the central processing unit 10 and has two input / output channels 1.
2, 13 connected to each input / output channel 12, 13
An array group 40 or an array group 41 is accessed via the storage directors 20 and 21.

The shared memory 3 for sharing control information between the two storage directors 20 and 21 in the array controller 14.
0 and the array group 40
An area for storing the values of the degeneration candidate spindle number 31, the abnormal end path number 32, and the normal end path number 33 as variables is placed in the area. Similarly, for the other array groups, an area for storing the values of the variables corresponding to the above is provided in the shared memory 30 for each array group.

[0013] Degeneration candidate spindle number 3 which is each variable
In the case of 1, the abnormal end path number 32, and the normal end path number 33, -1 is stored when the variable value is invalid, and an integer of 0 or more is stored when the variable value is valid. In the example of FIG. 1, since the array group 40 is composed of three spindles 41, 42, and 43 for the degeneration candidate spindle number 31, -1 to
An integer of 2 and an integer of -1 to 1 are stored in the abnormal end path number 32 and the normal end path number 33.

The initial values of the variables of the degeneration candidate spindle number 31, abnormal end path number 32, and normal end path number 33 are -1.

FIG. 2 is an I / O processing flow of the storage directors 20 and 21 when the array group 40 is selected. The storage directors 20 and 21 receive I / O commands such as read and write issued by the OS 11 and operate the subordinate array groups (step 101). The storage directors 20 and 21 store the operation results of the array group 40 in the first Are classified into "normal end" and "abnormal end" (step 10).
2) Second, the "abnormal termination" classified in the first is further classified into "suspension of spindle operation failure" and "suspect of non-spindle operation failure" (step 103). In the second category,
When it is recognized that a malfunction of one of the spindles 41, 42, 43 is the cause of the “abnormal termination”, that is, the spindles 41, 42, 43 in the array group 40
If one of them is degraded, the "abnormal end" which has the possibility of "normal end" is classified as "suspension suspected of spindle operation failure", and the others are classified as "suspected non-spindle operation failure".

As shown in FIG. 3, the storage directors 20, 21 store the operation result classification and the degeneration candidate spindle number 31, abnormal end path number 32, and normal end path number 3 obtained above.
Eight types of processing are performed in accordance with the combination of each variable value of No. 3 (entered in the previous value column in FIG. 3), and each variable value is changed according to the processing. In FIG. 3, each variable value after the change is displayed in the column of the after value.

At the time of "normal end", if a valid value is stored in the abnormal end path number 32, the number corresponding to the own storage directors 20, 21 is stored in the normal end path number 33, and the I / O normal The termination is reported to the OS 11 (item number 2).
If no valid value is stored in the abnormal end path number 32, the variable value is not updated (item number 1).

At the time of "abnormal termination" of "suspension of spindle operation failure", the abnormal termination path number 32 is invalid or the abnormal termination path number 32 is changed to the storage director 20 or 21 of another system.
And the degeneration candidate spindle number 31 is different from the suspected spindle number found as a result of this operation,
The normal termination path number 33 is invalidated, the I / O is abnormally terminated, and the I / O retry on the alternative path is requested by the status reported to the OS 11 at the time of abnormal termination (item number 3,
4).

If the abnormal termination path number 32 indicates the remote storage director 20 or 21, and the degeneration candidate spindle number 31 matches the suspected spindle number found as a result of this operation, the suspected spindles 41, 42, 43 Is degraded and an internal retry is performed (item 5). The abnormal termination path number 32 indicates the local storage director 20 or 21, and the normal termination path number 33 indicates the remote storage director 2.
If 0 or 21 is indicated, the I / O is abnormally terminated, the path is closed according to the status reported to the OS 11 at the time of the abnormal termination, and a retry on the alternative path is requested (item number 6).

When the abnormal end path number 32 indicates the local storage director 20 or 21 and the normal end path number 33 does not indicate the other system storage director, one of the suspected spindles 41, 42 and 43 is degenerated. Then, an internal retry is performed (item number 7).

After making these determinations, the suspected spindle number is stored in the degeneration candidate spindle number 31 and the number of the local storage director 20 or 21 is stored in the abnormal termination path number 32 (item numbers 3 to 7).

At the time of "abnormal end" which is not "suspension of spindle operation failure", the variable of the degeneration candidate spindle number 31 is invalidated, and the local storage director 2 is set to the abnormal end path number 32.
The number 0 or 21 is stored (item number 8).

The operation when the first spindle 41 fails will be described below. Each storage director 20 or 21 has I
Upon receiving / O, the array group 40 is operated. The result is "abnormal termination" of "suspension of spindle operation failure". OS 11 assigns first I / O to first storage director 2
In the case of issuing to 0, the first storage director 20 changes the variable values 31, 32, and 33 according to item No. 3 in FIG. 3, abnormally terminates the I / O, and reports the status to the OS 11 upon abnormal termination. Request I / O retry on the alternative route. The OS 11 issues a second I / O (retry I / O) to the second storage director 21 which is an alternative route. The second storage director 21 sets each variable value 3 according to item number 5 in FIG.
1, 32 and 33 are changed, the first spindle 41 is degenerated, the array group 40 after degeneration is operated, and the I / O processing is executed again.

Next, the failure of the first storage director 20 is determined by the first storage director 20.
The operation when the operation error of the spindle 41 is erroneously recognized will be described. OS 11 assigns first I / O to first storage director 2
When issued to 0, the first storage director 20 mistakenly regards the failure of the first storage director 20 as an operation failure of the first spindle 41, and the variable values 31, 3 according to item No. 3 in FIG.
2 and 33, the I / O is abnormally terminated, and the status reported to the OS 11 at the time of the abnormal termination indicates the I / O on the alternative route.
Request / O retry. If the alternative path is available, the OS 11 issues a second I / O (retry I / O) to the second storage director 21 which is the alternative path. Since this I / O can be processed normally, the second storage director 21
The variable values 31, 32, and 33 are changed in accordance with item No. 2 of, and the I / O normal end is reported to the OS 11. Next, when the third I / O is issued to the first storage director 21, the first storage director 21 changes the variables 31, 32, and 33 according to item No. 6 in FIG. Set the status to the recommended status and abnormally terminate the I / O.

The OS 11 responds to the abnormal termination of the first I / O.
If the first storage director 20 cannot use the alternative route, the OS 11 returns the second I / O (retry I / O) to the first storage director 20 again.
Issue to The first storage director 20 changes the variables 31, 32, and 33 according to item number 7 in FIG. 3, degenerates the first spindle 41, operates the array group 40 after degeneration, and restarts the I / O processing. Run.

The item number 1 in FIG.
This indicates that when the I / O processing ends normally when the values 2 and 33 are the initial values, the variable values 31, 32 and 33 are retained.

Item No. 4 in FIG. 3 represents two storage directors 2
If the separate spindles 41, 42, and 43 are candidates for degeneration at 0 and 21, the latest degeneration candidate spindle number is stored in the degeneration candidate spindle number 31.

Item No. 8 in FIG. 3 shows that when the “non-spindle operation failure suspected” occurs, the variable values 31, 32,
33, regardless of the status,
/ O is abnormally terminated.

Although the embodiment of the present invention has been described by taking the case of an array disk as an example, it is apparent that the same method can be applied to a magnetic tape device and other storage devices.

[0030]

The first effect is that erroneous degeneration due to erroneous recognition of a fault location can be prevented.

The reason is that according to the present invention, the result of I / O processing on a composite path can be used as a condition for performing array degeneration.

[Brief description of the drawings]

FIG. 1 is a block diagram of a computer system of the present invention.

FIG. 2 is an I / O processing flow of a storage director.

FIG. 3 is a table showing I / O processing conditions and processing types of a storage director;

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 10 Central processing unit 11 OS 12 and 13 I / O channel 14 Array control device 20 1st storage director 21 2nd storage director 30 Shared memory 31, 61 Reduction candidate spindle number 32, 62 Abnormal termination path number 33, 63 Normal termination path number 40 array group 41,51 first spindle 42,52 second spindle 43,53 third spindle

Claims (5)

(57) [Claims] 1. A system comprising: at least one set of an array group including a plurality of spindles and capable of performing redundant recording; a central processing unit; a plurality of storage directors; and a shared memory for sharing control information among the storage directors. An array control device, comprising: a computer system connecting the central processing unit and the array group via a plurality of the storage directors, wherein the shared memory may degenerate as a variable for each array group. A degenerate candidate spindle number indicating the spindle having the above, an abnormal end path number indicating the storage director used when abnormally ended, and a normal end path number indicating the storage director used when normally ended, The Storage Director uses the operating system (hereinafter
OS) receives an input / output instruction issued by the OS , activates a subordinate array group, and the storage director first classifies the operation result of the array group into a normal end and an abnormal end. If there is a possibility of normal termination if the array group is degenerated
Classified as suspected spindle operation failure, otherwise
Classifying the spindle operation failure suspected, classifying the operation result and the variable value of the shared memory before the operation , that is, the degeneration candidate spindle number and the abnormal end
A method for controlling an array group, comprising: performing a predetermined process based on a combination of a path number and a normal end path number, and changing each variable value. 2. An input issued by an OS from the storage director.
When an output instruction is received, and when the operation result is the normal end, the abnormal termination path number stores a valid value corresponding to the storage director receiving the input / output instruction, that is, the own storage director. 2. The array group control method according to claim 1, wherein a number corresponding to the own storage director is stored in a normal end path number, and the input / output command normal end is reported to the OS. 3. An input issued by an OS from the storage director.
It receives the output instruction, the scan operation results in the abnormal termination
When a pindle operation failure is suspected, an invalid value that does not correspond to any storage director is stored in the abnormal termination path number or other storage other than the storage director that has received the input / output instruction. Suspected spindle indicating the director and said degenerate candidate spindle number was found to have possibly failed as a result of this operation
If the number is different from the number corresponding to, the normal termination path number is set to an invalid value, the input / output instruction is abnormally terminated, and the OS is reported to
Status requests input and output commands retry alternative route to the OS at that, the abnormal termination routing number is shows the other system storage directors, the degenerate candidate spindle number the suspected spin
If you want to match the number that corresponds to the dollar, the suspect spin
If the abnormal termination path number indicates the own storage director and the normal termination path number indicates the other storage director, the I / O instruction is abnormally terminated, and the OS is terminated. Report to
In the status, the OS is requested to close the path and retry on the alternative path, and the abnormal termination path number indicates the own storage director,
If the normal end path number does not indicate the remote storage director, one of the suspected spindles is degenerated and an internal retry is performed. After these determinations, the degenerate candidate spindle number or the suspected spindle number is replaced with the abnormal one. 2. The array group control method according to claim 1, wherein a number corresponding to the own storage director is stored in the end path number. 4. An input issued by the OS from the storage director.
When the output command is received and the operation result is suspected to be the non- spindle operation failure due to the abnormal end, the degeneration candidate spindle number is invalidated, the abnormal end path number is set to the number of the own storage director, and the own storage is set. Stores the director number, abnormally ends the input / output instruction , and reports to the OS.
2. The array group control method according to claim 1, wherein a request is made to the OS to close the path and retry on an alternative path in the status to be notified . 5. An input issued by an OS from the storage director.
When the output command is received and the operation result is suspected to be the non- spindle operation failure due to the abnormal end, the degeneration candidate spindle number is invalidated, the abnormal end path number is set to the number of the own storage director, and the own storage is set. Stores the director number, abnormally ends the input / output instruction , and reports to the OS.
And requesting the OS to close the path and retry on an alternative path in the status to be notified.
3. The method for controlling an array group according to claim 3.