JPH0651914A - Magnetic disk subsystem - Google Patents

Abstract

PURPOSE:To reduce man-hour for retrial and to improve the throughput of an entire system by providing counters for counting the number of times of faults generated on an access passage respectively corresponding to plural memory directors. CONSTITUTION:When a memory director 2 is abnormally finished, a counter 5 counts the number of times of abnormal end and when the value reaches a value set in advance, it is confirmed the other memory director 3 is set in a usable state. Then, fault contents are edited, those contents and the enclosure of the director itself are reported to a host computer 1, and the self-close is performed. When the memory director 2 is newly activated after the self- enclosure, an I/O 6 is activated without by the memory director 3 without any competition with the other director since there is no usable memory director in a cluster 4 excepting for the memory director 3. Since the cluster 4 is closed when the self-enclosure of the memory director 3 is performed, the self-enclosure of the memory director 2 is not performed.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a magnetic disk subsystem, and more particularly to a magnetic disk subsystem having a plurality of magnetic disk devices and a plurality of storage directors provided on a plurality of access paths to the plurality of magnetic disk devices. Regarding

[0002]

2. Description of the Related Art In a conventional magnetic disk subsystem including a plurality of magnetic disk devices and a plurality of storage directors provided on a plurality of access paths to the plurality of magnetic disk devices, a host computer is provided with a magnetic disk device. Access to the magnetic disk device (I /
O) is started by the storage directors available in the same cluster by a plurality of access routes, and the host computer itself can execute I / O by which storage director.
It is a system that does not know whether the activation of O has been executed. At this time, if a failure occurs on the access path or in the storage director, the host computer may retry. At the time of this retry, since the host computer does not care which storage director performs the retry, if the retry is performed by a storage director other than the storage director on the access path where the failure occurred earlier, The retry is likely to end normally. As a result, the storage director having a fault on the access path participates also at the next activation, becomes an abnormal end each time, retries, and succeeds as a result of the retries.

[0003]

As described above, since the conventional magnetic disk subsystem has a plurality of access paths, even if some failure occurs in one access path and the processing ends abnormally, the retry is retried. At this time, if a storage director available on another access path retries and terminates normally, the storage director with the access path failure will be operated as it is, and the next new startup Since all storage directors including the storage director with a path failure compete for processing, the number of trials increases and the host computer is burdened with recovery processing, which reduces the throughput of the entire system. is doing.

[0004]

A magnetic disk subsystem according to the present invention comprises a plurality of magnetic disk devices and a plurality of storage directors provided on a plurality of access paths to the plurality of magnetic disk devices. In the system, a counter is provided for counting the number of failures occurring on the corresponding access path corresponding to each of the plurality of storage directors, and the host computer is provided with a counter when the count value of the counter exceeds a preset value. It was reported and self-obstructed.

[0005]

Embodiments of the present invention will now be described with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of the present invention.

The magnetic disk subsystem of FIG. 1 includes two storage directors 2 and 3, a plurality of clusters 4 having common information between the storage directors 2 and 3, and a plurality of magnetic disk devices (I / O). ) 6 is provided, the host computer 1 can perform the boot process in competition with the plurality of magnetic disk devices 6 by the storage directors 2 and 3. A counter 5 is provided in each of the storage directors 2 and 3. Each counter 5
Is the storage director 2 or 3 to which he belongs and I / O
The number of faults occurring in the storage path 7 between 6 and 6 is counted.

When the host computer 1 issues an instruction to activate the I / O 6 to the cluster 4 having the common information between the storage directors 2 and 3, which storage director 2 or 3 is in conflict with the other storage directors. I / O
It is decided whether or not to start 6 and the host computer 1
Is unaware of which storage director initiated the I / O. When the storage director 2 wins the competition and the storage director 2 activates the I / O 6, as a result of some failure in the storage path 7 from the storage director 2 to the I / O 6 and abnormal termination, the host computer 1 May retry. At the time of this retry, if the storage director 3 wins the conflict and activates the I / O 6, there is a high possibility that it will end normally, but if the storage director 2 wins the conflict and activates the I / O 6, It is likely to terminate abnormally.

When the storage director 2 terminates abnormally, the counter 5 counts the number of abnormal terminations, and when the count value reaches a preset value, it is confirmed that another storage director 3 is available. After confirming and editing the failure content and reporting the failure content and self-blocking to the host computer 1, self-blocking is performed.
After the storage director 2 is self-blocked, when the host computer 1 newly starts up, there is no storage director in the cluster 4 other than the storage director 3 in an available state. I / O6 is started without conflict with other storage directors.

When the storage director 2 terminates abnormally, if the storage director 3 is not in an available state, that is, if the storage director 3 is self-blocking, the cluster 4 is blocked when the storage director 2 self-blocks. Therefore, the storage director 2 does not self-close.

As described above, each storage director is provided with a counter, and when a failure occurs in the storage path from each storage director to the magnetic disk device, the failure occurrence is designated by counting the number of occurrences of the failure with the counter. Storage directors that exceed the specified value can be self-blocked, and the magnetic disk device can be activated by the storage director whose failure has not reached the specified value. This reduces the number of retries by the host computer and improves the overall system throughput. It becomes possible.

Further, by self-blocking the storage director whose failure has exceeded the specified value, there is no need for a process to prevent the host computer from accessing the storage director. Since it does not participate in the startup of, the number of recovery processes that are normally terminated by retrying by another route after abnormal termination is reduced, and the control for that is also reduced.
The load on the host computer can be reduced and the recovery processing time can be shortened.

Furthermore, each storage director reports to the host computer that it is self-blocking, and it is clarified that the storage director is not blocked due to a hardware failure, and the contents of the failure are reported by a caution event report. It is possible to report to the host computer.

[0014]

As described above, in the magnetic disk subsystem of the present invention, each storage director is provided with a counter, and when a failure occurs in the storage path from each storage director to the magnetic disk device, the failure occurs. By counting the number of times with a counter and reporting the fact to the host computer when the number of failure occurrences exceeds a specified value and self-blocking, it reduces the number of man-hours for retry by the host computer, There is an effect that it is possible to improve the throughput of the entire system. In addition, the process of preventing the host computer from accessing the storage director becomes unnecessary, the number of recovery processes to be normally completed by retrying by another route after abnormal termination is reduced, and the control therefor can also be reduced. Therefore, it is possible to reduce the load on the host computer and shorten the recovery processing time. Furthermore,
By reporting to the host computer that each storage director blocks itself, it is clarified that it is not a blockage due to a hardware failure of the storage director, and the contents of the failure are reported to the host computer by a caution event report. There is also an effect that it becomes possible to report.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention.

[Explanation of symbols]

 1 Host Computer 2 Storage Director 3 Storage Director 4 Cluster 5 Counter 6 Magnetic Disk Unit (I / O) 7 Storage Path

Claims (1)

[Claims] 1. A magnetic disk subsystem comprising a plurality of magnetic disk devices, and a plurality of storage directors provided on a plurality of access paths to the plurality of magnetic disk devices, the magnetic disk subsystem corresponding to each of the plurality of storage directors. A counter is provided for counting the number of failures that have occurred on the corresponding access route, and when the count value of the counter exceeds a preset value, it is reported to the host computer to self-block. Magnetic disk subsystem.