Method for fast queue restart after redundant I/O path failover

• US 5,944,838 A
• Filed: 03/31/1997
• Issued: 08/31/1999
• Est. Priority Date: 03/31/1997
• Status: Expired due to Term

• Current Assignee: Avago Technologies General IP PTE Limited (Broadcom, Inc.)
• Original Assignee: LSI Logic Corporation (Broadcom, Inc.)
• Inventors: Jantz, Ray M.
• Primary Examiner(s): Kizou, Hassan
• Assistant Examiner(s): NGUYEN, NGUYEN X
• Application Number:
  US08/828,614
• Time in Patent Office: 883 Days
• Field of Search: 711/114, 340/825.5, 395/401, 395/650, 714/6, 714/7, 714/9
• US Class Current: 714/6.22
• CPC Class Codes: G06F 11/1443 Transmit or communication errors
• CPC Class Codes: G06F 11/201 between storage system components
• CPC Class Codes: G06F 11/2092 Techniques of failing over between control units

Abstract

A redundant storage control module (also referred to as RDAC or multi-active controller) maintains a queue of pending I/O requests sent for processing via a first asynchronously operating I/O path. In the event of failure of the first asynchronously operating I/O path, the controller restarts the entire queue of pending I/O requests to a second I/O path without waiting for each request to individually fail from the first path. Some prior techniques required the RDAC module to await failure of each I/O request sent to the failed first I/O path before restarting each failed request on the secondary I/O path. Such techniques greatly extend the total time required to restart all operations sent to a failed I/O path, by awaiting the failure of all I/O requests previously sent to the first I/O path. Other known techniques provide non-standard features in the lower level driver modules to permit the higher level RDAC modules to directly manipulate dispatch queues maintained for each I/O path within the low level device drivers. Such prior techniques reduce the wait time for restart of failed I/O requests but at the expense of portability of the RDAC modules. The methods and structure of the present invention reduce the wait time for restart of failed I/O requests but in such a manner as to remain portable over a variety of host systems.

Claims

1. A method for restart of failed input/output operations in a system having a redundant controller management component and having a plurality of asynchronously operating I/O paths, the method comprising the steps of:

transferring at least one I/O request from said redundant controller management component to a first I/O path of said asynchronously operating I/O paths wherein said redundant controller management component has no access to said at least one I/O request as manipulated within said first I/O path;

storing in a pending I/O queue associated with said redundant controller management component an entry containing information regarding said at least one I/O request; and

restarting on a second I/O path of said asynchronously operating I/O paths I/O requests having entries in said pending I/O queue previously transferred to said first I/O path in response to receipt of indicia of failure of said first I/O path.

2. The method of claim 1, wherein each of said plurality of asynchronously operating I/O paths corresponds to one of a plurality of redundant multi-active controllers.

3. The method of claim 2, wherein each of said plurality of redundant multi-active controllers is a RAID storage controller.

4. The method of claim 1, wherein each of said asynchronously operating I/O paths comprises an I/O adapter.

5. The method of claim 1 wherein each of said asynchronously operating I/O paths comprises an intelligent I/O adapter.

6. The method of claim 1, wherein said at least one I/O request comprises a plurality of I/O requests and wherein the step of restarting is performed in response to a first said receipt of indicia of failure.

7. The method of claim 6, wherein the step of restarting includes the step of:

marking said entries in said pending I/O queue corresponding to the restarted I/O requests.

8. The method of claim 7, further comprising the step of:

disregarding receipt of indicia of failure received subsequent to said first receipt of indicia and corresponding to marked ones of said entries in said pending I/O queue.

9. The method of claim 1 further comprising the step of:

removing said entries from said pending I/O queue in response to successful completion of processing of said at least one I/O request.

10. A method for restart of failed input/output operations in a system having a redundant controller management component and having a plurality of asynchronously operating I/O paths, the method comprising the steps of:

transferring at least one I/O request from said redundant controller management component to a first I/O path of said asynchronously operating I/O paths wherein said redundant controller management component has no access to said at least one I/O request as manipulated within said first I/O path;

storing in a pending I/O queue associated with said redundant controller management component an entry corresponding to said at least one I/O request transferred to first I/O path;

retrieving, in response to receipt of indicia of failure generated in response to failure of said first I/O path, all entries from said pending I/O queue corresponding to I/O requests previously transferred to said first I/O path; and

sending said all entries to a second I/O path of said asynchronously operating I/O paths.

11. The method of claim 10, wherein each of said plurality of asynchronously operating I/O paths corresponds to one of a plurality of redundant multi-active controllers.

12. The method of claim 11, wherein each of said plurality of redundant multi-active controllers is a RAID storage controller.

13. The method of claim 10, wherein each of said asynchronously operating I/O paths comprises an I/O adapters.

14. The method of claim 10, wherein each of said asynchronously operating I/O paths comprises an intelligent I/O adapter.

15. The method of claim 10, wherein said at least one I/O request comprises a plurality of I/O requests and wherein the steps of retrieving and sending are performed in response to a first receipt of indicia of failure corresponding to a first failed I/O request.

16. The method of claim 15, wherein the step of retrieving includes the step of:

marking each retrieved entry in said pending I/O queue as having been resent to said second I/O path.

17. The method of claim 16, further comprising the step of:

disregarding receipt of indicia of failure received subsequent to said first receipt of indicia and corresponding to marked ones of said entries in said pending I/O queue.

18. The method of claim 10 further comprising the step of:

removing said entry from said pending I/O queue in response to completion of processing of a corresponding one of said at least one I/O request.

19. An apparatus for requeuing failed I/O requests within an I/O subsystem, said apparatus comprising:

a redundant controller management component;

a pending I/O queue associated with said redundant controller management component;

a first asynchronously operating I/O path coupled to said redundant controller management component and including means not accessible by said redundant controller management component for retaining I/O requests transferred from said redundant controller management component;

a second asynchronously operating I/O path coupled to said redundant controller management component and including means not accessible by said redundant controller management component for retaining I/O requests transferred from said redundant controller management component;

means in said redundant controller management component, responsive to receipt of an I/O request, for queueing said I/O request in said pending I/O queue;

means in said redundant controller management component, responsive to receipt of an I/O request, for transferring said I/O request to said first asynchronously operating I/O path;

means in said redundant controller management component for detecting a failure in the operation of said first asynchronously operating I/O path;

means, responsive to detection of a failure in operation of said first asynchronously operating I/O path, for retrieving all I/O requests from said pending I/O queue previously transferred to said first asynchronously operating I/O path; and

means for sending said all I/O requests to said second asynchronously operating I/O path.

20. The apparatus of claim 19, wherein said first asynchronously operating I/O path is a redundant multi-active controller and wherein said second asynchronously operating I/O path is a redundant multi-active controller.

21. The apparatus of claim 20, wherein said first asynchronously operating I/O path is a RAID controller and wherein said second asynchronously operating I/O path is a RAID controller.

22. The apparatus of claim 19, wherein said first asynchronously operating I/O path is an I/O adapter and wherein said second asynchronously operating I/O path is an I/O adapter.

23. The apparatus of claim 19, wherein said first asynchronously operating I/O path is an intelligent I/O adapter and wherein said second asynchronously operating I/O path is an intelligent I/O adapter.

24. The apparatus of claim 19, wherein the means for detecting a failure in operation of said first asynchronously operating I/O path is responsive to receipt of first indicia of failure of a first failed I/O request from said first asynchronously operating I/O path.

25. The apparatus of claim 24, further comprising:

means for discarding subsequent indicia of failure of I/O requests from said first asynchronously operating I/O path wherein said subsequent indicia of failure are received by said means for detecting subsequent to the receipt of said first indicia of failure.

26. A computer readable storage medium tangibly embodying programmed instructions for performing a method for restart of a failed input/output operations operable within a system having an redundant controller management component and having a plurality of asynchronously operating I/O paths, the method comprising the steps of:

transferring at least one I/O request from said redundant controller management component to a first I/O path of said asynchronously operating I/O paths wherein said redundant controller management component has no access to said at least one I/O request as manipulated within said first I/O path;

storing in a pending I/O queue associated with said redundant controller management component an entry containing information regarding said at least one I/O request; and

restarting on a second I/O path of said asynchronously operating I/O paths I/O requests having entries in said pending I/O queue previously transferred to said first I/O path in response to receipt of indicia of failure of said first asynchronously operating I/O path.

27. The storage medium of claim 26, wherein each of said plurality of asynchronously operating I/O paths corresponds to one of a plurality of redundant multi-active controllers.

28. The storage medium of claim 27, wherein each of said plurality of redundant multi-active controllers is a RAID storage controller.

29. The storage medium of claim 26, wherein each of said asynchronously operating I/O paths comprises an I/O adapter.

30. The storage medium of claim 26, wherein each of said asynchronously operating I/O paths comprises an intelligent I/O adapter.

31. The storage medium of claim 26, wherein said at least one I/O request comprises a plurality of I/O requests and wherein the method step of restarting is performed in response to a first said receipt of indicia of failure.

32. The storage medium of claim 31, wherein the method step of restarting includes the step of:

marking said entries in said pending I/O queue corresponding to the restarted I/O requests.

33. The storage medium of claim 32 wherein the method further comprises the step of:

disregarding receipt of indicia of failure received subsequent to said first receipt of indicia and corresponding to marked ones of said entries in said pending I/O queue.

34. The storage medium of claim 26 wherein the method further comprises the step of:

removing said entries from said pending I/O queue in response to successful completion of processing of said at least one I/O request.