Method and system for detecting path failure and performing input/output recovery in a multi-path input/output system

1. A computer system having at least one computer connected on a fabric switched network, with at least one adapter on the computer managing paths on the fabric to multiple hardware components connected on the network, comprising:

an operating system resident on the computer programmed to recognize an RSCN indicative of a path failure from a change in a hardware component on the fabric and to issue an input/output (I/O) request to the hardware component on another path in response thereto.

2. The system of claim 1 wherein the operating system is programmed to continue to perform I/O request with the hardware component which caused the RSCN, while attempting to recover the failed path to the hardware component.

3. The system of claim 2 wherein the operating system is programmed for detecting if a non-failed path is available upon receipt of an RSCN from a hardware component, and for selecting a non-failed path, if available, for continuing I/O operations thereon.

4. The system of claim 1 wherein the operating system is programmed for detecting if a non-failed path is available upon receipt of an RSCN from a hardware component, and for selecting a non-failed path, if available, for continuing I/O operations thereon.

5. The system of claim 1 wherein the operating system is programmed for detecting whether an RSCN received is a Fabric RSCN, a Domain RSCN, an Area RSCN, or a Port RSCN.

6. The system of claim 1 wherein the operating system is programmed such that if a Fabric RSCN is received, the I/O request is aborted, a new non-failed path is selected and new input/output request is issued, if necessary, on the new non-failed path, and an attempt is made to repair the failed path.

7. The system of claim 1 wherein the operating system is programmed such that if a Domain RSCN is received, the I/O request is aborted, a new non-failed path is selected and new I/O request is issued if necessary on the new non-failed path, and an attempt is made to repair the failed path.

8. The system of claim 1 wherein the operating system is programmed such that if an Area RSCN is received, the I/O request is aborted, a new non-failed path is selected and new I/O request is issued, if necessary, on the new non-failed path, and an attempt is made to repair the failed path.

9. The system of claim 1 wherein the operating system is programmed such that if a Port RSCN is received, the I/O request is aborted, a new non-failed path is selected and new I/O request is issued, if necessary, on the new non-failed path, and an attempt is made to repair the failed path.

10. A method of continuing issuing of input/output requests on a fabric having a plurality of paths connected to adapters on a computer managing input/output (I/O) requests, on a network which includes the fabric, in the event input/output requests are issued on a failed path, comprising:

detecting an RSCN issued on the fabric;

searching the fabric on which said I/O request was issued for a path failure;

determining if the RSCN detected is a Fabric RSCN, and if it is a Fabric RSCN,

determining the state of all paths from a specified adapter that is managing the I/O request on the failed path;

for a specific path, determining its state, and if the path has failed, recovering the path by attempting repair of the path;

if the repair is successful, updating non-file system structures to represent the repair state;

determining if the path was marked failed in the file system;

if marked failed, repairing the file system path information;

marking the path as capable if I/O requests; and

receiving I/O requests on the path if needed.

11. A method of continuing issuing of input/output requests on a fabric having a plurality of paths connected to adapters on a computer managing input/output requests, on a network which includes the fabric, in the event input/output (I/O) requests are issued on a failed path, comprising:

detecting an RSCN issued on the fabric;

upon an RSCN being detected, determining if it is a Domain RSCN;

determining the state of all paths on the adapter;

managing the I/O request with Domain IDs matching the Domain ID of the Domain RSCN; and

determining the state of one specific path on the adapter with a Domain ID

matching the Domain ID of the Domain RSCN, and if the path has failed,

recovering the path or selecting another path.

12. A method of continuing issuing of input/output requests on a fabric having a plurality of paths connected to adapters on a computer managing input/output requests, on a network which includes the fabric, in the event input/output (I/O) requests are issued on a failed path comprising:

upon an RSCN being detected, determining if it is an Area RSCN;

if it is an Area RSCN, determining the state of all paths on the adapter managing the-I/O request with Area IDs matching the Area ID of the Area RSCN;

determining the state of one specific path on the adapter with an Area ID matching the Area ID of the Area RSCN, and if the path has failed, recovering the path or selecting another path.

13. A method of continuing issuing of input/output requests on a fabric having a plurality of paths connected to adapters on a computer managing input/output requests, on a network which includes the fabric, in the event input/output (I/O) requests are issued on a failed path comprising:

detecting an RSCN on the fabric;

upon an RSCN being detected, determining if it is a Port RSCN;

if it is a Port RSCN, determining the state of all paths on the adapter managing the I/O request with Port IDs matching the Port ID of the Port RSCN;

determining the state of one specific path on the adapter with a Port ID matching the Port ID of the Port RSCN, and if the path has failed, recovering the path or selecting another path.

14. A method of continuing issuing input/output requests on a fabric having a plurality of paths connected to adapters on a computer managing input/output requests, on a network which includes the fabric, in the event input/output (I/O) requests are issued on a failed path, comprising:

detecting an RSCN issued on the fabric;

upon an RSCN being detected, determining if it is a Fabric RSCN, and if it is a Fabric RSCN, determining the state of all paths from a specific adapter that is managing the I/O request on the failed path; and

for a specific path, determining its state, and if the path has failed, one of selecting an alternate non-failed path on which to issue I/O requests or recovering the path.

15. The method of claim 14 , wherein said recovering step comprises recovering the path on which the initial I/O request was issued.

16. The method of claim 14 , wherein said recovering step comprises detecting if a non-failed path is available, and if a non-failed path is available, resume I/O operations on the detected non-failed path.

17. The method of claim 14 , wherein said selecting of an alternate path comprises:

aborting an outstanding I/O request on the failed path;

issuing a status request indicating selection of a non-failed path is needed;

determining if a non-failed path is available; and

if a non-failed path is available, resuming the I/O request on the non-failed path.