Enhancing reliability and robustness of a cluster

• US 7,047,439 B2
• Filed: 07/12/2004
• Issued: 05/16/2006
• Est. Priority Date: 03/29/2000
• Status: Expired due to Fees

• Current Assignee: Intel Corporation
• Original Assignee: Intel Corporation
• Inventors: Shah, Rajesh R.
• Primary Examiner(s): Beausoliel, Robert
• Assistant Examiner(s): MASKULINSKI, MICHAEL C
• Application Number:
  US10/889,493
• Publication Number:
  US 20050005187A1
• Time in Patent Office: 673 Days
• Field of Search: 714/9
• US Class Current: 714/5.11
• CPC Class Codes: G06F 11/0709 in a distributed system consisting of a plurality of standalone computer nodes, e.g. clusters, client-server systems
• CPC Class Codes: G06F 11/0793 Remedial or corrective actions recovery from an exception in an instruction pipeline G06F9/3861; by retry G06F11/1402; for recovering from a failure of a protocol instance or entity H04L69/40
• CPC Class Codes: G06F 11/1443 Transmit or communication errors
• CPC Class Codes: G06F 11/2005 using redundant communication controllers

Abstract

A cluster having a host connected thereto via a cluster interconnection fabric where a determination is made as to whether an error condition exists in an I/O controller connected to the host via the cluster interconnection fabric. This is achieved by attempting to communicate with the I/O controller a first predetermined time period after an inquiry by an operating system as to whether or not an I/O controller driver stack should be unloaded. Also, the operating system is commanded to unload the I/O controller driver stack upon a determination that the error condition still exists.

Claims

1. A method, comprising:

receiving a request from an operating system (OS) to unload an Input/Output (I/O) controller driver stack; and

delaying a response to the OS after determining the request is associated with a transient communication error between a host and an I/O controller, which is recoverable after a pre-determined period of time.

2. The method of claim 1 further comprising, unloading the I/O controller driver stack after determining the request is not associated with a transient communication error between the host and the I/O controller.

3. The method of claim 2 further comprising, periodically accessing the OS to attempt to automatically re-establish communication between the host and the I/O controller and, if successful, reloading the I/O controller driver stack onto the host.

4. The method of claim 1 further comprising, re-attempting to confirm communications between the host and the I/O controller driver stack after the pre-determined period of time elapses.

5. The method of claim 1 further comprising, accessing the OS to load the I/O controller driver stack into the host before receiving the request from the OS.

6. The method of claim 1 further comprising periodically attempting to re-establish lost communication between the host and the I/O controller at configurable intervals if no communication exists between the host and the I/O controller.

7. The method of claim 1 further comprising instructing the OS to unload the I/O controller driver stack from the host if a non-transient error condition is detected.

8. A system, comprising:

an Input/Output (I/O) controller driver stack; and

a control driver, wherein the control driver monitors communication errors between a host and an I/O controller, and if a communication error is transient, the control driver delays responding to an Operating System (OS) request to unload the I/O controller driver stack from the host for a predetermined period of time, during which if communication is re-established between the host and the I/O controller, the OS request is denied.

9. The system of claim 8, wherein the control driver attempts to re-establish communications between the host and the I/O controller at configurable periods of time if communications between the host and the I/O controller are lost.

10. The system of claim 8, wherein the control driver responds to the OS to unload the I/O controller driver stack if an error between the host and the I/O controller is determined to be non-transient.

11. The system of claim 8 further comprising:

one or more second I/O controller driver stacks; and

one or more second I/O control drivers, wherein each of the one or more second I/O controller driver stacks are associated with a unique one of the one or more second I/O control drivers forming a unique pair, and wherein each unique pair monitors communications between a unique second host and its unique I/O controller.

12. The system of claim 8, wherein the I/O controller driver stack advertises resources available to the host via the I/O controller.

13. The system of claim 12, wherein the I/O controller accesses the resources based on instructions received from applications processing on the host.

14. The system of claim 8, wherein, initially, the I/O controller driver stack is loaded within memory of the host and the host is in communication with the I/O controller.

15. A method, comprising:

receiving from an Operating System (OS) a request to unload an Input/Output (I/O) controller driver stack from memory of a host;

determining the request is associated with a transient error condition associated with a temporarily loss of communications between the host and an I/O controller; and

delaying a response to the OS for at least a predetermined period of time to determine if the loss of communications is resolved and if resolved, denying the request of the OS.

16. The method of claim 15 further including instructing the OS via the response to unload the I/O controller driver stack if the error condition cannot be resolved within the predetermined period of time.

17. The method of claim 16 further including periodically attempting to re-establish communications with the I/O controller at configurable intervals if the I/O controller driver stack is unloaded from the memory of the host.

18. The method of claim 17 further including instructing the OS to re-load the I/O controller driver stack if communications with the I/O controller is successfully re-established.

19. The method of claim 15 further including initially instructing the OS at startup of the machine to load the I/O controller driver stack into the memory of the host.

20. The method of claim 15 further including preventing access of applications processing on the host from accessing resources associated with the I/O controller if the I/O controller driver stack is not loaded in the memory of the host.