Storage device failover

• US 9,766,992 B2
• Filed: 03/31/2014
• Issued: 09/19/2017
• Est. Priority Date: 10/31/2013
• Status: Active Grant

• Current Assignee: Jpmorgan Chase Bank NA As The Agent
• Original Assignee: SanDisk Technologies LLC (Western Digital Corporation)
• Inventors: Wipfel, Robert, Van Assche, Bart, Gupta, Ranjan
• Primary Examiner(s): Bonzo, Bryce
• Assistant Examiner(s): GIBSON, JONATHAN D
• Application Number:
  US14/231,051
• Publication Number:
  US 20150121134A1
• Time in Patent Office: 1,268 Days
• CPC Class Codes: G06F 11/1482 by means of middleware or OS functionality
• CPC Class Codes: G06F 11/2094 Redundant storage or storage space G06F11/2056 takes precedence

Abstract

Techniques are disclosed relating to storage device failover. In one embodiment, a plurality of storage devices are represented as cluster resources to a cluster resource manager that manages cluster resources on a plurality of cluster nodes. An indication may be received that a failover operation is requested with respect to one of the plurality of storage devices. In response to the indication, the cluster resource manager may initiate the failover operation. In some embodiments, the failover operation includes changing a first access state of the storage device and a second access state of another storage device. In such an embodiment, the storage device and the other storage device may be associated with a logical unit number. In some embodiments, the storage device is located within a first of the plurality of cluster nodes; the other storage device is located within a second of the plurality of cluster nodes.

Claims

1. A method, comprising:

representing a plurality of storage devices as resources of a cluster, to a cluster resource manager executing within the cluster, that manages cluster resources on a plurality of cluster nodes, the cluster resource manager executing as a user-level application;

receiving, at a kernel-level storage driver executing within the cluster, a request from a client outside the cluster for a failover operation with respect to one of the plurality of storage devices, the failover operation requested by the client in response to the client detecting that a failure relating to the plurality of storage devices has occurred;

in response to the request, causing the cluster resource manager to initiate the failover operation; and

providing, before the failover operation has been performed, a response to the client indicating that the request has been processed.

2. The method of claim 1, wherein:

the failover operation includes changing a first access state of the storage device and a second access state of another storage device; and

the storage device and the other storage device are associated with a logical unit number.

3. The method of claim 1, further comprising:

prior to an initialization of the cluster resource manager, receiving a request to perform a failover operation; and

denying the request in response to the cluster resource manager not being initialized.

4. The method of claim 1, wherein:

a resource agent enables the cluster resource manager to execute as a user-level application; and

the resource agent communicates with the kernel-level storage driver of one or more of the plurality of storage devices and communicates with the cluster resource manager.

5. The method of claim 1, wherein:

the causing includes adjusting one or more parameters for the cluster resources; and

the one or more parameters are usable by the cluster resource manager to determine whether to fail over a cluster resource from one cluster node to another cluster node.

6. The method of claim 1, wherein:

the plurality of cluster nodes includes three or more cluster nodes; and

the three or more cluster nodes include one or more storage devices within the plurality of storage devices.

7. An apparatus, comprising:

a management module configured to initialize a cluster resource manager for a cluster, the cluster resource manager executing within the cluster as a user-level application, the cluster resource manager configured to make a determination whether to fail over a cluster resource from a first cluster node to a second cluster node, wherein the cluster resource corresponds to a logical disk having a logical unit number; and

a communication module configured to;

use a kernel-level storage driver to receive a first request within the cluster, from a client outside the cluster, to fail over the cluster resource, the request submitted by the client in response to the client detecting that a failure relating to the cluster resource has occurred; and

provide, while the cluster resource manager is being initialized, a response to the client, the response indicating that the first request cannot be processed.

8. The apparatus of claim 7, wherein the provided response causes the client to submit a second, subsequent request to fail over the cluster resource.

9. The apparatus of claim 7, wherein the communication module is further configured to:

determine that a shutdown associated with the cluster resource manager has been initiated; and

prevent processing requests to fail over cluster resources.

10. The apparatus of claim 9, wherein the communication module is further configured to prevent processing the requests by closing a port used to receive requests to fail over cluster resources.

11. The apparatus of claim 7, further comprising:

an agent module configured to communicate an indication of the determination to a small computer system interface (SCSI) core configured to manage one or more storage devices associated with the logical unit number.

12. The apparatus of claim 7, wherein the communication module is further configured to:

receive a second request to fail over another cluster resource; and

provide a response to the client, the response indicating that the second request has been processed before the failover of the other cluster resource has been performed.

13. A non-transitory computer readable medium having program instructions stored thereon, wherein the program instructions are executable by a computing system to cause the computing system to perform operations comprising:

receiving a small computer system interface (SCSI) protocol request associated with a plurality of storage devices, the request received by a kernel-level storage driver for a cluster comprising the plurality of storage devices, the SCSI protocol request issued by a client separate from the cluster in response to the client detecting that a failure relating to the plurality of storage devices has occurred;

adjusting, based on the request, one or more parameters associated with a policy of the computing system, wherein the policy is used to determine a distribution of a plurality of resources across a plurality of computer systems;

sending an indication of the determined distribution to one or more of the plurality of computer systems, wherein the indication is usable for a user-level cluster resource manager to perform a failover corresponding to a logical unit number of one of the plurality of storage devices; and

providing, before the failover has been performed, a response to the client indicating that the request has been processed.

14. The computer readable medium of claim 13, wherein:

the policy includes one or more rules specifying a relationship between storage devices within a first computer system of the plurality of computer systems and storage devices in a second computer system of the plurality of computer systems; and

the adjusting of the one or more parameters causes a change in the distribution according to the one or more rules.

15. The computer readable medium of claim 14, wherein:

the one or more parameters includes a first score of the first computer system; and

the distribution is determined responsive to the first score satisfying a threshold.

16. The computer readable medium of claim 13, wherein the received request is a set target port groups (STPG) command.

17. A non-transitory computer readable medium having program instructions stored thereon, wherein the program instructions are executable by a computing system of a cluster to cause the computing system to perform operations comprising:

receiving, at a kernel-level storage driver, a request to change an access state associated with one or more storage devices, the request received by a cluster comprising the one or more storage devices, the request sent by a client outside of the cluster in response to the client detecting that a failure relating to the one or more storage devices has occurred;

in response to the request, making, at a user-level cluster resource manager, a determination to fail over a cluster resource associated with the one or more storage devices; and

providing, before failing over the cluster resource has been performed, a response to the client indicating that the request has been processed.

18. The computer readable medium of claim 17, wherein the operations further comprise:

conveying an indication of the determination to the kernel-level storage driver, wherein the conveying is via a resource agent of the user-level cluster resource manager.

19. The computer readable medium of claim 17, wherein the cluster resource corresponds to a logical disk associated with the one or more storage devices.

20. The computer readable medium of claim 17, wherein the operations further comprise:

waiting to process requests to change access states until the kernel-level storage driver and the user-level cluster resource manager are executing.