Netapp, Inc.

Network Appliance, Inc.

Wayne A Booth

Melvin McGee

Feng Wang

Douglas W Coatney

Radek Aster

Brad A Reger

Timothy Vincent Lee

Keith Son

James Lin

John Fabrizio Bevilacqua

Richard I Ely

Technology is disclosed for stack isolation in a storage system including a storage network switch and multiple storage sub-systems (e.g., storage stacks). The storage network switch includes multiple ports and at least one of the ports is configured to dynamically connect to a device that can be either a storage controller device or a storage sub-system. The technology can receive an identification message indicating that a device is connected to a port of a storage network switch, determine based on the identification message at the storage network switch whether the device is a storage controller device or a storage sub-system, and transfer messages between the ports of the storage network switch such that the storage network switch prevents communications between storage sub-systems connected to the storage network switch, but allows communications between the storage sub-systems and storage controller devices connected to the storage network switch.

Stack isolation by a storage network switch

A storage system includes a storage server and a storage subsystem having an I/O module coupled to disks. The storage subsystem receives new firmware from the storage server for an I/O module in the storage subsystem. The storage subsystem stores state data for the I/O module. The storage subsystem reboots a CPU for the I/O module to load the new firmware and determines after reboot that the reboot was for loading the new firmware based on the stored state data. The storage subsystem initializes the I/O module to use the new firmware. The initialization of the I/O module is based on the stored state data, does not reset I/O routing hardware on the I/O module, and is independent of the I/O module processing a request to access data stored on disks coupled to the I/O module.

Non-disruptive firmware upgrade of a storage shelf

A system and method are provided for counting storage-related error events using a sliding window. This is accomplished by counting error events that occur within a sliding window of time and triggering a reaction based on such count. By this feature, the error events are counted with additional accuracy so that a reaction will be appropriately triggered. To this end, in various embodiments, more accurate error counting is afforded to avoid a situation, such as in fixed sampling window frameworks, where an appropriate reaction is not triggered due to a failure to count an appropriate number error events in close proximity.

System and method for counting storage device-related errors utilizing a sliding window

Method and system is provided where PHY state change (PHY CHANGE) notifications from one or more PHYs in a storage infrastructure are monitored as a potential error condition. The rate of PHY CHANGE notifications is monitored to determine if the rate of PHY CHANGE notifications may cause a loss of service or degrade I/O performance. An excessive rate of PHY CHANGE notification that may cause a loss of service is detected by comparing a current PHY CHANGE count with a burst threshold value. The current PHY CHANGE count is also compared to an operational threshold value to detect if the rate of PHY CHANGE notification may result in degradation of overall I/O performance. If the PHY CHANGE count for a PHY equals or exceeds the burst threshold value or the operational threshold value, then the PHY is disabled.

Method and system for minimizing unnecessary topology discovery operations by managing physical layer state change notifications in storage systems

The present invention is a method and a system for dynamic mapping of a fiber channel loop ID in an ALPA loop. Based on reserved address information for the fiber channel system and a number of select ID bits for a slot ID, a dynamic drive mapping table is created. A unique address may be assigned to each drive and each controller in the ALPA loop. The created drive mapping table may be stored on logic decoding circuitry of an adaptor card coupled to each disk drive in the ALPA loop. When fiber channel loop ID signals are sent from a backplane, the fiber channel loop ID signals are translated into seven bits within an ALPA address range based on the dynamic drive mapping table. The converted signals may be sent to the disk drive coupled the adaptor card at boot up time.

Dynamic address mapping of a fibre channel loop ID

A system and method are provided for avoiding categorization of non-error events as actual error events. This is accomplished by categorizing potential error events as actual error events or non-error events based on a proximity to a physical layer signal state change. By this feature, the result of the categorization may be used to more effectively avoid categorizing non-error events and accompanying noise, etc., as actual error events. To this end, in various embodiments, more accurate error counting is afforded to preclude inadvertent disabling, etc. of a mass storage device or associated link that would otherwise result from a flawed error count.

System and method for avoiding categorizing potential link error events as actual link error events based on a proximity to a physical layer signal state change

Method and system is provided where PHY state change (PHY CHANGE) notifications from one or more PHYs in a storage infrastructure are monitored as a potential error condition. The rate of PHY CHANGE notifications is monitored to determine if the rate of PHY CHANGE notifications may cause a loss of service or degrade I/O performance. An excessive rate of PHY CHANGE notification that may cause a loss of service is detected by comparing a current PHY CHANGE count with burst threshold value. The current PHY CHANGE count is also compared to an operational threshold value to detect if the rate of PHY CHANGE notification may result in degradation of overall I/O performance. If the PHY CHANGE count for a PHY equals or exceeds the burst threshold value or the operational threshold value, then the PHY is disabled.

Method and system for minimizing unnecessary topology discovery operations by managing physical layer state change notifcations in storage systems

A method and system for injecting a deterministic fault into storage shelves in a storage subsystem. The method comprises injecting a known fault condition on demand into a hardware component in a storage shelf to cause a failure of the storage shelf. The hardware component incorporates a circuit that is configurable to select between a normal operating condition and a faulty condition of the hardware component. The method further comprises verifying that a reported failure is consistent with the known fault condition.

Method and apparatus for deterministic fault injection of storage shelves in a storage subsystem

A drive shelf includes an embedded switching hub (ESH) that monitors Fibre Channel Arbitrated Loop (FC-AL) traffic, including loop initialization requests (LIRs), coming from drives attached to the ESH. When a number of LIRs issued by a single drive within a certain amount of time exceeds a threshold, the drive is bypassed. One threshold (a 'burst' threshold) measures the number of LIRs issued within a short period of time. If the number exceeds a threshold that would prevent other data from circulating through the loop, the drive is bypassed. A second threshold (a long-term threshold) measures the number of LIRs issued within a longer period of time. If the drive issues a number of LIRs at a slower rate than the burst threshold, but still above the long-term threshold, the drive is bypassed.

Growing community of inventors

Cupertino, CA, United States of America

Wayne A Booth