Bus Closing Unit for Uncorrectable Fault in Shared Disk Cluster

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

In shared disk type cluster systems, existing methods for stopping access to a shared disk during a failover are inefficient, particularly due to time-consuming processes like unmounting disks, fiber channel port closure, OS panic, and BMC-induced resets, which hinder high-speed failover.

Innovation Solution

A cluster system with a disk input/output unit, a fault detecting unit, and a bus closing unit that generates an uncorrectable fault on the bus to instantly stop access to the shared disk, utilizing a predetermined bus for data input/output operations and employing the ACPI BIOS to intentionally cause a fault and close the bus.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If unmounting disk is used to stop access to shared disk, then access control is achieved, but the process takes time and may fail when writing is ongoing

Engineering Contradiction:

Improveaccess control reliabilityVSAvoidfailover time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The invention extracts the disk access control function from the OS-level unmounting process and implements it at the hardware bus level. By closing the bus using an uncorrectable fault, the system immediately stops all disk access without waiting for OS-level unmounting to complete, thus reducing failover time while maintaining reliable access control.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The invention replaces the software-based OS unmounting mechanism with a hardware-level bus closing mechanism. Instead of relying on the OS to stop disk access, the system uses an uncorrectable fault to physically close the bus, eliminating the time delay and reliability issues associated with OS-level disk unmounting.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Reliability

If closing FC port is used to stop access to shared disk, then access control is achieved, but connection takes time and may fail depending on fault type

Engineering Contradiction:

Improveaccess control reliabilityVSAvoidconnection time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The invention extracts the access control function from the FC port closing mechanism and implements it at the bus level. By using an uncorrectable fault to close the bus, the system achieves immediate access control without the time delay and reliability issues of FC port closing, especially when the server cannot connect to the FC switch due to fault type.

Inventive Principle:

Principle #2Taking out (Extraction)

3Reliability

If OS panic is used to stop access to shared disk, then access control is achieved, but I/O data in HBA cache may be written and failover cannot be performed at high speed

Engineering Contradiction:

Improveaccess control reliabilityVSAvoidfailover speed

Core Design Contradiction:

ReliabilityVSSpeed

Solution Approach 1:

The invention extracts the access control function from the OS panic mechanism and implements it at the hardware bus level. By closing the bus using an uncorrectable fault before OS panic, the system prevents I/O data in HBA cache from being written to the shared disk, enabling high-speed failover while maintaining reliable access control.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The invention applies preliminary anti-action by closing the bus using an uncorrectable fault before allowing OS panic to occur. This preliminary action prevents the harmful effect of I/O data being written to the shared disk during failover, thereby enabling high-speed failover while maintaining data integrity.

Inventive Principle:

Principle #9Preliminary anti-action

4Reliability

If BMC reset or power block is used to stop access to shared disk, then access control is achieved, but the process takes time when BMC is busy and may cause OS panic when BMC is inactive

Engineering Contradiction:

Improveaccess control reliabilityVSAvoidstopping time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The invention extracts the access control function from the BMC-based reset or power blocking mechanisms and implements it at the bus level. By using an uncorrectable fault to close the bus, the system achieves immediate access control without the time delay when BMC is busy, and avoids the complexity of BMC state management while maintaining reliable access control.

Inventive Principle:

Principle #2Taking out (Extraction)

Data Source

PatentUS9501372B2Cluster system including closing a bus using an uncorrectable fault upon a fault detection in an active server

Publication Date: 2016.11.22 NEC CORP

AI summary

The present invention provides a cluster system that promptly stops access to a shared disk upon occurrence of abnormality. The cluster system is a cluster system where an active system server and a standby system server operate utilizing a shared disk. Each server includes: a disk input/output unit that accesses the shared disk by using data that is input and output via a predetermined bus; a fault detecting unit that, when a fault occurs in the active system server, detects the fault; and a bus closing unit that, when the fault detecting unit detects the fault, closes the bus by issuing an uncorrectable fault generation request to cause generation of an uncorrectable fault on the bus.