JP4830698B2 - Disk controller for performing RAID control using responsible LUN control and diagnostic control method - Google Patents

Description

The present invention relates to a RAID control method in a disk array subsystem, and relates to a control method when a failure occurs in a disk controller that performs responsible LUN control in a storage system having a plurality of disk controllers.
Note) LUN: Logical Unit Number

  FIG. 1 shows a device configuration example of a storage system.

  This system consists of a plurality of host devices (H0 ~) 10a, 10b as main elements, a plurality of disk controllers (C0 ~) 2a, 2b for controlling hard disks, and a controller router (CR) 3 for connecting the disk controllers. And a plurality of RAID groups (RG0˜) 5a, 5b composed of a plurality of hard disks. A plurality of hard disks divided into RAID groups are connected to and controlled by disk routers (C0˜) 2a, 2b in RAID group units by device routers (DR0˜) 4a, 4b. The host device is a general purpose computer or a PC server.

  When a plurality of RAID groups are controlled using a plurality of disk controllers, the connection form of each element can be freely changed physically in order to increase availability. In FIG. 1, only two components are shown for simplicity, but there may be three or more, and there may be connection routes omitted in the figure.

  In order to effectively use control resources of the host device or the like, during normal operation, control (responsible LUN control) is performed for the host device so that a controller in charge is assigned to each RAID group on a one-to-one basis. The connection between the disk controller (C) and the host device (H) is 1: 1, and the connection with the RAID group (RG) is also 1: 1. Each disk controller is connected by a controller router (CR) 3. In FIG. 1, the host device (H0) 10a, disk controller (C0) 2a, device router (DR0) 4a, RAID group (RG0) 5a, host device (H1) 10b, disk controller (C1) 2b, device router (DR1 ) 4b and RAID group (RG1) 5b.

  When each disk controller receives a command specifying a RAID group other than its own RAID group from the host device (H), it transmits it to the other disk controllers via the controller router (CR). Allow control.

  Usually, a disk controller as a master is determined at the time of start-up, etc., a RAID group in charge of each disk controller is determined, and device routers DR0 to DRn are set accordingly. Although this determination method is determined by a predetermined rule, details are omitted because it is not directly related to the present invention.

  Here, it is assumed that a certain disk controller (for example, C0) causes an abnormal operation. The disk controller (C0) 2a includes a diagnosis processing unit (DG) 23, and performs recovery processing such as retry, identification of a faulty part, separation (degeneration) processing, and the like when an abnormality is detected. The result and the abnormal status during operation are transmitted to the host device (H0) 10a, and the process is terminated. If the operation cannot be continued, such as when failure recovery processing cannot be performed or failure cannot be identified, the status is communicated to the host device (H0) 10a, and then self-reset and restart. . Or restart from power-off.

  Since the functions of the disk array subsystem 20 are complicated, abnormal operations with unknown causes may occur rarely due to firmware bugs or combinations of specific operations. Even if recovery processing, degeneration processing, or failure part cannot be identified, it is often recovered by restarting, but at the time of restarting or restarting from power-off, processing such as initial setting and initial diagnosis Take time. Although it is not so long, as shown in FIG. 2, the path from the host device (H0) 10a is cut during that time, and access to the RAID group (RG0) 5a becomes impossible. In a time zone in which access to the system is concentrated, the processing capability of the system can be seriously affected. In addition to this, there are cases where the firmware is replaced by improving the function of the disk controller (C), or the replacement of a failed part that has been in a degenerate operation due to a previous failure. At this time, the entire system is not stopped, but the disk controller (C) must be stopped. During that time, the RAID group (RG) in charge cannot be accessed, which similarly affects the processing capacity of the system. Will be affected.

Multiplexing and redundant configurations are widely used. For disk array management and recovery methods, for example,
There are many known examples as disclosed in Japanese Patent Laid-Open No. 2003-196036. However, a publicly known example from the viewpoint of shortening the period during which the path is cut off at the time of restarting or restarting after power-off is not detected.

  In the disk array subsystem, the path from the host device is cut off or the RAID group cannot be accessed due to a temporary stop or restart due to a disk controller failure or firmware function improvement. The purpose is to prevent things.

  In FIG. 2, the disk controller interface (Channel Adapter) portion is utilized as much as possible.

  The disk controller 2a is coupled to each other by another disk controller 2b in the system and the controller router 3, and the subordinate disk devices are coupled in units of RAID groups by the device router 4a to perform responsible LUN control.

  It also detects and diagnoses errors in the channel adapter unit 21 that controls the interface with the host device 10a and the controller router 3, and each functional unit 22 of the disk controller, and isolates, restores, degenerates, restarts, and restarts And a diagnostic processing unit 23 for performing

  When there is no problem with the function of the channel adapter 21 when restarting or restarting, the diagnosis control unit 23 sets the channel adapter not to take charge of the RAID group and operates separately from other functional units. To do.

  When an abnormality occurs, the diagnosis processing unit 23 of the disk controller notifies the host device 10a and the other disk controller 2b. If there is no abnormality in the channel adapter 21 as a result of the diagnosis process, the diagnosis processing unit 23 separates from the other parts. When restarting or restarting in this state, the channel adapter 21 performs an operation of passing a command or the like from the host device 10a to the controller router 3 through.

  At this time, other disk controllers determine the situation, and one of them operates as a master and takes over. And take over the LUN in charge of the abnormal disk controller. This operation itself is no different from the prior art.

  The disk controller 2a in which an error has occurred can be restarted or restarted (power is turned on again after power is turned off) when failure recovery processing cannot be performed or failure cannot be identified. When the host device 10a accesses the assigned RAID group 5a, the other disk controller 2b taking over takes over and responds. Even in the case of restarting, the channel adapter 21 operates normally when the power is turned on, so the time during which access from the host device 10a is not possible can be shortened.

(Example 1)
FIG. 3 shows an operation flowchart of the diagnosis processing unit when an abnormal state occurs in the disk controller (C) 2.
(S1) Error detection (s2) The diagnosis control unit performs diagnosis to identify the faulty part,
(S3) If recovery is possible, recovery processing such as retry (s31),
(S4) If degeneration is possible, detachment (degeneration) processing (s41) is performed.
(S51) When the operation can be continued, the result and the abnormal status during the operation are transmitted to the host device, and the process is terminated.

If you are unable to continue the operation, such as failure recovery processing, degeneration processing or failure cannot be identified,
(S5) Notify the status to the host device and other disk controllers.
(S7) If there is no abnormality in the channel adapter as a result of diagnosis, the mode is set to a mode (channel adapter separation mode) in which the channel adapter is operated separately from other parts. The channel adapter is set not to take charge of the RAID group.
(S8) Perform a self-reset and restart. Or restart from power-off.

1) Rebooting When the channel controller part is normal but the disk controller C itself cannot be used as it is, such as when the disk controller memory has failed, the disk controller itself is in an abnormal state. After that, only the channel adapter is operated independently.

2) Rebooting after power is turned off If it can be determined that the system cannot be recovered by rebooting, or if it fails to reboot, turn on the power again. When the power is turned on after the disk controller C is turned off, the channel adapter of the disk controller C operates independently.

FIG. 4 shows a processing flowchart in the case of restart or power-on.
(P1) Determine whether the mode is normal mode or channel adapter separation mode when the power is turned on by restart or power cycle. In the channel adapter separation mode, the channel adapter operates independently.

Meanwhile, the diagnostic processing unit (DG) 23
(P2) Make sure that the channel adapter 21 or disk is not operating,
(P3) Perform initial settings and function checks.
(P4) If the result is normal,
(P5) Make sure that the channel adapter 21 or disk is not operating,
(P6) Install the channel adapter and return as the channel controller.

  Since the channel adapter operates independently of the startup process of the disk controller itself (its main part), access from the host is temporarily disabled, but the system can be used as it is without disconnecting the path to the host device. .

  If the disk controller C0 fails, if the channel adapter CA0 starts up, the channel adapter CA0 passes through the command from the host device H0 to the other disk controller C1. Takes over to disk controller C1.

  As a result, although access from the host device H0 is temporarily unavailable, the host device H0 also keeps the time required for processing such as disk controller C0 restart, initial setting at reboot and disk path confirmation. The system can continue the process as it is without cutting the path.

Hardware configuration When disk controller C0 fails Operation flowchart of diagnosis processing unit Restart or power cycle processing

Explanation of symbols

10 Host device 20 Disk subsystem 2, 2a, 2b Disk controller 21 Channel adapter 22 Function unit 23 Diagnostic processing unit 24 Device adapter 3 Controller router 4, 4a, 4b Device router 5, 5a, 5b RAID group

Claims (4)

A disk controller that is coupled to each other by another disk controller in the disk array subsystem and the controller router, and that subordinate disk devices are coupled in RAID group units by the device router, and performs responsible LUN control for the host device.
It has a channel adapter unit that performs interface control with the host device and the controller router, and a diagnostic control unit that detects and diagnoses errors in each functional unit of the disk controller, and performs isolation, restart, and restart.
When there is no problem with the function of the channel adapter unit when restarting or restarting, the diagnostic control unit sets the mode in which the channel adapter and other functional units are operated separately, and restarts or restarts. A disk controller characterized by raising the disk. The disk controller according to claim 1,
When restarting or restarting in a mode in which the channel adapter is disconnected from other functional units,
A disk controller characterized by confirming that the channel adapter is not operating when a functional test other than the channel adapter is normally completed, and setting the normal mode. This is a diagnostic control method for a disk controller that is linked to each other by another disk controller in the disk array subsystem and the controller router, and the subordinate disk devices are linked in RAID group units by the device router, and performs the assigned LUN control for the host device. There,
Detect and diagnose errors in each functional part of the disk controller,
When it is determined that there is no problem in the function of the channel adapter unit that performs interface control with the host device and controller router,
When it is necessary to restart or restart,
Set the mode to operate the channel adapter and other functional parts separately,
A diagnostic control method characterized by restarting or restarting. A diagnostic control method for a disk controller according to claim 3,
When restarting or restarting in a mode in which the channel adapter is disconnected from other functional units,
A diagnostic control method, wherein a function test other than the channel adapter is performed, and if normal, it is confirmed that the channel adapter is not operating, and the normal mode is set.

Images