CN101651559A - Failover method of storage service in double controller storage system - Google Patents

Abstract

The invention provides a failover method of iSCSI Target storage service in a double controller storage system, and the failover steps are as follows: disconnecting a power supply of a failed controller by an electrical switch; taking over the iSCSI Target nodes and logic storage units of the failed controller; and binding the failed controller ip and carrying out arp broadcasting. The failover ofthe iSCSI target service can ensure continuous availability of the double controller storage system when single controller is failed.

Description

A method for failover of storage services in a dual-controller storage system

technical field

The invention relates to computer network storage technology, in particular to a method for iSCSI Target service failover on a dual-controller ip-san storage system and equipment.

technical background

SCSI (Small Computer System Interface, small computer system interface) is a high-performance computer peripheral device interface. Through this interface, all external devices connected to the PC can realize data transmission and distribution independent of the host through the HBA (Host Bus Adapter, Host Bus Adapter). The main function of the current SCSI protocol is to transfer commands, status and block data between the host computer and the storage device. The SCSI architecture defines the relationship between an initiator (host) and a target (eg, disk, tape) as a client/server exchange. The SCSI-3 application client is on the host, and it describes the I/O requests of high-level applications, file systems, and operating systems. The SCSI-3 device server is in the target device and is responsible for responding to requests. Client/server requests and responses are exchanged over an underlying transport and managed by an appropriate SCSI-3 service delivery protocol, such as FCP or iSCSI over a Gigabit Serial Link.

The iSCSI protocol (Internet Small Computer Systems Interface) is a protocol for mapping SCSI-related architectures through tcp/ip network transmission. The iSCSI initiator (iSCSI Initiator) logs in to the iSCSI target (iSCSI Target) through the tcp/ip network to transmit SCSI commands, Data and status, so that multiple clients can obtain multiple block devices on the same target through the tcp/ip network, mainly disks, and client applications can use these block devices as local block devices to achieve centralized storage .

With the development of network technology, the tcp/ip network bandwidth and quality in applications are getting higher and higher, iSCSI storage performance is improved, and applications are becoming more and more extensive, followed by the reliability of storage services and data security Increased demand. For a storage system that provides iSCSI storage services, if there is only one storage controller, when the storage controller in the application fails, the iSCSI storage application will be interrupted, which may cause service interruption or even data incompleteness. If dual controllers are used to take over services from each other, according to the probability theory, the availability of storage services will be greatly improved, because the possibility of simultaneous failure of dual controllers is much smaller. Cooperating with the disk array RAID, the dual-control storage system will realize the redundant backup function. The key to this system is the way the iSCSI Target storage service switches over in the event of a controller failure.

Contents of the invention

The purpose of the present invention is to overcome the problem of application interruption and data inconsistency in a single-controller storage system when the storage service will be interrupted when the controller fails and the storage service is interrupted, causing the application client to read and write data. The running service monitoring software on the dual-controller system will take over its storage service in a short enough time (within 60 seconds) when it detects that another controller fails (san storage ip-san based on ip network, San storage based on optical fiber switching (fc-san, etc.), so that the data reading and writing process of the application client continues normally after a short interruption.

The method of the present invention is realized in the following manner. When a single controller fails in a dual-control storage system, the iSCSI Target service is switched to another controller. The main structure of the dual-control storage system is that both controllers are connected to The same disk group can access all disks at the same time, as shown in Figure 1. In terms of operating system and software, each controller system is installed with iSCSI Target service software. The characteristics of the iSCSI Target software and service are: it can create more than two iSCSI Target Nodes; The block device acts as a SCSI logical unit LU and assigns the specified logical unit number LUN to the LU; a SCSI ID can be specified for each LU. Among them, SCSI ID is the identification of the logical unit, and a unique value is used to represent the logical unit. The specific implementation method is to specify and save a SCSI ID when adding an LU. When the iSCSI initiator executes the INQUIRY command and sets the EVPD in the INQUIRY command packet to bit and the PAGE CODE value is 0x83h, use the SCSI ID component response packet to respond to the LU identifier. The dual-controller storage system ensures that the configurations on the two controller systems are the same during startup and normal operation. The two controllers are identified as controller 0 and controller 1 respectively. In normal operation mode, the dual-controller system is about The iSCSI Target service configuration is as follows:

(1) Both controllers run the iSCSI Target service and create different iSCSI Target Nodes respectively. One of the differences is that the iSCSI Target Name is different. For the convenience of reference, the iSCSI TargetNode of controller 0 is iSCSITargetNode0, and the iSCSI Target Node of controller 1 is iSCSITargetNode1

(2) Divide the disk or RAID and LV provided by the shared disk group into two groups, and use two iSCSI Target Nodes on the two controllers as SCSI LUs, that is, iSCSI LUs, and assign a dual controller to each LU System-wide unique unit number LUN and SCSI ID. For ease of reference, they are denoted as LunGroup0 and LunGroup1, which are used by iSCSITargetNode0 and iSCSITargetNode1 respectively.

(3) The two controllers use different external network ip. For ease of reference, note that controller 0 uses Ip0, and controller 1 uses Ip1.

(4) Ensure that the iSCSI Initiator application client is connected to the two storage controllers through the network.

The distribution of iSCSI Target service elements during normal operation is shown in Table 1, and the application structure of the storage system is shown in Figure 1.

When the iSCSI Initiator application client is accessing (reading and writing) the LU on a controller, and the hardware or software system of the controller fails, causing the iSCSI data packet of the client to fail to respond, the iSCSI Target service on the faulty controller will be switched to another a controller.

The method of monitoring the running status of the iSCSI Target service during system operation is to first check whether the iSCSI Target service process exists, and if it does not exist, it indicates that the iSCSI Target service is stopped; secondly, try to connect to the iSCSI Target service to monitor the TCP port through the system loopback address 127.0.0.1, if connected Failure indicates that the iSCSI Target service is running abnormally, and then try to establish a session according to the iSCSI protocol and initiate a REPORTLUN request for the LUN list. If multiple attempts fail, it indicates that the iSCSI Target service is running abnormally. Otherwise, the iSCSI Target service is considered to be running normally. When the iSCSI Target service is started, listen to the loopback address 127.0.0.1.

The process of iSCSI Target service switching is as follows (take controller 0 failure as an example):

(1) Application client block read and write requests are blocked, iSCSI Initiator tries to retransmit over the network, if the TCP connection times out, delete the connection, try to rebuild the session until timeout, during which the dual-control storage system will complete the iSCSI Target service takeover.

(2) Controller 1 cuts off the power supply of controller 0 through an electronic switch to completely isolate its access to storage resources and external service response

(3) The running iSCSI Target service of controller 1 creates iSCSI Target node iSCSITarget0, and adds logical unit resource LunGroup0 to iSCSITarget0, and sets the LUN and SCSI ID of each LU according to the original configuration of controller 0.

(4) Controller 1 binds the external ip Ip0 on the original controller A, and broadcasts arp

(5) The application client receives the arp broadcast to update the MAC address of Ip0, and the iSCSI Initiator reconnects to iSCSITarget0 on controller 1 through Ip0, establishes a session, obtains the list information of the iSCSI storage logical unit, and queries the SCSI of each LU through the INQUIRY command ID, since then all suspended iSCSI LUs are available again, and data read and write resumes.

Table 2 shows the distribution of each element after the iSCSI Target service is taken over.

The beneficial effects of the invention are: when a single controller fails in a dual-controller storage system, the iSCSITarget service is switched to another controller, so as to avoid data incompleteness caused by interruption of iSCSI storage application services. Using dual controllers, which can take over services from each other, according to probability theory, the availability of storage services will be greatly improved, because the possibility of simultaneous failure of dual controllers is much smaller. Cooperating with the disk array RAID, the dual-control storage system will realize the redundant backup function.

Description of drawings

Figure 1 is a schematic diagram of a dual-controller failover structure.

Figure 2 is the distribution table of iSCSI Target service elements during normal operation;

Figure 3 is the distribution table of each element after the iSCSI Target service is taken over.

Detailed ways

The switching method of the present invention will be described in detail below with reference to the accompanying drawings.

In a dual-controller storage system, when a single controller fails, the iSCSI Target service is switched to another controller. The structure of the dual-controller storage system is that both controllers are connected to the same disk group and access all disks at the same time. In terms of operating system and software, each controller system is installed with iSCSI Target service software. The characteristics of the iSCSI Target software and service are: the ability to create more than two iSCSI TargetNodes; As a SCSI logical unit LU and assign a specified logical unit number LUN to the LU; a SCSI ID can be specified for each LU, where the SCSI ID is the identification of the logical unit, and a unique value is used to represent the logical unit. Adding an LU When the iSCSI initiator executes the INQUIRY command and sets EVPD in the INQUIRY command packet and the PAGECODE value is 0x83h, the SCSI ID component response packet is used to respond to the LU identifier. The dual-controller storage system starts and In normal operation, ensure that the configurations on the two controller systems are the same, and the two controllers are respectively identified as controller 0 and controller 1. In normal operation mode, the configuration of the iSCSI Target service of the dual-controller system is as follows:

(1) Both controllers run the iSCSI Target service, and create different iSCSI Target Nodes respectively. One of the differences is that the iSCSI Target Name is different. For the convenience of reference, record the iSCSI TargetNode of controller 0 as iSCSI TargetNode0, and the iSCSI Target Node of controller 1 is iSCSITargetNode1;

(2) Divide the disk or RAID and LV provided by the shared disk group into two groups, and use two iSCSI Target Nodes on the two controllers as SCSI LUs, that is, iSCSI LUs, and assign a dual controller to each LU The unique unit number LUN and SCSI ID within the system are recorded as LunGroup0 and LunGroup1 for easy reference, and are used by iSCSITargetNode0 and iSCSITargetNode1 respectively;

(3) The two controllers use different external network ip. For the convenience of reference, remember that controller 0 uses Ip0, and controller 1 uses Ip1;

(4) Ensure that the iSCSI Initiator application client is connected to the two storage controllers through the network.

Dual controllers are connected to the same disk array. In normal operation mode, both controllers run iSCSITarget services, but provide different iSCSI Target Nodes and masters with different disk resources and external ip; both controllers run monitoring software monitoring The controller's iSCSI Target service is running and communicates with each other. When a controller fails, the application client fails to access the iSCSI Target Node provided by the faulty controller. According to the iSCSI protocol and the client data read and write timeout retry mechanism, it will Keep waiting and retrying for a period of time. During this time, another controller will take over the iSCSI Target service of the faulty controller. The two controllers provide iSCSItargetNode0 for controller 0, controller 1, and one controller respectively. Bind IP0, and another controller provides iSCSITargetNode1 to bind IP1. Assume that controller 0 is faulty and controller 1 is in charge. The fault takeover steps are as follows:

(S1) Controller 1 cuts off the power supply of Controller 0 through an electronic switch to completely isolate its access to storage resources and external service response;

(S2) The iSCSI Target service that controller 1 is running creates iSCSITargetNode0, and adds storage resources on the original controller 0 to iSCSITargetNode0 in the form of a logical unit LUN;

(S3) Controller 1 binds the external ip IP0 on the original controller A, and performs arp broadcast;

(S4) The application client reconnects to iSCSITargetNode0 through IP0, reestablishes the session, obtains the iSCSI storage logical unit, and retransmits the data read and write request.

The method to detect whether the iSCSI Target service on the system is running normally is to first check whether the iSCSI Target service process exists, and then simulate the iSCSI Initiator to log in to the iSCSI TargetNode on the controller system and request the report logic unit list REPORTLUN to simulate the application client access. If it is normal, it is considered that the iSCSI Target service on this system is running normally, otherwise it is abnormal.

When the controller takes over, keep the newly added elements consistent with the failed controller. These elements include: iSCSITarget Node Name, storage device list and its corresponding LUN number, SCSI-ID, and external service ip.

Claims (4)

1. A method for failover of iSCSI Target storage services in a dual-controller storage system, characterized in that, when a single controller fails in a dual-controller storage system, the iSCSI Target service is switched to another controller, and the dual-controller storage system switches the iSCSI Target service to another controller. The structure of the control storage system is that two controllers are connected to the same disk group and access all disks at the same time. In terms of operating system and software, each controller system is installed with iSCSI Target service software. The characteristics of the iSCSI Target software and services Yes: can create more than 2 iSCSI TargetNodes; can use ordinary disks, RAID, logical volume LV and other block devices as SCSI logical unit LU and assign a specified logical unit number LUN to the LU; can specify a SCSI ID for each LU , where SCSI ID is the identification of the logical unit, using a unique value to represent the logical unit, specify and save a SCSI ID when adding an LU, when the iSCSI initiator executes the INQUIRY command and sets EVPD in the INQUIRY command packet and PAGECODE When the value is 0x83h, the SCSI ID component response packet is used to respond to the LU identification. The dual-controller storage system ensures that the configurations on the two controller systems are the same during startup and normal operation, and the two controllers are identified as controller 0 and controller 0 respectively. Controller 1, in normal operation mode, the configuration of the iSCSI Target service of the dual-controller system is as follows: (1) Both controllers run the iSCSI Target service, and create different iSCSI Target Nodes respectively. One of the differences is that the iSCSI Target Name is different. For the convenience of reference, record the iSCSI TargetNode of controller 0 as iSCSI TargetNode0, and the iSCSI Target Node of controller 1 is iSCSITargetNode1; (2) Divide the disk or RAID and LV provided by the shared disk group into two groups, and use two iSCSI Target Nodes on the two controllers as SCSI LUs, that is, iSCSI LUs, and assign a dual controller to each LU The unique unit number LUN and SCSI ID within the system are recorded as LunGroup0 and LunGroup1 for easy reference, and are used by iSCSITargetNode0 and iSCSITargetNode1 respectively; (3) The two controllers use different external network ip. For the convenience of reference, remember that controller 0 uses Ip0, and controller 1 uses Ip1; (4) Ensure that the iSCSI Initiator application client is connected to the two storage controllers through the network. 2. The failover method according to claim 1, wherein the dual controllers are connected to the same disk array, and in the normal operation mode, both controllers run iSCSI Target services, but provide different iSCSI Target Nodes Different disk resources and external ip from the main controller; both controllers run monitoring software to monitor the running status of the iSCSI Target service of the controller and communicate with each other. When a controller fails, the application client accesses the iSCSI Target provided by the faulty controller Node failure, according to the iSCSI protocol and the client data read and write timeout retry mechanism, it will maintain a period of waiting and retrying, during this period, another controller will take over the iSCSI Target service of the faulty controller, with The two controllers are controller 0 and controller 1. One controller provides iSCSITargetNode0 to bind IPO, and the other controller provides iSCSITargetNode1 to bind IP1. Assuming that controller 0 is faulty and controller 1 is in charge, the fault takeover steps are as follows: (S1) Controller 1 cuts off the power supply of Controller 0 through an electronic switch to completely isolate its access to storage resources and external service response; (S2) The iSCSI Target service that controller 1 is running creates iSCSITargetNode0, and adds storage resources on the original controller 0 to iSCSITargetNode0 in the form of a logical unit LUN; (S3) Controller 1 binds the external ip IPO on the original controller A, and performs arp broadcast; (S4) The application client reconnects to iSCSITargetNode0 through the IPO, reestablishes the session, obtains the iSCSI storage logic unit, and retransmits the data read and write request. 3. The failover method according to claim 1, wherein the method for detecting whether the iSCSI Target service on the system is running normally is to first check whether the iSCSI Target service process exists, and secondly simulate the iSCSI Initiator to log in to the iSCSI Target on the controller system Node also requests the report logic unit list REPORTLUN to simulate application client access. If both steps are normal, it is considered that the iSCSITarget service on the system is running normally, otherwise it is abnormal. 4. The failover method according to claim 1, characterized in that, when the controller takes over, keep the newly added elements consistent with the failed controller, these elements include: iSCSI Target Node Name, storage device list and its corresponding The LUN number, SCSI-ID, external service ip.

Images