JPH0934852A - Cluster system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To take over the processing at a high speed on the occurrence of a fault of a computer being a component of the cluster system. SOLUTION: Each of computers 10-1-10-3 is provided with a service processor 44 conducting recovery processing and re-configuration for a faulty part in response to the detection of a fault, and the service processor 44 is provided with an interrupt control section 54 receiving a fault from a component in the computer to generate an interrupt for a fault processing routine, and an auxiliary CPU 52 executing the processing by a fault processing routine with the interrupt by the interrupt control section 54, and the auxiliary CPU 52 analyzes a fault in the fault processing routine to discriminate whether or not the consecutive processing due to the fault is disable and informs a notice representing disable consecutive processing by its own computer to other computer when disable consecutive processing is discriminated.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a cluster system in which a plurality of computers are connected to each other, and more particularly to a cluster system for improving availability.

[0002]

2. Description of the Related Art Conventionally, in order to improve system availability and scalability of performance, a cluster system may be constructed by connecting a plurality of computers with each other. In such a cluster system, when one computer cannot continue its processing due to some failure, another computer takes over the interrupted processing to realize high availability of the entire system. In addition, by performing the same type of processing in a distributed manner on a plurality of computers, it is possible to improve the performance according to the number of computers that make up the cluster system.

[0003] For example, D. P. Cox, F.F. Lawlo
r, "Outline of HACMP / 6000 version 2.1" (IBM AIXpert 1994 No. 2)
Describes the high availability and scalability of performance realized by the cluster configuration. In addition,
Japanese Laid-Open Patent Publication No. 217666 discloses a failure detection method in a multiprocessor system that realizes distributed processing configured by connecting a plurality of processors like the cluster system described above.

Further, an example of a computer provided with a dedicated processor for disconnecting a faulty part in response to detection of a fault and performing a reconfiguration process and a restart process is described in J. O. Nichols
on, "The RISC System / 6000
SMP System "(Compcon Proce
edings 1995). Also, T.I.
B. Alexander et. al. , "Corpo
rate Business Servers: An
Alternate to Mainframes
for Business Computing ”
(June 1994 HP Journal)
Another example of performing similar fault handling is shown.

Conventionally, in detecting an abnormality of a computer constituting a cluster system, an operation monitoring signal called "heart beat" is transmitted between computers at regular time intervals, and it is detected that this signal is interrupted. Therefore, it is generally judged that the computer is abnormal.

[0006]

As described above, in the prior art, the operation monitor signal is mutually transmitted at every constant time interval, and it is judged that the computer is abnormal by detecting the interruption of this signal. However, even if the computer itself is not abnormal and the communication path connecting the computers is abnormal, the operation monitoring signal cannot be received, and there is a possibility that the computer is erroneously determined to be abnormal. That is, if there is an abnormality in the communication path between the computers, the computer that has transmitted the operation monitoring signal cannot determine that there is an abnormality in the communication path, and therefore the processing continues. On the other hand, the computer that receives the operation monitoring signal determines that an abnormality has occurred in the signal sending computer because the operation monitoring signal is not received, and then takes over the processing that was being executed on the partner computer. become. In such a case, the two computers operate so as to perform the same processing, but there is a problem that an address conflict problem occurs due to, for example, overlapping LAN network addresses.

In order to solve the above-mentioned first problem,
For example, L. Leaby, "NewAvailable
ity Features of Local Are
aVAXcluster Systems "(Vol.
3 No. 3 1991 DigitaI Techni
"Cal journal" discloses a technique for improving availability by providing a plurality of communication paths among computers that constitute a cluster system, that is, by multiplexing the communication paths. In this way, in the configuration in which the communication paths are multiplexed, if the operation monitoring signal does not arrive within the predetermined time, the receiving computer uses another communication path to check whether the transmitting computer is normal. Check.
As a result of the confirmation, it can be determined whether the sending computer is normal or abnormal. However, it takes time to confirm the status of the sending computer via another communication path after waiting for the arrival of the operation confirmation signal for a predetermined time, and the sending computer actually has an abnormality. If so, there is a problem that it takes time to switch the processing to the other computer.

The present invention solves the above-mentioned conventional problems and realizes improved availability and high reliability by promptly taking over processing by another computer in response to an abnormality in a computer forming a node. It is an object of the present invention to provide a cluster system capable of

[0009]

According to the present invention to achieve the above object, in a cluster system constituted by connecting a plurality of computers, a restoration process or a reconstruction process of a failed portion is performed according to the detection of a failure. A dedicated processor is provided for each computer, and the dedicated processor detects and analyzes a failure of its own computer, determines whether or not the processing cannot be continued due to the failure, and the processing cannot be continued. If it is determined that it is, the notification that the processing by the own computer cannot be continued is notified to the other computer.

In a preferred mode, the dedicated processor performs a recovery process for the failure when it is determined that the process can be continued.

According to the present invention to achieve the above object, in a cluster system constituted by connecting a plurality of computers, the computer executes a dedicated processor for performing restoration processing or reconstruction processing of a failed portion in response to detection of a failure. The dedicated processor includes an interrupt control unit that receives a failure occurrence signal from a component in the computer and generates an interrupt for a failure processing routine, and the failure processing by the interrupt from the interrupt control unit. The auxiliary CPU is provided with an auxiliary CPU that executes a process according to a routine.
In the failure processing routine, when the failure is analyzed, it is determined whether or not the processing cannot be continued due to the failure, and when it is determined that the processing cannot be continued,
It is configured to notify the other computer via the communication path that the processing by the own computer cannot be continued.

Further, in a preferred mode, the auxiliary CPU of the dedicated processor performs a recovery process for the failure when it is determined that the process can be continued. In another preferred aspect, the computers are connected to each other through two communication paths, and the auxiliary CPU of the dedicated processor notifies the notification that the processing by the own computer cannot be continued via the communication path. It is configured to notify other computers.

[0013]

[Operation] In the present invention, when a failure occurs in a computer, the failure information is notified to the interrupt control means of the dedicated processor. In response to the notification of this failure, the interrupt control means causes the auxiliary CPU to generate an interrupt and shifts the processing to the failure processing routine. The auxiliary CPU analyzes the failure factor from the notified failure information and determines whether or not the processing cannot be continued due to the failure. In the case of a failure that can continue the processing, recovery processing for the failure is performed. If the failure is such that processing cannot continue, the dedicated processor notifies the other computers that make up the cluster system via the communication path that "the failure has occurred in the local computer and the processing cannot continue". Notify the notification information of.
The other computer that has received the notification information takes over the process from the computer in which the failure has occurred by a predetermined method.

[0014]

Next, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 is a configuration block diagram showing an overall configuration of a cluster system to which the present invention is applied, and FIG. 2 is a configuration block diagram showing details of a computer constituting a node of the cluster system of FIG.

In FIG. 1, the cluster system 10 is
Node # 1 (computer 10-1), node # 2 (computer 1
0-2) and node # 3 (computer 10-3). These nodes are installed LAN adapter 30
-1 and 30-2, 30-3 and 30-4, 30-5 and 30
Two LANs 20-1 and 20-2 are connected to each other via -6.

The two systems of LANs 20-1 and 20-2 are composed of nodes # 1, # 2, # 3 constituting the cluster system 10.
It is used for exchanging data between devices, exchanging configuration information, and notification of abnormalities when a failure occurs. LAN 20-1, 20-2
Have redundant configurations, but they are not duplicated.

LAN adapters 30-1 and 30-2 connected to the LANs 20-1 and 20-2, respectively, are connected to the nodes # 1, # 2, and # 3 constituting the cluster system 10.
30-3 and 30-4, 30-5 and 30-6 are mounted, respectively. The LAN adapter (A) and the LAN adapter (B) are selectively used by a predetermined method such as round robin.

The configuration of the computers forming the nodes # 1, # 2 and # 3 will be described with reference to FIG. Here, node # 1
The computer 10-1 which comprises is shown. Other nodes #
The same applies to the computers 10-2 and 10-3 forming the second and the third # 3.

The computer 10-1 includes a CPU 41, a memory 42, a bus bridge 43, a service processor 44, a system bus 45, a LAN adapter (A) 30-1 and an LA.
An N adapter (B) 30-2 and an input / output bus 46 are provided.

The CPU 41 can access the memory 42 through the system bus 45, and likewise the system bus 4
It is possible to access various input / output adapters (not shown other than the LAN adapter) connected to the input / output bus 46 via the bus bridge 43 connected to 5. On the input / output bus 46, an adapter for connecting a disk device, a tape device, etc., a communication control adapter, etc. are selectively mounted.

The service processor 44 includes the auxiliary CPU 5
2, an auxiliary memory 53, an interrupt controller 54, a bus bridge 55, and a bus 51 connecting them.
The service processor 44 tells the computer 10-1 that
It is a dedicated processor having functions such as startup processing, configuration control processing, failure processing, and remote diagnosis support. For example, when a failure occurs, the interrupt control unit 54 causes the auxiliary CPU 52 to generate an interrupt based on the failure occurrence signal notified via the signal lines (L1, L2, L3), so that the failure processing is started. Transfer processing to the routine.

Next, the operation of the present embodiment configured as described above will be described. When the cluster system 10 is operating normally, the nodes # 1 to # 3 in FIG. 1 perform data communication between the nodes using either of the two systems of LANs 20-1 and 20-2. For example, in the case of a cluster system that performs database processing, database record information, lock information for updating records, and the like are exchanged between the nodes forming the cluster.

Next, the operation when a failure occurs in a node will be described. For example, it is assumed that a failure occurs in the CPU 41, the memory 42, etc. of the computer 10-1 that constitutes the node # 1. In FIG. 2, the CPU 41 and the memory 4
The fault information of No. 2 is notified to the interrupt control unit 54 of the service processor 44 via the signal line L3. In response to the notification of this failure, the interrupt control unit 54 causes an interrupt to the auxiliary CPU 53 through the bus 51, and shifts the processing to the failure processing routine.

The auxiliary CPU 53 executes failure processing based on the failure processing routine shown in the flowchart of FIG. That is, the auxiliary CPU 53 analyzes the failure factor from the notified failure information (step 301) and determines whether or not the processing cannot be continued due to the failure (step 302). For example, if the type of failure is a failure related to I / O processing and it is possible to recover from the failure by retrying the processing, it is determined that the processing can be continued. Otherwise, the processing continues at this node. Judge impossible. For example, as a failure in the memory 42, ECC
Errors and parity errors are possible. A CRC error or the like can be considered as an input / output processing related failure.

If the fault is such that processing can be continued,
Recovery processing for the failure is performed (step 303).
In this failure recovery process, the failure point is separated and retried.

In the case of a fault in which the processing cannot be continued, the service processor 44 passes the L via the bus bridge 55.
AN adapter 30-1, 30-2, LAN20-1, 2
0-2 notifies the other nodes # 2 and # 3 constituting the cluster system 10 of the notification information that “the failure has occurred in the node # 1 and the processing cannot be continued” (step 304). . Then the service processor 44
Initializes the part other than the service processor of node # 1 (step 305).

Other node # 2 that received the above notification information
In # 3, a node which takes over the process from the node # 1 in which the failure has occurred is determined by a predetermined method, and the process is taken over.

Although the present invention has been described with reference to the preferred embodiments, the present invention is not necessarily limited to the above embodiments. For example, in FIG. 1, a cluster system including three nodes has been described, but two nodes or four nodes are used.
It goes without saying that the present invention can be similarly applied to a cluster system including one or more nodes.

[0029]

As described above, according to the present invention, a dedicated processor for determining whether or not continuation of processing is possible due to a failure is mounted in response to a failure detection of a computer constituting a node. By this, it is possible to continue the processing interrupted to other nodes at high speed, so it is possible to shorten the time from the occurrence of a failure to the takeover of processing, and an operation confirmation signal to indicate that the local node is operating normally. Is sent between nodes instead of being exchanged between nodes, it notifies that there is an error in the local node and cannot continue processing.Therefore, even if there is an error in the communication path, the cluster system can maintain a consistent state and maintain reliability. Can be increased. Therefore, it is possible to improve the availability and high reliability of the cluster system.

[Brief description of drawings]

FIG. 1 is a block diagram illustrating an overall configuration of a cluster system to which the present invention is applied.

FIG. 2 is a block diagram illustrating an internal configuration of a node included in the cluster system of FIG.

FIG. 3 is a flowchart illustrating the processing contents of a failure processing routine by the auxiliary CPU.

[Explanation of symbols]

 10 cluster system 10-1, 10-2, 10-3 computer 20-1, 20-2 LAN 30-1 to 30-6 LAN adapter 41 CPU 42 memory 43 bus bridge 44 service processor 52 auxiliary CPU 53 auxiliary memory 54 interrupt Controller 55 Bus bridge

Claims (5)

[Claims] 1. In a cluster system configured by connecting a plurality of computers, each computer is provided with a dedicated processor that performs restoration processing or reconfiguration processing of a failed portion in response to detection of a failure, and the dedicated processor is Detects and analyzes the failure of the own computer and determines whether the failure prevents the processing from continuing, and when it is determined that the processing cannot continue, A cluster system characterized by notifying to the other computer that continuation is impossible. 2. The cluster system according to claim 1, wherein the dedicated processor performs recovery processing for the failure when it is determined that the processing can be continued. 3. A cluster system configured by connecting a plurality of computers, wherein the computer includes a dedicated processor for performing restoration processing and reconfiguration processing of a failed portion in response to detection of a failure, and the dedicated processor is An interrupt control means for receiving a fault occurrence signal from a component in the computer and generating an interrupt for a fault handling routine; and an auxiliary for executing processing by the fault handling routine by an interrupt from the interrupt control means. A CPU is provided, and the auxiliary CPU analyzes a failure in the failure processing routine to determine whether or not the processing cannot be continued due to the failure, and also determines that the processing cannot be continued. In this case, a notification that the processing by the own computer cannot be continued is notified to the other computer via the communication path. A cluster system characterized by: 4. The cluster system according to claim 3, wherein the auxiliary CPU of the dedicated processor performs recovery processing for the failure when it is determined that the processing can be continued. 5. The computer is connected to each other via two communication paths, and the auxiliary CPU of the dedicated processor sends a notification that the processing by the own computer cannot be continued via the communication path. 4. The computer of claim 3 is notified.
The cluster system described in.