JP2007052536A - Failure detection method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To detect a failure in a switching hub in a blade server which is equipped with a plurality of servers and a plurality of switching hubs. <P>SOLUTION: Signals via a plurality of switching hubs 4 and 5 are transmitted from a server 3 which is detected in failure to a server 2 which detects the failure at a fixed cycle. The server which monitors the failure holds a non-received time management table 8 and determines whether the non-received time of the server 3 and the switching hubs 4 and 5 continue a given period of time or not. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a failure detection method, and more particularly to a failure detection method for detecting a failure of a server or a switching hub.

  Conventionally, Patent Document 1 describes a failure detection system for an exchange switch that always detects a failure quickly regardless of whether a data transmission request is generated in each communication control processing module.

Conventionally, Patent Document 2 discloses a network connection system in which when a failure occurs in a switching hub or cable constituting a network, the operation is switched to a spare network so that the network can be continuously operated without being stopped. Is described.
JP 7-123135 A JP 2004-159205 A

However, conventionally, in a blade server device that communicates with a plurality of server groups via a plurality of switching hubs connecting the server groups, a failure in the switching hub cannot be detected.
An object of the present invention is to detect a failure of a server group and a switching hub included in the blade server device in a blade server device that communicates with a plurality of server groups via a plurality of switching hubs connecting the servers. is there.

In order to solve the above problems, the present invention comprises at least the following means. That is,
(1) A signal is transmitted at regular intervals to servers that detect a failure in a plurality of server groups.
(2) In a server that detects a failure, the time during which signals received from all server groups are interrupted is managed.
(3) In the server that detects the failure, the time during which the signal passing through each switching hub is interrupted is managed.

According to the solution of the present invention,
A plurality of servers; a first switching hub that connects the plurality of servers; and a second switching hub that connects the plurality of servers, the first or second switching hub for communication between the servers In the failure detection method in the blade server device, which is performed via the server, and the server communicates with the external network via the first or second switching hub,
A function in which a server other than the specific server transmits first and second constant periodic signals to the specific server determined in advance via the first and second switching hubs. When,
A function for the specific server to receive first and second constant period signals from the other servers via the first and second switching hubs, respectively;
A function for the specific server to identify a transmission source server and a switching hub via the first and second constant periodic signals;
In the first switching hub, the specific server detects that the first constant period signal from all the other servers has been interrupted for a predetermined time and has received the second constant period signal. The ability to detect the failure that occurred,
In the second switching hub, the specific server detects that the second constant periodic signal from all the other servers has been interrupted for a predetermined time and has received the first constant periodic signal. The ability to detect the failure that occurred,
A fault detection method is provided.

  According to the present invention, since the failure occurrence state of the switching hub can be managed, by switching the communication with the external network that the server was performing in the interface via one switching hub to be routed via the other switching hub It becomes possible to continue communication when a failure occurs in one switching hub.

  In addition, according to the present invention, when a failure occurs in the other switching hub, the server switches the communication with the external network that was performed at the interface via the other switching hub to pass through the one switching hub. In this case, communication can be continued.

  According to the present invention, in a blade server device that communicates with a plurality of server groups via a plurality of switching hubs connecting the servers, it is possible to detect a failure of the server group and the switching hub included in the blade server device. .

Embodiments of the present invention will be described below with reference to the drawings.
FIG. 1 shows a configuration diagram of a server blade apparatus according to the present invention. The blade server device 1 includes a server 2 that detects a failure, a server 3 (3a, 3b, 3c) from which a failure is detected, a two-system switching hub 4 and a switching hub 5, and external interfaces 8 and 9. ing. Although the number of servers in this embodiment is four, it can be freely selected as necessary.

  The blade server device 1 is connected to the external network via the switching hub 4 via the external interface 8 and is connected to the external network via the switching hub 5 via the external interface 9.

  The server 3 in which a failure is detected has a function of transmitting a signal 6 (6a, 6b, 6c) with a fixed period via the switching hub 4 to the server 2 that detects the failure. Further, the server 3 that detects a failure has a function of transmitting a signal 7 (7a, 7b, 7c) having a constant period via the switching hub 5 to the server 2 that detects the failure.

  On the other hand, the server 2 that detects a failure manages a time (non-reception time) after the signal 6 passing through the switching hub 4 or the signal 7 passing through the switching hub 5 is interrupted, that is, a non-reception time management table. 50. In addition, the specific server 2 that detects the failure can instruct the servers 3a to 3c to select the switching hub 4 or 5 via communication with the external network.

Next, FIG. 2 shows a non-reception time management table 50.
The non-reception time management table 50 stores the non-reception time for each of the servers 3a to 3c and the switching hubs 4 and 5. In the following description of the flowchart, as an example, a predetermined measurement start value is set, and the predetermined value is subtracted every predetermined period to correspond to the non-reception time. Data corresponding to the above can be used, and a method of adding a predetermined value can also be used.

  The outline of the operation of the present embodiment is as follows. In other words, signals from the plurality of switching hubs 4 and 5 are transmitted at regular intervals from the server 3 that detects the failure to the server 2 that detects the failure. The server 2 that monitors the failure holds the non-reception time management table 50, and determines whether the non-reception times of the server 3 and the switching hubs 4 and 5 continue for a certain time.

FIG. 3 is a flowchart showing a non-reception time management table setting processing routine.
When receiving a signal, the server 2 that detects a failure activates a non-reception time management table setting processing routine 60 as illustrated (reception of a fixed period signal).

  When the server 2 that detects the failure receives the signal 6a having a constant period, the server 2 identifies that the transit switching hub is the switching hub 4 (61), and sets the measurement start value for the switching hub in the data 52 of the non-reception time management table 50. Setting (62), the signal transmission server is identified as the server 3a (64), and the server measurement start value is set in the data 51a of the unreceived time management table 50 (65).

  Note that the measurement start value can be a predetermined value. Here, as an example, the measurement start value is a positive value corresponding to the number of cycles for detecting the occurrence of a failure. Further, the determination of the switching hub 4 or 5 can be made, for example, by an input port, and the determination of the signal transmission servers 3a to 3c is made from the switching hub 4 or 5 or the transmission source data from the servers 3a to 3c, for example. Can be determined by including it in the fixed period signal and the server 2 receiving it.

  When the server 2 that detects the failure receives the signal 7a with a fixed period, the transit switching hub determination specifies the switching hub 5 (61), and the measurement start value for the switching hub is stored in the data 53 of the unreceived time management table 50. (63), and the signal transmission server is identified as the server 3a (64), and the server measurement start value is set in 51a of the unreceived time management table (65).

  When the server 2 that detects the failure receives the signal 6b having a fixed period, the relay switching hub determination specifies that the switching hub 4 is the switching hub 4 (61), and the switching hub measurement start value is stored in the data 52 of the unreceived time management table 50. (62), the signal transmission server is identified as the server 3b (64), and the server measurement start value is set in the data 51b of the unreceived time management table (66).

  When the server 2 that detects the failure receives the signal 7b having a fixed period, the transit switching hub determination is specified as the switching hub 5 (61), and the measurement start value for the switching hub is stored in the data 53 in the non-reception time management table 50. (63), and the signal transmission server is identified as the server 3b (64), and the server measurement start value is set in the data 51b of the unreceived time management table (66).

  When the server 2 that detects the failure receives the signal 6c with a fixed period, the relay switching hub determination specifies that the switching hub 4 is the switching hub 4 (61), and the switching hub measurement start value is stored in the data 52 of the non-reception time management table 50. (62), the signal transmission server is identified as the server 3c (64), and the server measurement start value is set in the data 51c of the unreceived time management table (67).

  When the server 2 that detects the failure receives the signal 7c with a fixed period, the transit switching hub determination is specified as the switching hub 5 (61), and the measurement start value for the switching hub is stored in the data 53 of the unreception time management table 50. (63), the signal transmission server is identified as the server 3c (64), and the server measurement start value is set in the data 51c of the unreceived time management table (67).

Next, FIG. 4 is a flowchart showing the failure detection processing routine 80.
The server 2 that detects the failure activates the failure detection processing routine 80 by cyclic activation.
The server 2 that detects a failure sequentially selects a device that detects the failure (86), and subtracts 1 or a predetermined number from each set value (data 51a to 51c, 52, 53) of the unreceived time management table 50. (81). The server 2 determines whether or not the unreceived time has been subtracted from the measurement start value to 0 or less (82), and if it has been subtracted to 0 or less, one of the devices (servers 3a to 3c, switching hubs 4 and 5). (Or a plurality of failures) is detected, and the device is described as a failure in an appropriate storage means, and / or notified / output to a predetermined device section (84). If it is greater than 0, no failure is detected (83). The failure detection processing is performed for all of the server device 3 and the switching hub devices 4 and 5 (85), and the failure of each device is detected.

  The non-reception time is set to 0 as the start time, for example, 1 or a predetermined number of additions are performed every time a signal is received in a certain period, and a failure is detected when a predetermined threshold is exceeded. Also good. In addition, the failure can be detected by an appropriate counting method.

  FIG. 5 shows a signal interruption state when a failure occurs in the server 3a. In this case, the signals 6a and 7a are interrupted, and only the unreceived time data 51a is subtracted to 0 or less to detect a failure in the server 3a.

  FIG. 6 shows a signal interruption state when a failure occurs in the server 3b. In this case, the signals 6b and 7b are interrupted, and only the unreceived time data 51b is subtracted to 0 or less to detect a failure in the server 3b.

  FIG. 7 shows a signal interruption state when a failure occurs in the server 3c. In this case, the signals 6c and 7c are interrupted, and only the unreceived time data 51c is subtracted to 0 or less to detect a failure in the server 3c.

  FIG. 8 shows a signal interruption situation when a failure occurs in the switching hub 4. In this case, the signals 6a and 6b and 6c are interrupted, and only the non-reception time data 52 is subtracted to 0 or less, and a failure in the switching hub 4 is detected.

  FIG. 9 shows a signal interruption situation when a failure occurs in the switching hub 5. In this case, the signals 7a, 7b, and 7c are interrupted, and only the unreceived time data 53 is subtracted to 0 or less to detect a failure in the switching hub 5.

  According to the present embodiment, as described above, the server 2 that detects a failure manages the failure of the server 3 and the switching hubs 4 and 5 by managing the signal non-reception time from the server 3 via the switching hub. When the failure of the switching hub 4 or the switching hub 5 occurs, the server 2 that detects the failure switches the switching hub that is in communication with the external network in the server 3 so that the server 3 is connected to the external network. Communication can be continued.

  The present invention can be applied to a blade server device, various communication devices such as various server devices, a computer system, and the like.

The block diagram of the server blade apparatus in this invention. Non-reception time management table 50. The flowchart which shows a non-reception time management table setting process routine. 7 is a flowchart showing a failure detection processing routine 80. The figure which shows the interruption state of the signal when a failure generate | occur | produces in the server 3a. The figure which shows the interruption state of the signal when a failure generate | occur | produces in the server 3b. The figure which shows the interruption state of the signal when a failure generate | occur | produces in the server 3c. The figure which shows the signal interruption condition when a failure generate | occur | produces in the switching hub. The figure which shows the signal interruption condition when a failure generate | occur | produces in the switching hub.

Explanation of symbols

1 blade server device, 2 failure detection server, 3 server, 4 switching hub, 5 switching hub, 50
Unreceived time management table, 60 Unreceived time management table setting processing routine, 80 Failure detection processing routine

Claims (4)

A plurality of servers; a first switching hub that connects the plurality of servers; and a second switching hub that connects the plurality of servers, the first or second switching hub for communication between the servers In the failure detection method in the blade server device, which is performed via the server, and the server communicates with the external network via the first or second switching hub,
A function in which a server other than the specific server transmits first and second constant periodic signals to the specific server determined in advance via the first and second switching hubs. When,
A function for the specific server to receive first and second constant period signals from the other servers via the first and second switching hubs, respectively;
A function for the specific server to identify a transmission source server and a switching hub via the first and second constant periodic signals;
In the first switching hub, the specific server detects that the first constant period signal from all the other servers has been interrupted for a predetermined time and has received the second constant period signal. The ability to detect the failure that occurred,
In the second switching hub, the specific server detects that the second constant periodic signal from all the other servers has been interrupted for a predetermined time and has received the first constant periodic signal. The ability to detect the failure that occurred,
Fault detection method including   The specific server detects a failure occurring in the other server by detecting that both of the first and second constant period signals from any of the other servers are interrupted for a predetermined time. The failure detection method according to claim 1, further comprising a function.   Corresponding to each device of the plurality of servers and the first and second switching hubs, the first or second is based on a table storing information indicating the non-reception time of either the first or second constant period signal. 3. The device according to claim 1, wherein when the fixed period signal of 2 is received, information indicating a non-reception time corresponding to each device is set to detect which device has failed. Failure detection method. The failure detection in the blade server device according to any one of claims 1 to 3, wherein the specific server instructs the other server to select a switching hub via communication with an external network. Method.

Images