FAILURE DETECTION METHOD AND FAILURE DETECTION SYSTEM

• US 20130159790A1
• Filed: 11/28/2012
• Published: 06/20/2013
• Est. Priority Date: 12/19/2011
• Status: Active Grant

• Current Assignee: Fujitsu Limited
• Original Assignee: Fujitsu Limited
• Inventors: Watanabe, Yukihiro, Matsumoto, Yasuhide, Iikura, Fumi
• Granted Patent: US 9,141,452 B2
• Time in Patent Office: Days
• US Class Current: 714/48
• CPC Class Codes: G06F 11/006 Identification G06F11/2289 takes precedence
• CPC Class Codes: G06F 11/0751 Error or fault detection not based on redundancy power supply failures G06F1/30; network fault management H04L41/06

Abstract

A failure detection method executed by a system having a plurality of nodes and detecting failures at an information processing device base on a plurality of types of messages output by the information processing device, the method includes collecting messages belonging to a first set, which is a part of the plurality of types, at a first node; collecting messages belonging to a second set different from the first set, which is a part of the plurality of types, at a second node; generating first failure candidate information based on messages of the type belonging to the first set, collected at the first node; generating second failure candidate information based on messages of the type belonging to the second set, collected at the second node; and detecting failure at the information processing device, based on the first failure candidate information and the second failure candidate information.

Claims

1. A failure detection method executed by a system having a plurality of nodes and detecting failures at an information processing device base on a plurality of types of messages output by the information processing device, the method comprising:

collecting messages belonging to a first set, which is a part of the plurality of types, at a first node;

collecting messages belonging to a second set different from the first set, which is a part of the plurality of types, at a second node;

generating first failure candidate information based on messages of the type belonging to the first set, collected at the first node;

generating second failure candidate information based on messages of the type belonging to the second set, collected at the second node; and

detecting failure at the information processing device, based on the first failure candidate information and the second failure candidate information.

2. The failure detection method according to claim 1,

wherein the types of messages collected at the first node and the second node are changed based on the number of times of output of each of the plurality of types of messages from the information processing device.

3. The failure detection method according to claim 2,

wherein information of frequency of appearance of each of the plurality of types of messages is obtained, and

wherein the types of messages collected at the first node and the second node are changed such that the frequency of appearance is equalized.

4. The failure detection method according to claim 3,

wherein the type of messages collected at the first node and the second node are changed in order of being sorted by the frequency of appearance of each of the plurality of types of messages.

5. The failure detection method according to claim 1,

wherein the plurality of types is divided into a plurality of sets of which elements do not overlap, the plurality of sets including the first set and the second set, with a node having been assigned to each of the plurality of types as a collection destination, and

wherein, at the time of collecting messages, a storage unit storing information of numbers corresponding to each of the plurality of types is referred to so as to obtain a numbers corresponding to the plurality of types of messages, and collection destination nodes are determined based on a remainder of having divided the numbers by the number of collection destination nodes.

6. The failure detection method according to claim 5,

wherein the collection destination nodes are determined by whether the remainder is odd.

7. The failure detection method according to claim 1, further comprising:

dividing failure patterns of types of messages corresponding to particular failures into a plurality of partial patterns,

wherein collecting messages of a type belonging to the first set at the first node includes assigning a first partial pattern out of the plurality of partial patterns to the first node,

wherein collecting messages of a type belonging to the second set at the second node includes assigning a second partial pattern out of the plurality of partial patterns to the second node,

wherein generating information of the first failure candidate includes generating information of the first failure candidate in the event that the first partial pattern has been detected from a combination of message types collected at the first node, and

wherein generating information of the second failure candidate includes generating information of the second failure candidate in the event that the second partial pattern has been detected from a combination of message types collected at the second node.

8. The failure detection method according to claim 7,

wherein, in the event that the first partial pattern is not detected from a combination of types of messages collected at the first node, information indicating negation of the particular failure is generated,

wherein, in the event that the second partial pattern is not detected from a combination of types of messages collected at the second node, information indicating negation of the particular failure is generated, and

wherein, in the event that information indicating negation of at least any of the particular failures has been generated, the particular failures are not generated.

9. The failure detection method according to claim 1,

wherein, when collecting messages at the first node and the second node, output point-in-time of the messages is obtained, and

wherein, at the time of accepting input of time range information, information of the first failure candidates and information of the second failure candidates is generated based on the messages collected at the first node and the second node, of which the output point-in-time belongs within the time range.

10. The failure detection method according to claim 1,

wherein the plurality of nodes includes

a first-stage node group including the first node and the second node and including three or more nodes,

an intermediate node group including one or more nodes, and

a final-stage node;

and wherein information of a plurality of failure candidates including the first failure candidate generated by the first-stage node group and the second failure candidate generated by the second-stage node group is narrowed down in stages, and the narrowed down information of failure candidates is collected at the final stage node;

and wherein a failure of the information processing device is detected based on the information of failure candidates collected at the final stage node.

11. A failure detection system to detect failures at an information processing device base on a plurality of types of messages output by the information processing device, the system comprising:

a first node configured to collect messages belonging to a first set, which is a part of the plurality of types, and generate first failure candidate information based on messages of the type belonging to the first set that has been collected;

a second node configured to collect messages belonging to a second set different from the first set, which is a part of the plurality of types, and generate second failure candidate information based on messages of the type belonging to the second set that has been collected; and

a third node configured to detect failure of the information processing device based on the information of the first failure candidates and the information of the second failure candidates, and consolidate information of the first failure candidates and the second failure candidates.