JPH1091478A - Agent managing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an agent managing method which can apply a fault adaptation resumption processing without affecting online performance. SOLUTION: An agent transmitted from a terminal 1 programs a management processing content description part. The agent autonomously sets the trace levels of respective places by the content of the management processing content description part, it autonomously returns to the arbitrary place and re-executes the processing for analysis. Thus, a situation in which program in a center is processed for the agent transmitted from the terminal 1 is grasped on a real time basis. When the retrieval agent which requires time for the processing is used, a system operator detects and analyzes the part of a case in short time.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] The present invention relates to an agent, in which a purpose of a caller is transmitted from a terminal to a center, in cooperation with a place in a translation engine of the center,
The present invention relates to an agent management method in an agent type communication system that performs autonomous intelligent information processing.

[0002]

2. Description of the Related Art Conventionally, to acquire log information for analyzing a program, a log acquisition macro is embedded in all target processing programs in a center, a common statistical information collection program is called, and log information is described in a DB. , Log information was edited and managed with an editing tool. The setting of the log acquisition necessity of the managed agent is performed by setting a table for each managed agent by a command, and the necessity of calling the statistical information collection program in each processing program is determined by each processing program. The problem was dealt with by individually modifying (such as removing macros).

[0003] In an agent type communication system,
By taking advantage of the features of the agent and programming the log acquisition function in the agent, the managed agent can hold and move the log to acquire the log, pass it to the management agent, and notify the management agent. I was getting logs.

[0004]

However, there is a problem that it takes a long time in search-type processing, and if a problem occurs during the processing, it is difficult to re-execute the analysis. Also, if you raise the trace level in advance,
There is a problem that the amount of trace information to be held increases, which causes a throughput and a response delay of the entire system. Furthermore, since agents move between places autonomously, it is difficult to identify the place where an abnormality has occurred, and the whole log is stitched together to investigate, so the failure analysis has been prolonged. It is a fact.

SUMMARY OF THE INVENTION An object of the present invention is to provide a failure adaptive restart process capable of autonomously acquiring detailed trace information of an arbitrary level from an arbitrary place in response to a failure in response to the problem of agent behavior management and failure analysis. An object of the present invention is to provide an agent management method applicable without affecting online performance.

[0006]

In order to solve the above-mentioned problems, an agent transmitted from a terminal has a management processing content description section and trace information stored therein. By keeping the processing log at the time, the effective trace level is set at the required place according to the monitoring result, and the synchronization with the agent and the place is synchronized, so that the restart processing for the analysis localized to the failure place Is performed without affecting online performance.

[0007]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below in detail with reference to the drawings.

FIG. 1 shows a configuration of an agent-type communication system showing an example of an embodiment of the present invention. A user transmits an agent 3 which has programmed a request from a terminal 1 to a center 2. The agent 3 enters the translation engine 4 in the center 2, and places 5-1, 5-2,...
Is processed while exchanging information. Also, the management agent in the place notifies the management node 6 of the trace information of the agent at any time. The management agent is activated by the translation engine 4 only when necessary, and is stopped after the processing is completed, so that the CPU and the memory can be effectively used.

FIG. 2 shows a program configuration held by the agent, which comprises a request content description section 3a, a management processing content description section 3b, and a management trace information storage section 3c. The agent that does not require monitoring does not hold any program other than the request content description unit 3a.

FIG. 3 shows a flow of processing of the agent in the center. The agent transmitted from the terminal 1 enters the place 5-1 in the translation engine 4. If monitoring is required for analysis during the process,
Receiving the management processing content description section and the management trace information storage section from the management agent existing in each place,
Start monitoring.

When an agent requiring monitoring enters the place 5-1 in the translation engine 4, the management trace information storage unit 3c currently held is transferred to the management agent of the place 5-1. It updates the counter value and notifies the management node 6 of the received management trace information. The agent and the management agent are assigned to each of the places 5-1, 5-2, 5-
The same processing is performed in 3, 5-4, 5-5, and 5-6, and the agent itself returns to the terminal 1 with the trace result.

In the normal processing, monitoring can be realized according to the flow shown in FIG. 3, but when a failure occurs, the processing is temporarily terminated, and
After confirming the trace information and estimating the failure location, the reproduction test is performed again, and the analysis is prolonged. As shown in FIG. 4, the agent-type communication system uses an agent-specific function to request the existing agent 21 in the place 11 for the next processing, or to cause the translation engine to execute the processing in the place 12 in response to the occurrence of an event. To create a new agent 22,
When sending the same agent to a plurality of places such as searching, the replication agents 23 and 24
There is a feature that creates. Hereinafter, the features of the present invention will be described in consideration of these.

FIG. 5 shows an example of an item for which the management agent recognizes a failure. Each time the agent 3 moves to each place, the agent 3 stores it in the management trace information storage section 3c.
Error information and the like corresponding to these items are stored.
The management agent of each place recognizes from the error information that a failure has occurred in the agent or the place.

FIG. 6 shows a flow of processing for resuming the analysis when the agent itself has failed.

In FIG. 6, when an agent 41 stops in place 31 due to a failure or the like, the place 3
The management agent 42 resident in 1 detects an abnormality because there is no transfer of trace information from the agent 41, and notifies the management node 6 of a failure. Further, the agent 41 which cannot be moved to the next processing due to a failure or the like due to the function of the agent is temporarily stored in the stopped agent storage place 32.

The management agent 43 of the stop agent storage place 32 sends the designated place 33 to be restarted.
Receiving the trace information 3b and the processing content information 3c relating to the stop agent 41 from the management agent 44 of
The state of the stop agent 41 is returned to the state of the designated place 33, the trace level is reset, and a return flag indicating that the resumption processing for breaking is turned on,
The stop agent 41 'is transmitted to the designated place 33. The stop agent 41 'sent back receives the normal processing and moves from the designated place 33 while increasing the trace level of each place.

FIG. 7 shows a flow of processing for resuming analysis when a place has an abnormality.

In FIG. 7, when the place 51 is stopped due to a failure or the like, the agent 61 entering the failure place 51 receives an abnormality notification of the place 51 from the management agent 62 of the failure place 51. At this time,
If the monitoring is off, the agent 61 is set to on, sets the trace level, turns on a return flag indicating restart processing for analysis, and is transmitted to the designated place 52.

At this time, the agent 61 returns the state of the agent 61 to the state of the designated place 52 from the management trace information carried by the agent 61, and when the agent of the return flag is turned on at the designated place 52, the agent 61 By matching the state of the designated place 52 with the state of the agent 61 based on the interaction information of the processing performed with the agent 61, the restart processing for analysis is enabled.

The returned stop agent 61 'is
The user moves from the designated place 52 by receiving normal processing while increasing the trace level of each place. Designated place 5
When the agent 61 ′ sent back to the second place enters the failed place 51 again, the return flag for analysis of the management processing content description section 3 b held by the agent is set to off, and the agent Processed as usual.

Next, instead of the normal mode in which a single agent moves and performs processing, a restart processing when a place stops due to a failure or the like while performing processing specific to an agent as shown in FIG. Is described.

FIG. 8 shows a flow of processing for resuming analysis when a new agent generation processing is being performed.

In FIG. 8, an agent 81 enters a place 71, and a new agent 8
2 and requesting the new agent 82 to work, if no failure occurs in the place 71, the management processing content description section and the management trace information storage section held by the request agent 81 are transferred to the new agent 82. Delegate acquisition of trace information for subsequent processing.

However, when the place 71 creating the new agent 82 is confirmed to be abnormal due to a failure or the like, the analysis restart processing is performed for the created agent 8.
2, the agent 81 of the creation source returns to the designated place, and executes the analysis resumption processing.

At this time, the created new agent 82
Is stopped, and the subsequent processing is not performed. When the agent for the resumption processing for analysis reenters the place 71, a new agent is created again, the return flag for analysis is set to off, and the subsequent acquisition of trace information, etc. Take over the process to a new agent,
Subsequent processing is performed by the new agent.

FIG. 9 shows a flow of processing for resuming analysis in the case where branch processing by duplication is being performed.

In FIG. 9, the agent 84 enters the place 72 and searches for information.
When a plurality of agents 85 and 86 are copied and branched for processing, if the place 72 does not fail, the management processing content description section and the management trace information storage section held by the agent 84 are copied. Agent 8
5 and 86, and the acquisition of trace information accompanying the subsequent processing is delegated to each of the replication agents 85 and 86.

However, when the place 72 where the agent is duplicated is confirmed to be abnormal due to a failure or the like, the resuming process for analysis is performed by the duplicated agent 85 or 86.
Instead, the duplication source agent 84 returns to the designated place and executes the analysis restart processing.

At this time, the plurality of copied agents 8
5, 86 are stopped at the place 72, and the subsequent processing is not performed. Further, when the replication source agent 84 for performing the resumption processing for analysis enters the place 72 again, the plurality of agents are duplicated again, the return flag for analysis is set to off, and the subsequent trace information is set. The duplication agent takes over the processing such as the acquisition of the data, and the duplication agent executes the subsequent processing.

FIG. 10 shows a flow of a process of resuming analysis when a meet process with an existing agent is being performed.

In FIG. 10, an agent 88 performs an information exchange meet with an agent 89 already existing in the place 73 to perform a process, and requests the existing agent 89 for a subsequent process. Agent 88 places 7
3, when exchanging information with the existing agent 89, if a failure does not occur in the place 73, the management processing content description section and the management trace information storage section held by the agent 88 are transferred to the existing agent 89. Acquisition of trace information accompanying the processing of (1) is delegated to the existing agent 89.

However, when the place 73 where the agent is exchanging the information of the processing request is confirmed to be abnormal due to a failure or the like, the resuming process for analysis is not performed by the existing agent 89 which exchanged the information. The agent 88 that has entered the place 73 returns to the designated place, and executes analysis resumption processing.

At this time, the existing agent 89 that has received the work request by exchanging information with the agent 88 stays in the place 73, does not perform any subsequent processing, and the management agent 90 uses the trace information to change the state of the existing agent 89. To the time when the agent 88 started processing at the place 73, and the place 73 is returned by the translation engine to the state at the time when the agent 88 started processing at the place 73 using the variation journal information of the place 73. .

When the resuming process for the analysis is completed and the agent 88 reenters the place 73,
Information is exchanged again with the existing agent 89, the return flag for analysis is set to off, and subsequent processing such as acquisition of trace information is taken over by the existing agent 89, and the subsequent processing is executed by the existing agent 89.

Next, the details of the method of returning the state of the agent and the place will be described.

FIG. 11 shows a flow of a restart process when returning the state of all agents and all places in the translation engine.

In FIG. 11, the translation engine 4 stores the state variation journal information 91 in a database in the translation engine at regular intervals. After the management agent 111 of the failure discovery place 101 detects the failure,
The trace information and the time stamp 112a are received from the agent 112 receiving the monitoring of No. 1 and notified to the management node 6.

The management node 6 transmits a temporary stop agent 113 to the translation engine 4 and each place in order to temporarily stop the agent entering the translation engine 4. Time stamp 11 indicating restart processing start time
According to the instruction of the suspension agent 113 including 3a,
All places return the state together with the entire translation engine using the variation log of the translation engine 4 until the state at the designated time.

All the agents in the translation engine 4 use the trace information held by the translation engine 4 to perform the processing at the specified time state and at the specified time.
Then, the process returns to step 3 and the restart processing for analysis is performed in a state where the entire synchronization has been performed from that point.

FIG. 12 shows a flow of restart processing when only the states of agents and places involved in a failure in the translation engine are returned.

In FIG. 12, when the management agent 132 detects a failure based on the trace information of the agent 131 and returns to the place 123 designated to acquire more detailed trace information, the agent 1
Using the trace information and the time stamp held by 31, places 122, 12 in reverse order to the target place
Move 4,123. At this time, based on the processing time and the agent ID in each place, in each place, the state of the processing contents and the like with the corresponding agent is partially returned to the designated time from the variation journal information 141 and 142 held in the place. To go.

Then, when the agent 131 goes to the target designated place 123 and the state of each place to be processed is set to a state before the agent 131 enters, the restart processing for analysis is partially started. .

In this case, since the process is a partial restart process,
There is no need to suspend the translation engine, and agents unrelated to other failures can continue processing as usual.
Can send a new agent.

[0044]

As described above, according to the present invention,
In an autonomous agent-type system in which agents embedding the purpose of each individual are processed between various places in the center and produce results, trace information for analysis is used for quick investigation of failure points and failure analysis in an autonomous agent type system Acquisition processing is performed in response to a failure, the agent itself identifies the reproduction place by interaction with the management agent, and by performing the resumption processing focused on that, by collecting a large amount of logs constantly online It can localize the decrease in throughput and response performance,
Also, in cases where it takes a long time to process results such as a search-type agent and it is difficult to repeat the test in the event of a failure, trace information acquisition is resumed during normal processing. By doing, efficient failure analysis becomes possible.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of an agent-type communication system showing an example of an embodiment of the present invention;

FIG. 2 is a configuration diagram of a program held by an agent.

FIG. 3 is a diagram showing a flow of processing of an agent in the center.

FIG. 4 is a diagram showing functions specific to an agent.

FIG. 5 is a diagram showing an example of an item for recognizing a failure;

FIG. 6 is a diagram showing a flow of processing for resuming analysis when an agent itself has failed;

FIG. 7 is a diagram showing a flow of processing for resuming analysis when a place has an abnormality;

FIG. 8 is a diagram illustrating a flow of a process of resuming analysis at the time of a place failure when a new agent generation process is being performed;

FIG. 9 is a diagram showing a flow of a process of resuming analysis at the time of a place failure when branch processing by replication is being performed;

FIG. 10 is a diagram showing a flow of a process for resuming analysis at the time of a place failure when a meet process is performed with an existing agent;

FIG. 11 is a diagram showing a restart processing flow when returning the state of all agents and all places in the translation engine.

FIG. 12 is a diagram showing a restart processing flow when only the states of agents and places involved in a failure in the translation engine are returned.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Terminal, 2 ... Center, 3 ... Agent, 4 ... Translation engine, 5-1-5-6 ... Place, 6 ... Management node.

Claims (5)

[Claims] A terminal transfers an agent containing a program and data to a center having a translation engine for interpreting and processing the agent, and links the agent program with a place in the translation engine that contains the program and data. While executing autonomously, the agent capable of programming the management processing content description section holds the trace information including the processing place ID, the counter value indicating the processing order of the place, the time stamp, and the interaction information with the place. In the agent management method that manages the agent by using
There is always a management agent that receives the trace information of the agent and notifies the management node or notifies the terminal of the return to the terminal. Based on the monitoring result of the agent, the management agent stops the agent to be processed, stops a place, executes an illegal process, etc. When an error occurs, an error notification is sent to the agent, and the agent returns to an arbitrary place specified in the trace information held therein, and performs the monitoring of the place specified in the management processing content description section for more detailed monitoring. An agent management method, comprising resetting a trace level and restarting processing from the place. 2. When the agent creates a new agent for a specific process, duplicates the agent, or exchanges information with an existing agent for processing, the management process description held by the agent. 2. The agent management method according to claim 1, wherein the agent and the trace information are taken over by an agent requesting the processing, and the acquisition of the trace information is continued. 3. When an agent returns to an arbitrary place for analysis, by setting a return flag in a management processing content description section of the agent, the agent can be distinguished from normal processing. The agent management method according to claim 1, wherein: 4. When the agent returns to any place for analysis, the management node suspends the translation engine and uses the trace information held by the agent to
The status of all places in the translation engine is returned up to the point indicated by the time stamp based on the difference information of the status change periodically output by the translation engine, and the status of all the agents in the translation engine is also managed by the management agent. 2. The agent management method according to claim 1, wherein the information is returned to the current time from the trace information, and after the states of the agent and the place are matched, a restart process for failure analysis is performed. 5. When the agent returns to an arbitrary place for analysis, the state is returned to the processing time at the designated place by using the trace information and the time stamp held by the agent. Using the time stamp at the time of processing the agent, based on the interaction time with the agent and the ID of the agent in the variation journal information output by the place itself, the state of the agent and the place until the interaction time, Extract information,
2. The agent management method according to claim 1, wherein a restart process for autonomous failure analysis from an arbitrary place of the agent is performed by synchronizing with the state of the agent.