Skip to main content

Analyzing the transactions

Hitmap​

Home > Select Project > Analysis > HITMAP

Select a project in the initial screen of the WhaTap monitoring service, and then select Analysis > HITMAP under Project Menu.

Note

Hitmap can be also accessed through the hitmap widget in Dashboard > Application Dashboard.

Detailed analysis​

Hitmap

number1 agent area​

You can filter by selecting agents linked with the current project. If you select the Icon Magnifier icon, the HITMAP Transaction chart can be viewed in the number2 area.

number2 Hitmap transaction chart​

If transactions with latency issues are included, they appear in orange or red on the Hitmap Transaction chart. If only normal transactions are included, they appear in blue.

number3 Selection area​

If you drag the real-time hitmap chart, the number5 TX Trace list appears where number4 agent list and transaction information can be viewed in the selected area.

number4 Agent list​

You can check the numbers of transactions and errors included in the selected area on the chart. If you click the desired application in the list, the information details of the application appear in number5 Transaction List.

number5 TX Trace list​

The TX Trace list contains information details. If you select a desired transaction from the list, the Statistics information window appears. In the Statistics window, you can see the detailed trace details for the transaction. For more information, see the following.

Understanding the hitmap patterns​

The hitmap is a distribution chart whose X-axis is the transaction end time, and Y-axis is the response time. A normal web application displays a distribution concentrated in a few seconds or less.

hitmap-pattern

Analyzing the hitmap lines​

  • Vertical line (LOCK symptom) pattern

    If a temporary lock (not only DB lock) occurs during transaction processing, the processing is queued. If the lock is released, pending transactions are finished together at a similar time. This creates vertical lines as follows:

    hitmap-lock

    Detecting locks in vertical line patterns is a very powerful concept. Especially in the microservice architecture, locks from backend systems can be equally detected.

    hitmap-front-api-db

    The vertical lines for response patterns in the front applications are also detected when any lock occurs in the DB used by the back-end system.

  • Horizontal line (timeout) pattern

    In the 10-second timeout condition, if the resources are insufficient, the transactions wait for 10 seconds until a timeout error occurs. At this time, horizontal lines appear near 10 seconds of the hitmap as follows:

    hitmap-timeout

    If there is a logic to retry after timeout, the horizontal lines repeat every 10 seconds as shown in the figure. The following is the hitmap for failures.

    hitmap-error

    (1) The response time increased in the section and (2) the red lines in the section are horizontal line patterns. (1) ConnectionPool is exhausted due to section load, (2) the secondary timeout failure occurred due to insufficient ConnectionPool.

Using the pattern analysis​

A line in the transaction response distribution indicates a bottleneck. If the lock is temporary, a vertical line is generated. If the bottleneck is release by timeout, a horizontal line is generated.

hitmap-pattern-analysis

When analyzing a problem, you can quickly pinpoint a transaction included in the line for selective analysis.

Analyzing the machine learning based response patterns​

This function automatically detects abnormalities after analyzing the hitmap patterns with an aid of machine learning technology and then issues an alert.

Machine learning analytics

Abnormal pattern example​

It learns abnormal patterns from hundreds of TB of performance data every month and issues an alarm when a pattern similar to the learned abnormal one is detected.

  • Vertical line pattern

    hitmap-abnormal1

  • Horizontal line pattern

    hitmap-abnormal2

  • Composite pattern

    hitmap-abnormal3

  • Hitmap alarm

    hitmap ai alert

Note

For more information about the analysis method of the hitmap transaction chart, see the following.

Criteria for exceptional handling (WARNING) of the Java agent​

This guide provides the criteria for the Java agent's exceptional handling for errors that occur in the application environment using the Java language. The following content has been written based on Java Agent v2.2.27.

Error display on the service screen​

hitmap, Trace analysis

Most Error level that can be seen throuth the Hitmap widget and Trace analysis window are for the WARNING level.

Error handling when passing the Spring ExceptionHandler​

In the Spring framework, passing the org.springframework.web.servlet.DispatcherServlet.processHandlerException method is treated as an error. However, if a developer uses the Try-Catch statement, WhaTap does not handle it as an error because it does not pass ExceptionHandler.

Status Code 400 or higher error handling​

If the HTTP response code is 400 or higher, it is treated as an error even if no exception class occurs.

Error handling when calling the JDBC driver​

The errors that occur when making a JDBC connection call or SQL execution call, are treated as errors.

Error handling for HTTP calls​

The errors that occur during HTTP calls appear in the SLOW_HTTPC or httpc error format.

WhaTap error handling​

The WhaTap monitoring service handles errors depending on specific conditions.

  • TOO_MANY_RECORDS: 1000 or more fetches

  • SLOW_SQL: SQL time of 30000 ms or more

  • SLOW_HTTPC: HTTP call elapsed time of 10000 ms or more

  • httpc error: HTTP call status code of 400 or more

  • status error: Transaction status code of 400 or more

Transaction error step INFO processing​

Some errors can be displayed as normal or ignored through the agent settings. The level displayed in the Hitmap widget is INFO (blue).

Note

Agent settings can be configured by editing the whatap.conf file or using Management > Agent CONFIG.. For more information, see the following.

  • biz_exceptions: Displays the record error.

    whatap.conf
    biz_exceptions=io.home.test.baseapp.base.exception.custom.CustomParameterizedException

  • ignore_exceptions: Does not display the record error.

    whatap.conf
    ignore_exceptions=org.springframework.web.util.NestedServletException

  • status_ignore

    whatap.conf
    status_ignore=408,500,200

  • httpc_status_ignore

    whatap.conf
    httpc_status_ignore=403,401

  • status_ignore_set

    whatap.conf
    status_ignore_set=/api/posts/test/timeout:200,/error:500

  • httpc_status_url_ignore_set

    whatap.conf
    httpc_status_url_ignore_set=/api/posts/test/timeout:408,/api/apache/httpclient/posts/timeout:500