All Products
Search
Document Center

Thread profiling

Last Updated: Mar 25, 2019

As a code-based diagnosis tool, ARMS TProf automatically captures stack snapshots for slow calls and restores the first scene of code implementation.

Scenarios

The following describes typical scenarios:

  • If slow calls occur during peak traffic of promotions and sales, ARMS TProf quickly locates faulty codes.

  • If a large number of slow calls occur in the system, ARMS TProf automatically saves the first scene.

  • If occasional slow calls cannot recur because the business is too complex, ARMS TProf can restore the real execution trajectory.

Usage

Set thread profiling parameters

  1. Log on to the ARMS console. In the left-side navigation pane, chooseApplication Monitoring > Applications .

  2. On the Applications page, click the name of the target application.

  3. In the left-side navigation pane, click Application Settings, and click the Custom Configuration tab on the right.

  4. In the Thread Profiling Settings area, you can turn on or off the thread profiling control switch, and set the trigger threshold for slow calls.

Note:

  • The listener is started only when the service call RT exceeds the threshold (1,000 ms by default) and lasts until the call ends or the consumed time exceeds 15 seconds.

  • We recommend that you set the threshold to the 99th percentile of the call RT. For example, if 100 calls are listed in ascending order by RT, the 99th one is represented by the 99th percentile.

View thread profiling details through API snapshots

  1. Log on to the ARMS console. In the left-side navigation pane, chooseApplication Monitoring > Applications .

  2. On the Applications page, click the name of the target application.

  3. In the left-side navigation pane, click Interface Call, and then click the Interface Snapshot tab on the right.

  4. On the Interface Snapshot tab page, click the target TraceId . The Traces tab page is displayed in a new window.

  5. In the Thread Profiling column, click the magnifier icon. The Thread Profiling dialog box is displayed.

Note:

  • The actual RT is the actual execution time for the service call, which is not affected by thread profiling.

  • The listening RT is the time consumed for TProf listening. Usually, the listening RT is approximately the actual RT minus the trigger threshold for slow calls.

View thread profiling details through multi-dimensional query

  1. Log on to the ARMS console. In the left-side navigation pane, chooseApplication Monitoring > Multi-dimensional Query .

  2. On the Trace Query tab page, select Thread Profiling Snapshots Only from the Parameter Name drop-down list, and click Search.

  3. Click the target TraceId from the search results. The Traces tab page is displayed in a new window.

  4. In the Thread Profiling column, click the magnifier icon.

FAQ

Why are some slow calls not listened to?

A: If a large number of slow calls occur in a short period of time, to ensure the stability of the business system, the ARMS instance limits listening resources for some slow calls.