DocsICP FilingConsole
Documentation

FAQ

Updated at:
Copy as MD
Product
My Favorites

What is the relationship between the Node.js Performance Platform runtime and the community Node.js runtime?

  • The Node.js Performance Platform runtime is fully compatible with the corresponding versions of the community Node.js runtime. For more information, see the version mapping.

Does the Node.js Performance Platform runtime affect performance?

  • The Node.js Performance Platform runtime writes monitoring data to memory in the main thread every minute and uses a separate log thread to write logs to a file. Therefore, the performance impact is negligible.

  • When you perform a fault diagnosis, the diagnostic feature runs for 3 minutes and then automatically returns to the normal running state.

What extra features does the Node.js Performance Platform runtime provide?

  • Runtime memory status monitoring for the Node.js V8 virtual machine.

  • libuv runtime status monitoring.

  • Online fault diagnosis features, such as heap snapshots, CPU profiles, and GC traces.

Why does the console show 0 instances after I deploy the Node.js Performance Platform runtime?

  • Step 1: Check whether agenthub started successfully. Run the following command to check if an agenthub instance is running.

    u@h:~$ agenthub list
|- App ID -|- PID -|---- Start Time -----|--- Config Path ------------------------------------|
|    12345 | 29015 | 2018-07-11 16:53:44 | /path/to/your/config.json                          |
|----------|-------|---------------------|----------------------------------------------------|
u@h:~$

    If no agenthub instance is running, start agenthub in DEBUG mode and check the ~/.agenthub.log file for error details.

    DEBUG=* agenthub start config.json

cat ~/.agenthub.log

  • Step 2: Check the agenthub configuration. On the Instances page, click View running Node.js processes and then click Check process.

Diagnostic operation failure

Diagnostic operations must run sequentially. If you start a new operation before the current one is complete, the new operation fails. An operation is considered complete when the dump button on the Files page is enabled. A diagnostic report takes a few seconds to generate. A heap snapshot can take from several seconds to several minutes, depending on the heap size. Other operations run for 3 minutes.

Agenthub abnormal exit

If the agenthub list command cannot find the agenthub instance after you start it by running agenthub start config.json, start agenthub in DEBUG mode.

DEBUG=* agenthub start config.json

Then, check the ~/.agenthub.log file for error information.

After you fix the error, restart agenthub normally.

agenthub stop all # Stop agenthub
agenthub start config.json

How to deploy multiple applications on one server

We strongly recommend that you deploy only one application per server.

If you must deploy multiple applications on the same server, you can use one of the following two methods:

  1. Request a different App ID and App Secret for each application. Use NODE_LOG_DIR to specify different paths for the runtime logs. Ensure that these paths match the logdir path in the config.json file. This lets you start multiple agenthub instances.

  2. Start one agenthub instance. Point the runtime logs for all applications to the same directory. The default is /tmp. You can add error_log and packages to the array in the configuration file.

Use one of these methods for your deployment.

How to use the Node.js Performance Platform runtime with applications managed by pm2

  • If the community Node.js runtime and pm2 are already installed before you install the Node.js Performance Platform runtime:

    • After you install the Node.js Performance Platform runtime, reinstall pm2. Ensure that the output of which pm2 contains the .tnvm field.

    • Kill all pm2 processes, including the PM2 v0.15.8: God Daemon daemon process.

    • Restart the application using pm2.

      $ ENABLE_NODE_LOG=YES pm2 start app.js

  • If the community Node.js runtime and pm2 are not installed before you install the Node.js Performance Platform runtime:

    • After you install pm2, start the application directly using pm2.

      $ ENABLE_NODE_LOG=YES pm2 start app.js

Console displays system monitoring data but not process monitoring data

On the Instances page, click View running Node.js processes and then click Check process to investigate.

How to handle multiple instances with the same hostname

Add the "agentidMode": "IP" configuration item to the configuration file.

For example, assume you have two instances that both have the hostname app_container. Their IP addresses are 10.10.12.124 and 10.10.12.123. If you add "agentidMode": "IP" to the configuration, the instance IDs become app_container_12124 and app_container_12123. For more information, see the details.

What are slow HTTP logs?

  • HTTP requests with a response time (RT) greater than 400 ms.

What are abnormal logs?

  • These are logs generated by your application that contain an error stack. They are located in the log file specified by error_log in the config.json file.

What are module dependencies?

  • These are the npm modules specified by packages in the config.json file. Modules that have security risks are highlighted in red.

What do the system monitoring metrics mean?

Basic information

The latest reported values for each metric. The metrics are described below.

Long-term data

Memory

  • memory_sys: The percentage of system memory usage.

  • memory_node: The percentage of total system memory used by all Node.js processes combined.

CPU

  • cpu_sys: The percentage of system CPU usage.

  • cpu_node: The sum of the CPU usage percentages of all Node.js processes.

Load

  • load1: The average load over 1 minute.

  • load5: The average load over 5 minutes.

  • load15: The average load over 15 minutes.

  • The following is reference information for Load. The Load value is normalized. For an N-core CPU, the actual load is Load × N:

    • 0.7 < Load < 1: Good. New tasks can be processed promptly.

    • Load = 1: Tasks may soon require additional waiting time to be processed. This situation requires attention.

    • Load > 5: Tasks require a long waiting time. Intervention is needed.

    • Typically, you should check load15 first. If it is high, check load1 and load5 to determine if there is a downward trend. A load1 value greater than 1 for a short period is not a cause for concern. However, a Load value that remains high for a long time requires attention.

    • Understanding Linux CPU Load - when should you be worried?

QPS

The sum of HTTP requests per second processed by all Node.js processes in the instance.

GC

  • gc_avg: The average percentage of time spent on garbage collection across all Node.js processes.

  • gc_max: The maximum percentage of time spent on garbage collection by a single Node.js process within a one-minute interval.

Apdex

Application Performance Index

The Node.js Performance Platform defines user satisfaction based on the HTTP response time (RT). The default threshold, T, is 100 ms:

  • satisfied: RT <= T

  • tolerating: T < RT < 4 * T

  • frustrated: RT > 4 * T

  • Formula: satisfied + tolerating / 2

  • For example:

    • If, in one minute, 60% of HTTP requests are satisfied, 30% are tolerating, and 10% are frustrated:

    • The Apdex score is 0.6 + (0.3 / 2) = 0.75

Apdex detail

This provides a detailed breakdown of HTTP requests by response time.

node process count

The number of Node.js processes running in the instance.

Disk

The disk usage of the instance.

What do the process monitoring metrics mean?

Overall heap information

  • rss: The actual memory used by the process, including both on-heap and off-heap memory.

  • heap_total: The total size of the heap memory.

  • heap_used: The amount of heap memory currently in use.

GC Information

  • scavenge_duration: The percentage of time spent on scavenge garbage collection.

  • marksweep_duration: The percentage of time spent on marksweep garbage collection.

Heap space composition

The size of each memory space in the heap: code_space, lo_space, map_space, new_space, and old_space.

QPS Trend

The number of HTTP requests per second handled by this process.

CPU Trend

The parameters now, cpu_15, cpu_30, and cpu_60 indicate the CPU usage of the process over the last 1 s, 15 s, 30 s, and 60 s, respectively.

Timer trends

This indicates the number of timers in the process.

  • All timers in the js layer with the same timeout value are counted as a single entry in this statistic.

Libuv Handle Trend

The statistics for libuv handles in this process, where:

  • Each TCP connection uses one handle.

  • Each open file uses one handle.

How to configure alert items

For more information, see User Guide: Configure Alerts.

How does the Node.js Performance Platform perform fault diagnosis?

For more information, see the Fault Diagnosis User Guide.

What is the difference between abnormal logs and performance logs?

  • Abnormal logs are generated by the application.

  • Performance logs are written by the runtime to the directory specified by NODE_LOG_DIR (the default is /tmp) if you set ENABLE_NODE_LOG=YES. By default, performance logs are not written. A new log file is created each day. For example, the file node-20171010.log contains the logs from October 10, 2017.

How to determine if there is a memory leak

  • The memory monitoring metrics section shows that memory usage continuously increases over time.

How to determine if there is a CPU hot spot function

  • CPU usage remains consistently high.