How to improve the profiling of NodeJS applications

[hrvacadaIM]

Hi.

First of all thank you for this nice tool. We have used the https://github.com/Granulate/gprofiler/blob/master/deploy/ecs/gprofiler_task_definition.json file to create task on our ECS clusters. We are getting some metrics in GProfiler Studio, but we would like to have also better insight into our own code and not just underlying code of NodeJS. Is there a possibility to get better insight into that part with passing some arguments? For example we found this option, but I am not sure, if this is needed at all or it can be passed to ECS gprofiler task:

NodeJS profiling options
--nodejs-mode: Controls which profiler is used for NodeJS.

• none - (default) no profiler is used.
• perf - augment the system profiler (perf) results with jitdump files generated by NodeJS. This requires running your node processes with --perf-prof (and for Node >= 10, with --interpreted-frames-native-stack). See this NodeJS page for more information.

Or, if it is possible to know from GProfiler Studio which part of the own code is causing the memory and cpu usage to increase. We have selected My code and set a filter for particular Service running in ECS, but that didn't give us any better insight in the issue.

I am thankful for any valuable insight.

[Jongy]

Hi @hrvacadaIM :) I'm glad to hear gProfiler is useful for you.

If you're using NodeJS and not passing --nodejs-mode=perf, then I figure you're getting a native profile of NodeJS/V8 itself and not of your code.
To gain visibility into your NodeJS code itself (i.e to see JavaScript function names) you'll need to pass --nodejs-mode=perf to the profiler; and yes - it can be done in the ECS task definition, just add it to the command array like this:

            "command": [
                "-cu",
                "--token",
                "<TOKEN>",
                "--service-name",
                "<SERVICE NAME>",
                "--nodejs-mode=perf"
            ],

Additionally, you need to add the --perf-prof flag to your NodeJS application itself (i.e change node myscript.js to node --perf-prod myscript.js) as described here. This is required to have NodeJS generate debugging information that gProfiler will use.

After you do that - gProfiler will show you a profile of your JavaScript code, and you'll be able to figure out what parts of your code are consuming CPU time.
As for memory usage - gProfiler is (currently) only a CPU profiler, not a memory profiler, so it won't help with that :( If you want to investigate the memory usage of your app, I suggest you search for a memory profiler.