OpenProfiling NodeJS - Google Cloud Storage Exporter

This exporter is advised when profiling distributed applications where retrieving from each container file system can be hard. It will just write the profile to a remote S3-compatible server, the file name in each bucket will follow the following format:

const name = `${profile.kind}-${profile.startTime.toISOString()}.${profile.extension}`

Where the profile kind can be: HEAP_PROFILE, CPU_PROFILE or PERFECTO And the extension can be either: heaprofile, cpuprofile or json

Advantages

• Centralization of every profile, prefered when using containers

Drawbacks

• You need to have a S3 compatible running (or AWS S3 itself) and manage it yourself.

How to use

In the following example, when the profile will be done it will be written to the remote S3 bucket:

import { ProfilingAgent } from '@openprofiling/nodejs'
import { GcloudStorageExporter } from '@openprofiling/exporter-gcs'
import { InspectorCPUProfiler } from '@openprofiling/inspector-cpu-profiler'
import { SignalTrigger } from '@openprofiling/trigger-signal'

const profilingAgent = new ProfilingAgent()
profilingAgent.register(new SignalTrigger({ signal: 'SIGUSR2' }), new InspectorCPUProfiler())
profilingAgent.start({
  exporter: new GcloudStorageExporter({
     // Alternatively, you might pass it via the GOOGLE_APPLICATION_CREDENTIALS env variable
    keyFilename: 'some/where/key.json',
    // Alternatively, you might pass it via the GCLOUD_PROJECT env variable
    projectId: 'my-project', 
    /*
    * name of the bucket to create
    */
    bucket: 'test'
  }
})

Development

When developing against this package, you might want to run a fake gcs server with Minio to be able to run tests and verify the behavior:

docker run -d --name fake-gcs-server -p 4443:4443 fsouza/fake-gcs-server