feat(node): Add breadcrumbs for child_process and worker_thread

[timfish]

I saw these were available in diagnostics_channel so this PR adds a default integration that includes breadcrumbs for both child process and worker thread start, stop and error.

[github-actions]

size-limit report 📦

Path	Size	% Change	Change
@sentry/browser	22.73 KB	-	-
@sentry/browser - with treeshaking flags	21.53 KB	-	-
@sentry/browser (incl. Tracing)	35.12 KB	-	-
@sentry/browser (incl. Tracing, Replay)	71.86 KB	-	-
@sentry/browser (incl. Tracing, Replay) - with treeshaking flags	62.26 KB	-	-
@sentry/browser (incl. Tracing, Replay with Canvas)	76.21 KB	-	-
@sentry/browser (incl. Tracing, Replay, Feedback)	88.99 KB	-	-
@sentry/browser (incl. Tracing, Replay, Feedback, metrics)	90.82 KB	-	-
@sentry/browser (incl. metrics)	27 KB	-	-
@sentry/browser (incl. Feedback)	39.87 KB	-	-
@sentry/browser (incl. sendFeedback)	27.38 KB	-	-
@sentry/browser (incl. FeedbackAsync)	32.17 KB	-	-
@sentry/react	25.49 KB	-	-
@sentry/react (incl. Tracing)	38.09 KB	-	-
@sentry/vue	26.91 KB	-	-
@sentry/vue (incl. Tracing)	37.02 KB	-	-
@sentry/svelte	22.87 KB	-	-
CDN Bundle	24.11 KB	-	-
CDN Bundle (incl. Tracing)	36.96 KB	-	-
CDN Bundle (incl. Tracing, Replay)	71.65 KB	-	-
CDN Bundle (incl. Tracing, Replay, Feedback)	76.99 KB	-	-
CDN Bundle - uncompressed	70.7 KB	-	-
CDN Bundle (incl. Tracing) - uncompressed	109.73 KB	-	-
CDN Bundle (incl. Tracing, Replay) - uncompressed	222.24 KB	-	-
CDN Bundle (incl. Tracing, Replay, Feedback) - uncompressed	235.46 KB	-	-
@sentry/nextjs (client)	38.06 KB	-	-
@sentry/sveltekit (client)	35.74 KB	-	-
@sentry/node	125.45 KB	+0.24%	+305 B 🔺
@sentry/node - without tracing	94.54 KB	+0.3%	+289 B 🔺
@sentry/aws-serverless	104.07 KB	+0.25%	+263 B 🔺

View base workflow run

[lforst]

Sorry for letting this one slip a bit. I have a concern with the value behind the individual things we log. Right now it seems we do:

• child/worker spawned
• child/worker exited

If I put myself into the seat of a user that looks at an error in Sentry, I don't think I care to much when a worker/child is created. I mostly care about whether I receive some faulty messages from a worker (I don't think we can instrument this reliably), or whether a child/worker dies unexpectedly.

What do you think about limiting the created breadcrumbs to error exits of children only?

[timfish]

I mostly care about whether I receive some faulty messages from a worker (I don't think we can instrument this reliably)

Both Worker and ChildProcess are EventEmitters, so we could use the 'newListener' event to wrap users 'message' handlers and add things like tracing or to add some more context if their handler functions throw.

What do you think about limiting the created breadcrumbs to error exits of children only?

Yep makes sense, I'll change this!

[lforst]

I mostly care about whether I receive some faulty messages from a worker (I don't think we can instrument this reliably)

Both Worker and ChildProcess are EventEmitters, so we could use the 'newListener' event to wrap users 'message' handlers and add things like tracing or to add some more context if their handler functions throw.

Ah that's good to know. I still have concerns with how spammy this potentially gets so let's keep this in our head in case we want to instrument it in the future. Or maybe we can make it optional 🤔

[timfish]

Or maybe we can make it optional

We can also just add it later in another PR!

I guess we already do something similar in browserApiErrorsIntegration to wrap setTimeout, setInterval, etc? Why is that browser only and not needed in Node?