[SVLS-8249] Generates UUID and passes env vars for Node by Chronobreak · Pull Request #46 · DataDog/datadog-serverless-compat-js

Chronobreak · 2026-04-24T18:25:18Z

What does this PR do?

Generates UUID for Windows named pipes
Passes DD_APM_WINDOWS_PIPE_NAME for the binary
Passses DD_TRACE_AGENT_URL for the Node tracer

Motivation

https://datadoghq.atlassian.net/browse/SVLS-8249

Additional Notes

Describe how to test/QA your changes

This is running with the sm.suffix:ddk-js46

logs
traces
metrics - switch to metrics tabs from traces

Corresponding changes to come in serverless-compat-self-monitoring

Avoid pipe collisions when multiple Azure Functions share a namespace by generating a fresh pipe name per process in start() before spawning the compat binary. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

Replace the graceful-skip guard with a thrown error so any future format change that violates the Windows named-pipe length limit fails loudly instead of silently reverting to user-supplied values. Also clarifies the call site lives inside start()'s existing try/catch so the error is caught and spawn is aborted. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

Chronobreak · 2026-05-05T14:52:20Z

+      !process.env.DD_APM_WINDOWS_PIPE_NAME &&
+      !process.env.DD_TRACE_AGENT_URL


Current behavior is that it'll only generate the UUID only if both are not set (i.e. respecting user's env vars)

However I was curious about the default behavior when one-but-not-the-other is set and I think it results in a situation where traces are dropped

Only DD_APM_WINDOWS_PIPE_NAME=foo set → the spawned binary listens on \.\pipe\foo, but dd-trace-js falls back to its default agent URL (HTTP localhost:8126) and never reaches the pipe. Traces drop.

Only DD_TRACE_AGENT_URL=unix://./pipe/foo set → dd-trace-js sends to \.\pipe\foo, but the binary picks its own default pipe name and listens elsewhere. Traces drop.

# Conflicts: # src/index.spec.ts

Chronobreak · 2026-05-19T19:53:45Z

+  if (pipeName) {
+    process.env.DD_TRACE_AGENT_URL = `unix://./pipe/${pipeName}`;
+    return;
+  }
+
+  if (traceAgentUrl) {
+    const match = traceAgentUrl.match(/\/pipe\/(.+?)\/?$/);
+    if (match) {
+      process.env.DD_APM_WINDOWS_PIPE_NAME = match[1];
+    }
+    return;
+  }


If one is set but not the other, we make sure they're both set

dd-trace-js's agent exporter wraps DD_TRACE_AGENT_URL in `new URL()`, and its `parseUrl` skips the unix-pathname rewrite for URL objects. For `unix://./pipe/<n>` that yields `socketPath=/pipe/<n>` — not a valid Windows pipe path. Emit `unix:\\.\pipe\<n>` instead so pathname already contains the canonical form. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

Without DD_DOGSTATSD_WINDOWS_PIPE_NAME, the serverless-compat Rust binary binds UDP 8125 and panics with AddrInUse when co-located with another function on the same Windows plan (EP2/P1V3). Mirror the APM auto-gen path for the dogstatsd pair (DD_DOGSTATSD_WINDOWS_PIPE_NAME server-side, DD_DOGSTATSD_PIPE_NAME client-side). Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

…trace span-stats dd-trace's span-stats writer (exporters/span-stats/writer.js) builds request options with `protocol: url.protocol` and no socketPath. With the backslash form `unix:\\.\pipe\<name>`, url.protocol is `'unix:'`, which Node 22's ClientRequest rejects with ERR_INVALID_PROTOCOL, crashing the worker. Monkey-patch http(s).request/get to strip `protocol: 'unix:'` and attach socketPath from DD_APM_WINDOWS_PIPE_NAME before forwarding to the original. This preserves span-stats functionality without modifying dd-trace. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

Lewis-E · 2026-06-01T19:11:15Z

+  if (!pipeName) return;
+  const socketPath = `\\\\.\\pipe\\${pipeName}`;
+
+  for (const module of [require('http'), require('https')]) {


Do you think this could impact any other features outside of the span export you're targeting? patching all of HTTP is pretty broad. Can this impact the user's app as well as the tracer?

I think talking to the tracer team about the long term fix should at least be a ticket.

Only fires for requests carrying protocol: 'unix:'

This could easily be an app calling redis or similar. We need to scope this to specifically http as used in the one module of the tracer we want to be patching. Y'gotta sit on Claude here; a comment is not a fix. Try making it explain all the possible risks across diverse customers with complicated apps before asking it to fix the issue as well as explaining the need for the absolute minimum blast radius.

Thanks so much - took a look again and tightened to scope further to only apply to span-stats writer traffic! I'll log the additional ticket for the node tracer team and circle back here.

Ticket: https://datadoghq.atlassian.net/browse/SVLS-9290

Please test this with an app making http calls, so we can make sure there are no side effects.

Thanks so much - tested here, getting outbound http calls

BridgeAR

I believe the fix should actually be in dd-trace. I am opening a PR there. When it is fixed, we can just update dd-trace-js and that should resolve the issue, if I am not mistaken.

Chronobreak and others added 3 commits April 21, 2026 12:12

docs: design spec for unique named-pipe identifiers

fd30f95

Avoid pipe collisions when multiple Azure Functions share a namespace by generating a fresh pipe name per process in start() before spawning the compat binary. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

Add support for Windows Named-pipes to Node js

6288426

Chronobreak requested a review from a team as a code owner April 24, 2026 18:25

Chronobreak requested review from apiarian-datadog and removed request for a team April 24, 2026 18:25

Chronobreak marked this pull request as draft April 24, 2026 18:25

Chronobreak removed the request for review from apiarian-datadog April 24, 2026 18:25

Chronobreak added 2 commits April 24, 2026 14:26

Remove outdated Claude exploration

e8869e8

Add check to not override if there are existing env vars

859e25e

Chronobreak commented May 5, 2026

View reviewed changes

Chronobreak added 4 commits May 5, 2026 11:04

Address case where only one env var is set

0c709b8

Simplify tests

106862c

Merge branch 'main' into ddk/svls-compat-named-pipes

c6cc074

# Conflicts: # src/index.spec.ts

Cleanup

74395f7

Chronobreak marked this pull request as ready for review May 18, 2026 18:04

Chronobreak requested a review from Lewis-E May 19, 2026 19:52

Chronobreak commented May 19, 2026

View reviewed changes

Chronobreak and others added 6 commits May 21, 2026 17:11

Nits

764f383

Fix comment

66d5884

Clarify comment

68500e7

Lewis-E reviewed Jun 1, 2026

View reviewed changes

Comment thread src/index.ts Outdated

Lewis-E reviewed Jun 1, 2026

View reviewed changes

Comment thread src/index.ts

Lewis-E reviewed Jun 1, 2026

View reviewed changes

Comment thread src/index.ts Outdated

Chronobreak added 2 commits June 2, 2026 16:24

Address PR comments

a1acb76

Scope rewrite to only patch span-stats writer traffic

80e13b7

Lewis-E approved these changes Jun 10, 2026

View reviewed changes

BridgeAR suggested changes Jun 23, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[SVLS-8249] Generates UUID and passes env vars for Node#46

[SVLS-8249] Generates UUID and passes env vars for Node#46
Chronobreak wants to merge 17 commits into
mainfrom
ddk/svls-compat-named-pipes

Chronobreak commented Apr 24, 2026 •

edited

Loading

Uh oh!

Chronobreak May 5, 2026

Uh oh!

Chronobreak May 5, 2026

Uh oh!

Chronobreak May 19, 2026

Uh oh!

Lewis-E Jun 1, 2026

Uh oh!

Lewis-E Jun 3, 2026 •

edited

Loading

Uh oh!

Chronobreak Jun 4, 2026

Uh oh!

Lewis-E Jun 10, 2026

Uh oh!

Lewis-E Jun 10, 2026

Uh oh!

Chronobreak Jun 17, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

BridgeAR left a comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

		!process.env.DD_APM_WINDOWS_PIPE_NAME &&
		!process.env.DD_TRACE_AGENT_URL

Uh oh!

Conversation

Chronobreak commented Apr 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What does this PR do?

Motivation

Additional Notes

Describe how to test/QA your changes

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Lewis-E Jun 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

BridgeAR left a comment

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Chronobreak commented Apr 24, 2026 •

edited

Loading

Lewis-E Jun 3, 2026 •

edited

Loading