test: test-http-pipeline-flood failures on Win32

[stefanmb]

• Version: Node 6.0.0
• Platform: Windows 7 Enterprise SP1, X86
• Subsystem: test

test-http-pipeline-flood is failing with an approximate rate of 1/1000 runs on my machine. This is what the failing output looks like:

assert.js:90
  throw new assert.AssertionError({
  ^
 AssertionError: Unexpected data received
    at Object.exports.fail (node\test\common.js:462:10)
    at Socket.req.socket.on node\test\parallel\test-http-pipeline-flood.js:45:46)
    at emitOne (events.js:101:20)
    at Socket.emit (events.js:188:7)
    at readableAddChunk (_stream_readable.js:172:18)
    at Socket.Readable.push (_stream_readable.js:130:10)
    at TCP.onread (net.js:535:20)

There have been at least two previous PRs related to fixing this test in #4789 and #3636, and the test has previously failed during unrelated CI, see for example: #5419 (comment)

[santigimeno]

I've run a couple of stress test jobs on Win10 but could not reproduce it:
https://ci.nodejs.org/job/node-stress-single-test/681/
https://ci.nodejs.org/job/node-stress-single-test/682/

/cc @nodejs/platform-windows

[stefanmb]

@santigimeno Thanks for running the stress test jobs. With my machine I can reproduce it fairly rapidly, time allowing I'll try to debug it later this week. One detail that may or may not be relevant is that the machine is a VMWare instance. I wanted to have the issue open in case others are experiencing this as well, or if there are any more failures during ci runs.

[Trott]

It might be useful to do a resource comparison between your VM and CI. Memory, CPU, that sort of thing...

[stefanmb]

@Trott I did briefly look at the VM config, it doesn't look too underpowered:

Xeon E7-2870 @ 2.40Ghz / 8GB RAM / Windows 7 Enterprise

When I get a chance I'll try to debug this further.

[jasnell]

Closing. If this is still an issue we can reopen

[gibfahn]

We're seeing this on every test run.

[Trott]

@gibfahn In CI? Link? On Windows in CI, I'm seeing parallel/test-zlib-failed-init failing a lot and some other stuff here and there, but not this.

[gibfahn]

Sorry, two different uses of we. Inside IBM on our CI we're seeing this every run, on a variety of machines. I have no idea what the reason for it is.