Investigate flaky test - sequential/test-performance-eventloopdelay

[mhdawson]

Test

sequential/test-performance-eventloopdelay

Platform

linuxone

Console output

node:assert:399
    throw err;
    ^

AssertionError [ERR_ASSERTION]: The expression evaluated to a falsy value:

  assert(histogram.min > 0)

    at Timeout.spinAWhile [as _onTimeout] (/home/iojs/build/workspace/node-test-commit-linuxone/test/sequential/test-performance-eventloopdelay.js:65:7)
    at listOnTimeout (node:internal/timers:568:17)
    at processTimers (node:internal/timers:510:7) {
  generatedMessage: true,
  code: 'ERR_ASSERTION',
  actual: false,
  expected: true,
  operator: '=='
}

Node.js v18.0.0-pre

Build links

https://ci.nodejs.org/job/node-test-commit-linuxone/30015/

Additional information

No response

[richardlau]

According to https://github.com/nodejs/reliability/issues?q=is%3Aissue+test-performance-eventloopdelay this first showed up on 22 December 2021.

[richardlau]

Walking backwards from commits landed on 22 December 2021, I've started two stress runs around 23637e9:

665b404: https://ci.nodejs.org/view/Stress/job/node-stress-single-test/306/nodes=rhel7-s390x/ (✔️ no failures in 1000 runs)
23637e9: https://ci.nodejs.org/view/Stress/job/node-stress-single-test/307/nodes=rhel7-s390x/ (❌ 100 failures in 1000 runs)

[richardlau]

I've added some debug to the test

diff --git a/test/sequential/test-performance-eventloopdelay.js b/test/sequential/test-performance-eventloopdelay.js
index f262e9de3f..eddc9348fa 100644
--- a/test/sequential/test-performance-eventloopdelay.js
+++ b/test/sequential/test-performance-eventloopdelay.js
@@ -62,7 +62,7 @@ const { sleep } = require('internal/util');
       histogram.disable();
       // The values are non-deterministic, so we just check that a value is
       // present, as opposed to a specific value.
-      assert(histogram.min > 0);
+      assert(histogram.min > 0, `histogram.min (${histogram.min}) > 0`);
       assert(histogram.max > 0);
       assert(histogram.stddev > 0);
       assert(histogram.mean > 0);

and kicked off another stress run (based on current master branch): https://ci.nodejs.org/view/Stress/job/node-stress-single-test/309/nodes=rhel7-s390x/console
and it looks like in the failure case histogram.min is 0:
e.g.

18:00:25 + /usr/bin/python3.6 tools/test.py -p tap --logfile test20.tap --mode=release sequential/test-performance-eventloopdelay
18:00:25 TAP version 13
18:00:25 1..1
18:00:32 not ok 1 sequential/test-performance-eventloopdelay
18:00:32   ---
18:00:32   duration_ms: 7.169
18:00:32   severity: fail
18:00:32   exitcode: 1
18:00:32   stack: |-
18:00:32     node:assert:399
18:00:32         throw err;
18:00:32         ^
18:00:32     
18:00:32     AssertionError [ERR_ASSERTION]: histogram.min (0) > 0
18:00:32         at Timeout.spinAWhile [as _onTimeout] (/home/iojs/build/workspace/node-stress-single-test/nodes/rhel7-s390x/test/sequential/test-performance-eventloopdelay.js:65:7)
18:00:32         at listOnTimeout (node:internal/timers:559:17)
18:00:32         at processTimers (node:internal/timers:502:7) {
18:00:32       generatedMessage: false,
18:00:32       code: 'ERR_ASSERTION',
18:00:32       actual: false,
18:00:32       expected: true,
18:00:32       operator: '=='
18:00:32     }
18:00:32     
18:00:32     Node.js v18.0.0-pre

[Trott]

Happened again today (and I'm sure many, many other days, but since it's marked as flaky, people probably don't log it).

https://ci.nodejs.org/job/node-test-commit-linuxone/nodes=rhel8-s390x/31271/console

00:16:18 not ok 3582 sequential/test-performance-eventloopdelay # TODO : Fix flaky test
00:16:25   ---
00:16:25   duration_ms: 7.211
00:16:25   severity: flaky
00:16:25   exitcode: 1
00:16:25   stack: |-
00:16:25     node:assert:399
00:16:25         throw err;
00:16:25         ^
00:16:25     
00:16:25     AssertionError [ERR_ASSERTION]: The expression evaluated to a falsy value:
00:16:25     
00:16:25       assert(histogram.min > 0)
00:16:25     
00:16:25         at Timeout.spinAWhile [as _onTimeout] (/home/iojs/build/workspace/node-test-commit-linuxone/test/sequential/test-performance-eventloopdelay.js:65:7)
00:16:25         at listOnTimeout (node:internal/timers:559:17)
00:16:25         at processTimers (node:internal/timers:502:7) {
00:16:25       generatedMessage: true,
00:16:25       code: 'ERR_ASSERTION',
00:16:25       actual: false,
00:16:25       expected: true,
00:16:25       operator: '=='
00:16:25     }
00:16:25     
00:16:25     Node.js v18.0.0-pre
00:16:25   ...

[Trott]

Is this a bug in the test and the docs? Is it possible for the minimum recorded event loop delay to be 0 (or rather, to round to 0) on a sufficiently fast machine? @jasnell

[jasnell]

Most likely test bug. I'll be able to look tomorrow