investigate flaky test-inspector-contexts in CI

[gireeshpunathil]

• Version:
• Platform:
• Subsystem:

sequential/test-inspector-contexts is flaking (timing out) a lot in the CI. Discovered while I was running CI for progressing the C&L PRs.

Example failure:

13:05:40 not ok 1307 sequential/test-inspector-contexts
13:05:40   ---
13:05:40   duration_ms: 148.541
13:05:40   severity: fail
13:05:40   exitcode: 1
13:05:40   stack: |-
13:05:40     timeout
13:05:40     Testing context created/destroyed notifications
13:05:40   ...

https://ci.nodejs.org/job/node-test-binary-windows-2/COMPILED_BY=vs2017,RUNNER=win2016,RUN_SUBSET=0/4208/console

cc @nodejs/testing @nodejs/v8-inspector

[Trott]

Happened again today.

https://ci.nodejs.org/job/node-test-binary-windows-2/4494/COMPILED_BY=vs2017,RUNNER=win2016,RUN_SUBSET=1/console

21:44:04 not ok 1311 sequential/test-inspector-contexts
21:44:04   ---
21:44:04   duration_ms: 150.925
21:44:04   severity: fail
21:44:04   exitcode: 1
21:44:04   stack: |-
21:44:04     timeout
21:44:04     Testing context created/destroyed notifications
21:44:04   ...

test-azure_msft-win2016-x64-3

[Trott]

I wonder if session.post() is failing but since no callback is sent, the error is never received?

[Trott]

I notice that if I modify mainContextPromise such that it never executes, the test still passes. Will add a common.mustCall() to make sure the test suites actually run.

[Trott]

Confirmed that if session.post() errors, it is ignored with no callback. Will add that to the PR.

[Trott]

#30649 (comment)

Welp, this at least has the desired effect of supplying more information about why test-inspector-contexts sometimes hangs in CI.

13:29:09 not ok 1312 sequential/test-inspector-contexts
13:29:09   ---
13:29:09   duration_ms: 149.140
13:29:09   severity: fail
13:29:09   exitcode: 1
13:29:09   stack: |-
13:29:09     timeout
13:29:09     Testing context created/destroyed notifications
13:29:09     Checking/waiting for GC.
13:29:09   ...

So it looks like it just runs the first global.gc() over and over and the context is never garbage collected/destroyed?

[Trott]

@nodejs/inspector @nodejs/vm @nodejs/testing Any thoughts on what to do next to troubleshoot?

[targos]

Is this still relevant?

[Trott]

Is this still relevant?

I'm not seeing it using ncu-ci walk pr or ncu-ci walk commit so I'm going to optimistically close this.