AIX, FreeBSD: test-cluster-disconnect-handles intermittent failures

[mhdawson]

• Version: master
• Platform: AIX
• Subsystem: cluster

I've seen a number of failures on this test over the last little while:

https://ci.nodejs.org/job/node-test-commit-aix/2942/nodes=aix61-ppc64/console

not ok 127 parallel/test-cluster-disconnect-handles
  ---
  duration_ms: 60.107
  severity: fail
  stack: |-
    timeout
  ...

[Trott]

Also FreeBSD.

https://ci.nodejs.org/job/node-test-commit-freebsd/6249/nodes=freebsd10-64/console

not ok 226 parallel/test-cluster-disconnect-handles
  ---
  duration_ms: 60.78
  severity: fail
  stack: |-
    timeout

[Trott]

Seems like this comment in the test kinda sorta acknowledges that there is a race condition:

          // It can happen that the first continue was sent before the break
          // event was received. If that's the case, send also a continue from
          // here so the worker exits

[Trott]

The test uses --debug so it's going to break when we update V8 at some point in the next few months.

Not sure if the right thing to do is just get rid of the test or to rewrite it to use something else like --inspect.

/cc @nodejs/testing @nodejs/diagnostics @jkrems @s3ththompson @eugeneo

[Trott]

Hopefully I didn't typo any of the things in the form, but stress tests for...

AIX: https://ci.nodejs.org/job/node-stress-single-test/1097/nodes=aix61-ppc64/console

FreeBSD: https://ci.nodejs.org/job/node-stress-single-test/1098/nodes=freebsd10-64/console

[Trott]

Oh, maybe the failure could be due to a stuck debugger test on one of the hosts preventing usage of the TCP port. I know @thefourtheye has been trying to get the debugger tests working again, and that's involved running them in CI for the first time in a very long time. And some of them hung...

[Trott]

/cc @bnoordhuis who wrote the test in the first place, especially on the question of whether it makes sense to port the test to something that doesn't use the to-be-removed debug protocol or whether it makes sense to simply delete the test. (Basically: Is the debug protocol part of what's actually being tested? Doesn't seem like it to me, but ¯\(ツ)/¯ )

[Trott]

PR where the test landed: #3677

Maybe /cc @indutny on how to proceed as well...

[mhdawson]

Another failure here: https://ci.nodejs.org/job/node-test-commit-aix/nodes=aix61-ppc64/3511/console

[mhdawson]

I think this may also require cleanup after a failure as I saw it consistently fail a few times on the same machine and then there were processes hanging around on that machine. @bnoordhuis can you help out so that we can decide if we can just delete the test or if its needs to be fixed ?