node-daily-v14.x-staging is failing on smartos

[richardlau]

Extracted out of #3131:

https://ci.nodejs.org/view/Node.js%20Daily/job/node-daily-v14.x-staging/ has been failing since 6 December. Cursory glance appears to be:

On macOS
(Fixed in #3131.)

On smartos there's one or more stray Node.js process being left behind:

11:29:41 ps awwx | grep Release/node | grep -v grep | cat
11:29:42  574381 ?        S  0:01 /home/iojs/build/workspace/node-test-commit-smartos/nodes/smartos20-64/out/Release/node -e setInterval(()=>{}, 99)
11:29:42  574413 ?        S  0:01 /home/iojs/build/workspace/node-test-commit-smartos/nodes/smartos20-64/out/Release/node -e setInterval(()=>{}, 99)
11:29:42 make[1]: *** [Makefile:515: test-ci] Error 1

Needs further investigation. The CI jobs are configured to fail if leftover Node.js processes are detected. As far as I can tell the tests themselves have passed. This may have started after the recent migration of the smartos machines (cc https://github.com/orgs/nodejs/teams/platform-smartos @bahamat @sxa ) -- I had clean CI's for Node.js 14.21.2 on December 8: https://ci.nodejs.org/job/node-test-commit-smartos/46911/nodes=smartos20-64/

[richardlau]

I've narrowed down node -e setInterval(()=>{}, 99) to coming from test-child-process-exec-abortcontroller-promisified. This test is passing but leaving behind Node.js child processes.

Curiously this test doesn't run the node child process on Linux and instead uses sleep 2m. This was done in nodejs/node#37572 -- prior to that PR all platforms used sleep 2m apart from Windows (which was flaky). If I switch the test to use sleep 2m on SmartOS the test still passes but doesn't leave behind any node child processes (no surprise since it's not creating any) but also is not leaving behind any sleep processes.

I'm not entirely sure why we didn't have the child processes being left behind on smartos before the migration -- perhaps this is something timing related? We're not seeing the leftover child processes on main or current but are consistently seeing them on Node.js 14 and 16 SmartOS CI runs when we were not prior to the migration.

I'm going to open a PR to switch back to sleep 2m for the test on POSIX platforms to unblock the Node.js 14 and 16 CI.

[richardlau]

I'm going to open a PR to switch back to sleep 2m for the test on POSIX platforms to unblock the Node.js 14 and 16 CI.

PR: nodejs/node#46276

[bahamat]

Looking at the issue, and the proposed fix, I think this is the correct approach. I can't mark approval on the PR, so I'm noting it here instead.