Repository navigation
Investigate flaky test-fs-readfile-tostring-fail on macOS #10742
Description
- Version: master
- Platform: macOS
- Subsystem: test
This test just failed recently on an osx1010 node. It appears it just timed out?:
not ok 405 parallel/test-fs-readfile-tostring-fail
---
duration_ms: 60.203
severity: fail
stack: |-
timeout
/cc @nodejs/testing
Activity
Trott commentedon Jan 11, 2017
On an idle machine, this test is fast. But it gets slow if I run multiple instances of it in parallel. I wonder if it should be moved to
sequentialso that it does not compete for resources with other tests.Vague benchmark: If I run 96 instances of this test simultaneously, it takes about 55 seconds to complete. If I run 96 instances of test-fs-readfile-empty.js simultaneously, it takes 2 seconds. test-fs-readfile-error.js: 7 seconds. test-fs-readfile-fd.js: 2 seconds.
Trott commentedon Jan 11, 2017
Test fails reliably in CI if 32 instances are run at once. (Might not even need that many.) https://ci.nodejs.org/job/node-stress-single-test/1101/nodes=osx1010/console
Same job moved to sequential is reliable: https://ci.nodejs.org/job/node-stress-single-test/1102/nodes=osx1010/console
Will submit a PR to move it.
test: move resource intensive test to sequential
test: move resource intensive test to sequential
test: move resource intensive test to sequential
test: move resource intensive test to sequential
test: move resource intensive test to sequential
test: move resource intensive test to sequential
test: move resource intensive test to sequential
test: move resource intensive test to sequential
refack commentedon Oct 29, 2017
Flake is back (even in
sequential) - https://ci.nodejs.org/job/node-test-commit-osx/13607/nodes=osx1010/consoleTrott commentedon Oct 30, 2017
@refack Can you open a separate issue for things like this where the new failure is different (and likely unrelated) to the issue reported previously? The issue reported here is a timeout. The test was taking more than 60 seconds to run. The one you link to ls a different kind of failure and is very unlikely to be related.
Trott commentedon Oct 30, 2017
(I opened it as #16601.)