-
Notifications
You must be signed in to change notification settings - Fork 29.7k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Investigate flaky parallel/test-net-write-slow on FreeBSD #7516
Comments
And again: https://ci.nodejs.org/job/node-test-commit-freebsd/3107/nodes=freebsd10-64/console:
|
Stress test to confirm it's reproducible via stress test: https://ci.nodejs.org/job/node-stress-single-test/789/nodes=freebsd102-64/console |
Failures were both on test-digitalocean-freebsd10-x64-2. Stress test (with no failures yet) is also on test-digitalocean-freebsd10-x64-2. ¯_(ツ)_/¯ |
Other recent failures besides 3080 and 3107: 3089: https://ci.nodejs.org/job/node-test-commit-freebsd/3089/nodes=freebsd10-64/console
3087: https://ci.nodejs.org/job/node-test-commit-freebsd/3087/nodes=freebsd10-64/console
3085: https://ci.nodejs.org/job/node-test-commit-freebsd/3085/nodes=freebsd10-64/console
3084: https://ci.nodejs.org/job/node-test-commit-freebsd/3084/nodes=freebsd10-64/console
3083: https://ci.nodejs.org/job/node-test-commit-freebsd/3083/nodes=freebsd10-64/console
|
Seems significant that /cc @nodejs/platform-freebsd |
Stress test came up clean. I repeat: ¯_(ツ)_/¯ |
Stress test label for FreeBSD has been updated to match that of node-test-pull-request. I don't think it's any different than the test I already ran, but just in case, and because I'm running out of things to try without spinning up my own FreeBSD instance: https://ci.nodejs.org/job/node-stress-single-test/791/nodes=freebsd10-64/ |
This test was introduced in 0f0af55a after 0.6.10 and before 0.6.11. It was intended to fix nodejs/node-v0.x-archive#2678. So it should fail in 0.6.10 and pass in 0.6.11 and later. However, for me, it does not fail in 0.6.10 in its current form. (You'll need to comment out the The original form fails or succeeds in both 0.6.x versions depending on how the two timeout values are tweaked. In fact, the same is true of the current form of the test running Node 6.2.2. On my laptop, if I change A factor of two is not a lot of head room, in my opinion. At this point, I'm not sure if the thing to try is change /cc @koichik who wrote the test originally, in case there's additional light to be shed. |
@Trott both freebsd machines have JOBS=2 set (and have 2 vcpus) |
Trying to run on all platforms as a stress test to see if any of them fail: https://ci.nodejs.org/job/node-stress-single-test/792/ |
Still occurring on FreeBSD but still unable to replicate in stress test. They do seem to cluster. (Could the time of day be relevant? Something else running on those machines at a particular time?) Here it is in 3118. https://ci.nodejs.org/job/node-test-commit-freebsd/3118/nodes=freebsd10-64/console:
Here it is in 3130. https://ci.nodejs.org/job/node-test-commit-freebsd/3130/nodes=freebsd10-64/console
Here it is in 3131. https://ci.nodejs.org/job/node-test-commit-freebsd/3131/nodes=freebsd10-64/console:
3132:
3135:
|
Trying something else that probably won't pan out but getting desperate: Running five |
Hey! That worked! 3138 failed! 3139 succeeded. Other three are yet to run... |
3 of the 5 runs had this test fail. That's actually potentially useful... |
Test runs of
Tests run of
Good enough for me! PR'ing a change (although I still wonder if this test isn't a candidate for removal). |
Increase socket timeout so that there is enough time to reliably run the test on FreeBSD. Fixes: nodejs#7516
https://ci.nodejs.org/job/node-test-commit-freebsd/3080/nodes=freebsd10-64/console:
The text was updated successfully, but these errors were encountered: