Getting build failures due to subj (the build was restarted since, link obsolete) occasionally, roughly once every 40-50 jobs (by perception).
What puzzles me is that the host that has failed to resolve is homebrew.bintray.com, the same that was used in lots of previous downloads in the same job, so it must be in the local DNS cache. Isn’t it installed in MacOS builders or something, or has a duration shorter than the job time limit?
I restarted the build I’ve linked to in the previous message, so that link is no longer valid.
Hit this again today: https://travis-ci.org/skvark/opencv-python/jobs/510371063#L8559 (this is a PR build so I can’t restart it and this link should stay valid). This time, homebrew.bintray.com and github.com failed to resolve. This happened shortly after the 20 minutes mark.
Got a failure in Linux, too: https://travis-ci.org/skvark/opencv-python/jobs/510375131#L4477 , for pypi.org, with many retries.
When all the ~350 jobs that I scheduled yesterday finish, I’ll be able to give the full stats on how frequent this error is.
| osx | DNS fails | linux | DNS fails | Σ jobs | Σ fails | |
|---|---|---|---|---|---|---|
| https://travis-ci.org/skvark/opencv-python/builds/510371038 | 39 | 1 | 47 | 86 | 1 | |
| https://travis-ci.org/skvark/opencv-python/builds/510373939 | 40 | 48 | 88 | 0 | ||
| https://travis-ci.org/skvark/opencv-python/builds/510375066 | 20 | 48 | 1 | 68 | 1 | |
| https://travis-ci.org/skvark/opencv-python/builds/510386616 | 40 | 48 | 88 | 0 | ||
| https://travis-ci.org/native-api/opencv-python/builds/510386513 (1 job restarted) | 41 | 1 | 48 | 89 | 1 | |
| https://travis-ci.org/native-api/opencv-python/builds/510854165 (14 jobs restarted) | 34 | 11 | 48 | 82 | 11 | |
| Σ | 214 | 13 | 287 | 1 | 501 | 14 |
| avg | 35,79 |
Had a surge of DNS errors in the last build, updated the table.