Type of Request

bug

Infrastructure Environment

In looking at why a test (web_tool_tests_1_2 ) is timing out around the 45 minute mark even though it is marked with timeout: 60, I realized that many tests specify a timeout - which seems to apply to the entire task, but no test_timeout_secs entry, which applies to one of the steps that actually does the work. I don't know what the default is, or which kinds of tasks use that secondary timeout property, but there was another test which was timing out similarly and we solved it by adjusting the test_timeout_secs entry to better match the regular timeout entry.

It would probably be a good idea to either run through and manually double-check the entries that have long run times, and/or have the test_timeout_secs default to a large fraction (90%? 100%?) of the total timeout entry in the absence of an explicit value if it has access to that information.

See #168863

What is happening?

Tasks are timing out without reaching the total test timeout value.

Steps to reproduce

Look at the history of test timeouts, in particular web_tool_tests_1_2 is currently in this condition. Given the number of entries in .ci.yaml that have timeout entries without test_timeout_secs values, there are likely others in this situation.

Expected results

Tests should rarely time out before their explicit whole-test timeout value.