Open ahsonkhan opened 1 year ago
These are live tests, connectivity issues can cause them to fail.
Should we consider not running them in CI (and only on nightly), if we expect them to be brittle? Alternatively, we add some retries to improve reliability.
Should we consider not running them in CI (and only on nightly), if we expect them to be brittle? Alternatively, we add some retries to improve reliability.
These tests aren't any more brittle than all of the other HTTP transport tests.
All tests which interact with network services will have some level of failure associated with them. We absolutely can consider removing all HTTP transport related tests from our CI pipeline, but that means that we will not have any test coverage on our HTTP transports in our CI Pipeline.
Over the past 30 days (the extent of errors recorded in AzureDevOps), this test has a 99.93% reliability. There have been 2 failures recorded over that interval.
IMHO that is not sufficiently high a level to justify disabling the test.
The test is failing on an unrelated PR with a timeout, both on x86 and x64
Win2022_Win32Api_debug_tests_winhttp_x86
https://dev.azure.com/azure-sdk/public/_build/results?buildId=3246625&view=logs&j=85bfc3e6-dd05-5f88-257f-550486204084&t=744927fc-7997-531b-7e25-698f166ee20fAlso here:
Win2022_Win32Api_debug_tests_winhttp_x64
https://dev.azure.com/azure-sdk/public/_build/results?buildId=3246625&view=logs&j=fe290b03-58c1-5285-793c-3227e2272b55&t=271ddb24-2791-55d7-7060-2a0f97b70f60Related issue on the libcurl side: https://github.com/Azure/azure-sdk-for-cpp/issues/4133
cc @LarryOsterman