Closed
Description
There are a few tests that fail due to the underlying Docker issues. Some examples:
- shared memory test - fails from time to time on Azure Pipelines
- stdout/stderr ordering
- Couchbase k/v
Since we agreed in #1594 that we should only rerun the tests that are known to be flaky, we should have a mechanism (e.g. with a help form JUnit) to rerun them automatically, N times (where N is ~1..3)
Such test cases (important! not classes) should be explicitly marked (e.g. with an annotation) as flaky. This will also make it possible to find usages of that annotation and discover all flaky tests in the project.
"Flaky test" here means a test that fails randomly, due to some timing issues, or CIs problem with resource allocation or the underlying Docker issues