Closed schrolla closed 5 days ago
Short term work around may include changing the nightly functional test parameters such that conflicting testplans are not run. Long term, the workflow would benefit from more selective concurrent execution.
We've started looking down this path, but the current matrix setup (while generally it supports this type of concurrency) likely requires some serious restructuring to be able to define the right keys at the right level to actually prevent concurrent runs against the same tenant. TBD
🐛 Summary
The current nightly functional testing workflow uses a matrix to build a set of concurrent functional test run jobs across products/variants/test tenants. However, if the parameters include test plans for the same product against the same test tenant and they are run in concurrently, then the tests may fail incorrectly as both jobs running at the same time make conflicting changes to the tenant configuration. The result is incorrect test results (failures) on one or more of the jobs.
To reproduce
Steps to reproduce the behavior:
Expected behavior
Test jobs with a potential to conflict (same product/test tenant) will not be run concurrently.
Any helpful log output or screenshots
Example showing a failed test run in both test tenant 3 and 6 when standard and G5 variant test jobs are run in the same tenant concurrently.