If our "test farm tests" are running in Azure and are interrupted, they can leave EC2 instances running. This is bad because it costs money and reduces the number of EC2 instances we have available.
The two main ways this happens is:
- Something goes wrong with the tests and Azure kills them for taking too long. A log of this happening from when we used to use Travis is at https://gist.github.com/bmw/074c100f0a2d6ac3326e31efa224ccf5.
- A test run in Azure is canceled.
If #6586 doesn't happen first, we should try and fix this.
cc @maximillianh
If our "test farm tests" are running in Azure and are interrupted, they can leave EC2 instances running. This is bad because it costs money and reduces the number of EC2 instances we have available.
The two main ways this happens is:
If #6586 doesn't happen first, we should try and fix this.
cc @maximillianh