Mass test cases cloud mode testing, occasional network fluctuations can lead to the overall task termination

[xu-xiang]

Network fluctuations occur occasionally; during long-term testing, any test case may experience network unavailability, causing the entire task to terminate. It seems there is no current way to ignore such errors.

output:
expect_error: true

It appears ignoring errors might be possible by writing overrides for each test individually (though it seems there’s no global configuration available in overrides.yaml). In unstable network environments, there should be a global parameter to accept occasional network failures instead of causing the whole task to fail.

[theseion]

Thanks @xu-xiang, we'll look into it.

[etiennemunnich]

+1 seems like a good opp for connection pooling/retry logic per task?

[theseion]

Maybe. Pooling could potentially be an issue due to races when writing the log.