Testing Farm release 2025-01.7

This release is introducing hard memory limits on our worker containers, possibly affecting some of the requests. In case you hit the problem you should discover that easily from the request error. The request error will link you to our error TFE-1 for information how to deal with the problem.

Testing Farm workers are now correctly enforcing a memory limit for worker containers, greatly improving stability of all requests and mitigating requests to get stuck in running state. This can be breaking for some of the tests, see 💔 Breaking changes section for details. (TFT-3333 TFT-3334 TFT-3338 TFT-1762)

Testing Farm now uses by default non-spot instances in AWS. Due to the elevated usage of dedicated Testing Farm instances, the spot instances are terminated more frequently, causing random test failures. Until we are able to handle this problem gracefully, i.e. by automatically restarting the affected requests, we decided to move the testing to non-spot instances. This will affect the cost of testing, but we do not expect the change to have large impact. We will be monitoring the situation going forward. (TFT-3371)

A bug in upstream dramatiq was found and patch in Artemis provisioner, greatly improving the stability of provisioning and mitigate outages. (TFT-3331)

The pool field under the provision step in tmt plans is now honored in the single-host pipelines. (TFT-3368)

List of important packages bundled in the worker image.