Testing Farm release 2026-03.2

Testing Farm release 2026-03.2 was deployed on 2026-03-20 🎉.

🚀 Upgrades

tmt was updated to 1.69 from 1.68.

Testing Farm CLI tool was updated to v0.0.33, please update your installation.

Artemis was updated to v0.0.87 on both ranches.

The test pipeline runtime was migrated from Python 3.9 to Python 3.12. (testing-farm/gluetool-modules!1023)

🔮 Upcoming Breaking Changes

We are planning to lock down repositories which can be used to run tests from on the Red Hat ranch due to security concerns. Only repositories matching a pre-configured allowlist will be accepted. The public ranch will remain unaffected. We will provide an easy way to allow additional repositories via a merge request workflow to a dedicated configuration repository. All currently used repositories will be pre-populated in the allowlist before the lockdown is activated — we will work hard to not break any current users. See Repository Allowlists for details. (TFT-4358)

🎯 Highlights

RHEL-9.9.0 and RHEL-10.3 are now available in image mode. (TFT-4374)

Initial support for AMD SEV-SNP Confidential VMs added in Artemis — provisioning of confidential instances on AWS is now possible. The feature is not yet complete and not user-facing — stay tuned for upcoming releases. (TFT-4391)

Fixed archiving timeouts and incomplete artifacts that could cause requests to finish with missing test results. (TFT-4321)

🚀 Improvements

Onboarded Fedora 44 compose to the Red Hat ranch. (TFT-4443)

Added more AWS bare metal instance types on both ranches — 16 new aarch64 types (Graviton2 through Graviton4) and 10 new x86_64 types (Broadwell through AMD Turin), each with 100/250/500 GiB disk size options. Availability of AWS bare metal instances has also been improved. See documentation for details. (testing-farm/infrastructure!1261)

Increased frequency of CTC/RC compose checks to reduce delays in QE milestone testing. (TFT-4385)

Updated detection of special composes (CTC, RC baseline, candidate). (TFT-4447)

API response now mentions that timeout only activates once the request is in a running state. (TFT-3844)

Mitigated Kubernetes ConfigMap size limit (1 MiB) in Artemis Helm deployment by splitting large configurations. (TFT-3852)

Added --skip-guest-setup option to the CLI. (testing-farm/cli!211)

Added retry logic for wrapper script download in CLI container. (testing-farm/cli!212)

📝 Documentation

Added FAQ entry about BeakerLib libraries failing in image mode composes and how to fix them by using the FMF identifier format.

Added FAQ entry about unsigned package errors on Fedora 45+ due to stricter GPG signature verification defaults in dnf4.

Added FAQ entry about using dnf/yum with standalone Python runtime.

Documented webhook notification payload format. (testing-farm/docs/root!253)

🐞 Bugfixes

Fixed archiving timeouts and incomplete artifacts in destroy phase — improved parallel archiving shutdown and ensured destroy-stage artifacts are always collected. (TFT-4321)

Fixed race condition in temporary file creation in the archive module that caused failures in busy plans with many nodes. (TFT-4419)

Fixed race condition when downloading Koji/Brew builds in multi-plan image mode requests. (TFT-4418)

Fixed beaker-harness-container service failure on image-mode machines causing premature machine reclamation. (TFT-4446)

Fixed Beaker driver not exposing additional guest logs. (testing-farm/artemis!1554)

Fixed dnf/yum Ansible tasks failing when using standalone Python runtime by installing the profiles collection with proper dnf module support. (TFT-4325)

Added --checksum to rsync options to avoid redundant file syncing during archiving. (testing-farm/infrastructure!1266)

Fixed CVE exposure in test pipeline dependencies (tornado, pyopenssl). (testing-farm/gluetool-modules!1045, testing-farm/gluetool-modules!1046)

📦️ Packages

List of important packages bundled in the worker image.

❯ podman run --entrypoint rpm quay.io/testing-farm/worker-public:2026-03.2 -q fmf tmt podman beakerlib | sort | uniq
beakerlib-1.33.2-1.fc42.noarch
fmf-1.7.0-1.fc42.noarch
podman-5.8.1-1.fc42.x86_64
tmt-1.69.0-1.fc42.noarch

🔎 Stats

The average error rate for Testing Farm was approximately 3.09%.