- Reuse bounded Vagrant startup batches when machine state already exists.
- Avoid falling back to a concurrent five-guest startup during retries.
- Keep instance reconciliation after all batches complete.
- Avoid concurrent SSH startup bursts against the disposable guests.
- Ensure every guest gathers interface facts before network setup.
- Prevent unreachable hosts from falling back to the default eth0 interface.
- Use eth1 for Ubuntu and Debian Bento guests when present.
- Keep enp0s8 for Rocky Linux guests and other images.
- Apply the selection consistently across all Molecule scenarios.
- Replace legacy eth1 overrides with the enp0s8 private NIC used by Bento guests.\n- Keep flannel, Calico, Cilium, and kube-vip scenarios on the Vagrant private network.
- Use published Bento boxes for Ubuntu 26.04, Debian 13, and Rocky Linux 10.1.\n- Pin exact VirtualBox amd64 box versions in CI.\n- Remove the obsolete Ubuntu SSH workaround and update fixtures.
- restore host-only routing required by outside verification
- assign deterministic adapter MACs to the five default guests
- pin disposable full-mesh neighbor entries from live Ansible facts
- move the five-node cluster NICs to a named VirtualBox internal network
- preserve NAT connectivity for provisioning and downloads
- retain serialized creation and early neighbor identity validation
- serialize five-node Vagrant creation to avoid VirtualBox host-only races
- refresh and verify the primary neighbor mapping before convergence
- include interface MAC details in failure diagnostics
- pin kube-vip and cluster traffic to the private guest interface\n- disable disposable guest firewalls and verify API reachability before joins\n- keep control-plane orchestration on the primary and preserve failure diagnostics
- Retry transient k3s release downloads with bounded backoff.\n- Bound failure diagnostics and validate guest release connectivity.\n- Discover repository-owned Molecule state under the actual project root.
* Change to FQCN with ansible-lint fixer
Since ansible-base 2.10 (later ansible-core), FQCN is the new way to go.
Updated .ansible-lint with a production profile and removed fqcn in skip_list.
Updated .yamllint with rules needed.
Ran ansible-lint --fix=all, then manually applied some minor changes.
* Changed octal value in molecule/ipv6/prepare.yml
* Update pre-commit actions
This was done by running "pre-commit autoupdate --freeze".
* Remove pre-commit only dependencies from requirements.in
Including them in the file would create the illusion that those were the
versions actually used in CI, but they are not. The exact versions are
determined by the pre-commit hooks which are pinned in
.pre-commit-config.yaml.
* Ansible Lint: Fix role-name[path]
* Ansible Lint: Fix name[play]
* Ansible Lint: Fix key-order[task]
* Ansible Lint: Fix jinja[spacing]
* Ansible Lint: Fix no-free-form
* Ansible Lint: Fix var-naming[no-reserved]
* Ansible Lint: Fix yaml[comments]
* Ansible Lint: Fix yaml[line-length]
* Ansible Lint: Fix name[casing]
* Ansible Lint: Fix no-changed-when
* Ansible Lint: Fix fqcn[action]
* Ansible Lint: Fix args[module]
* Improve task naming
* fix(requirements.txt): Use pip-compile
* fix(lint): Remove anchors from molecule since they aren't yet supported via lint
* fix(lint): Remove anchors from molecule since they aren't yet supported via lint
* Test CentOS 7 in CI
* Drop support for CentOS, test on Rocky and Debian
* Fix reset playbook for Rocky Linux
* Fix typo
* Disable firewalld during testing
Co-authored-by: Techno Tim <timothystewart6@gmail.com>
* Molecule: Derive overrides.yml location from scenario dir
# Conflicts:
# molecule/default/molecule.yml
# molecule/ipv6/molecule.yml
* Molecule: Add single_node scenario
* Fix get_nodes test for the case of empty groups
* Test cluster using molecule
* Fix detection of first control node
* Include --flannel-iface and --node-ip as k3s arguments
* Store logs of k3s-init.service as GitHub job artifacts
Timothy Stewart
Richard Holmboe
Balázs Hasprai
Simon Leiner