Commit Graph

15 Commits

Author SHA1 Message Date
unkinben c39caeb8bb fix: preload correct rke2 airgap images to break CNI bootstrap deadlock
ci/woodpecker/pr/ruby-validate Pipeline was successful
ci/woodpecker/pr/puppet-lint Pipeline was successful
ci/woodpecker/pr/yamllint Pipeline was successful
ci/woodpecker/pr/bolt-validate Pipeline was successful
ci/woodpecker/pr/erb-validate Pipeline was successful
ci/woodpecker/pr/epp-validate Pipeline was successful
ci/woodpecker/pr/ruby-check Pipeline was successful
ci/woodpecker/pr/puppet-validate Pipeline was canceled
New el9_8 nodes boot rke2 v1.33.11 (flannel v0.28.4 / calico v3.31.5) from
the rolling latest/1.33 repo, but the airgap preload pinned v1.33.4
(flannel v0.27.2 / calico v3.30.2), so canal's images were never on disk and
containerd fell back to the artifactapi mirror VIP, which is unreachable until
the flannel overlay it would provide is up. Classic bootstrap deadlock.

- bump pinned rke2_version 1.33.4 -> 1.33.11 so the versionlock, RPM and
  preloaded bundle all match the canal image tags the running binary requests
  (the default rke2-images bundle already contains the canal CNI images)
- wire the airgap archive source to the container_archive_source parameter
  (previously declared but unused) and refresh its default to the artifactapi
  github remote, so the pre-CNI-reachable source is hiera-overridable per node

Claude-Session: https://claude.ai/code/session_01JUoARVdmhxKQHyyyp1pxeT
2026-08-08 18:11:27 +10:00
unkinben aeae26711f Convert RKE2 registries to template, disable default endpoints (#474)
## Summary
- Replace static `registries.yaml` with EPP template driven by `rke2::registries` hash
- Add `disable-default-registry-endpoint: true` to all mirrors — RKE2 will only use artifactapi and never fall back to upstream registries
- Registry configuration now fully managed via hiera data (`roles/infra/k8s.yaml`)

Reviewed-on: #474
Co-authored-by: Ben Vincent <ben@unkin.net>
Co-committed-by: Ben Vincent <ben@unkin.net>
2026-06-29 22:30:48 +10:00
benvin 97d21c81c5 feat: make rke2 registries.yaml conditional on manage_registries (#472)
Add/Remove the registries.yaml file based on the manage_registries
boolean. We are leaving it on default=false now as the artifactapi
server was broken.

---------

Co-authored-by: Ben Vincent <ben@unkin.net>
Reviewed-on: #472
2026-06-27 07:50:31 +10:00
unkinben ceacfc85ae feat: restart rke2 when registries.yaml is deployed (#465)
- ensure we restart rke2 to pickup registries.yaml changes
- add a comment to registries.yaml to force a restart

Reviewed-on: #465
2026-05-06 23:11:20 +10:00
unkinben 0d412aebdb chore: deploy rke2 registries.yaml (#462)
ensure all new docker pulls are actioned through artifactapi

Reviewed-on: #462
2026-05-06 22:17:59 +10:00
unkinben c4d28d52bc chore: remove helm deploys from puppet (#444)
- migrate helm deployments to terraform

Reviewed-on: #444
2026-01-30 20:52:51 +11:00
unkinben 1077bdcbc1 chore: update ceph gpgkey (#438)
- stop checking ceph gpgkey (fixme)
- use artifactapi for retrieving large rke image bundle

Reviewed-on: #438
2026-01-16 23:51:11 +11:00
unkinben efbbb6bcb1 feat: moderate the k8s install (#403)
- only install a base config
- wait for 3 masters before deploying helm charts
- remove cluster-domain
- manage nginx ingres via rke2 helmconfig

Reviewed-on: #403
2025-10-12 17:50:24 +11:00
unkinben a9c959d924 fix: remove unicode from ceph-csi-yaml (#400)
Reviewed-on: #400
2025-09-21 00:41:06 +10:00
unkinben b224cfb516 fix: cattle-system namespace (#399)
- cattle-system namespace is created earlier than helm
- leave namespaces.yaml to manage cattle-system namespace (required
  before installing helm/rancher)

Reviewed-on: #399
2025-09-21 00:21:41 +10:00
unkinben 4c9204858e feat: define node-token from puppet (#398)
- define the token on the bootstrap node too, so node-token is defined
  for new clusters

Reviewed-on: #398
2025-09-20 22:25:56 +10:00
unkinben 571a9b25a7 fix: resolve rke2-server errors (#397)
- kubectl yaml files must not use underscores
- replace unicode hyphen with ascii hyphen

Reviewed-on: #397
2025-09-20 18:40:18 +10:00
unkinben 762f415d2d feat: k8s helm rework (#396)
- remove helm-generated-yaml, replace with helm execs
- template/parameterise ceph csi

Reviewed-on: #396
2025-09-20 17:40:41 +10:00
unkinben 4e77fb7ee7 feat: manage rancher, purelb, cert-manager (#395)
This change will install rancher, purelb and cert-manager, then
configure a dmz and common ip pool to be used by loadbalancers. The
nginx ingres controller is configured to use 198.18.200.0 (common) and
announce the ip from all nodes so that it becomes an anycast ip in ospf.

- manage the install of rancher, purelb and cert-manager
- add rancher ingress routes
- add nginx externalip/loadBalancer

Reviewed-on: #395
2025-09-14 20:59:39 +10:00
unkinben 6e4bc9fbc7 feat: adding rke2 (#394)
- manage rke2 repos
- add rke2 module (init, params, install, config, service)
- split roles::infra::k8s::node -> control/compute roles
- moved common k8s config into k8s.yaml
- add bootstrap_node, manage server and token fields in rke2 config
- manage install of helm
- manage node attributes (from puppet facts)
- manage frr exclusions for service/cluster network

Reviewed-on: #394
2025-09-14 13:27:49 +10:00