Documentation of the controller states/conditions and their interaction#264
Documentation of the controller states/conditions and their interaction#264
Conversation
This is a documentation of the various controllers and their interactions.
notandy
left a comment
notandy
left a comment
There was a problem hiding this comment.
The file name is incorrect, since it includes way more than just an diagram.
Please consider splitting parts.
Also please put any docs into a docs/ directory, like it's common with Go project layouts
|
|
||
| ### Lifecycle Controllers | ||
|
|
||
| 1. **HypervisorController** (`hypervisor_controller.go`) |
There was a problem hiding this comment.
Numbered list is a bit confusing since the "order" doesn't matter.
| - **Labels:** `kubernetes.io/hostname`, `topology.kubernetes.io/region`, `topology.kubernetes.io/zone`, `kubernetes.metal.cloud.sap/{bb,cluster,name,type}`, `worker.garden.sapcloud.io/group`, `worker.gardener.cloud/pool` (plus any keys from global LabelSelector) | ||
| - **Annotations → Spec:** `nova.openstack.cloud.sap/aggregates` → `Spec.Aggregates` (comma-split, zone appended), `nova.openstack.cloud.sap/custom-traits` → `Spec.CustomTraits` (comma-split) | ||
| - **Label → Spec:** `cobaltcore.cloud.sap/node-hypervisor-lifecycle` presence → `Spec.LifecycleEnabled=true`, value `"skip-tests"` → `Spec.SkipTests=true` | ||
| - **Status:** Node internal IP → `Status.InternalIP` |
There was a problem hiding this comment.
Could this be split up into one key per line?
| 3. **AggregatesController** (`aggregates_controller.go`) | ||
| - Manages OpenStack aggregate membership | ||
| - Applies different aggregates based on lifecycle phase | ||
| - Coordinates with onboarding and termination flows |
There was a problem hiding this comment.
A short explanation about the related spec and status field would be good
| - `"ha"`: Disables compute service but does **not** create an Eviction CR (the HA service handles evacuation) | ||
|
|
||
| 5. **EvictionController** (`eviction_controller.go`) | ||
| - Executes VM migration off the hypervisor |
There was a problem hiding this comment.
main point missing: reconcilles on evicition CRDs
| 6. **DecommissionController** (`decomission_controller.go`) | ||
| - Handles hypervisor offboarding | ||
| - Cleans up OpenStack service and resource provider | ||
| - Runs during node termination |
There was a problem hiding this comment.
Which a litte bit more information. like how it monitors node termination and what are the steps.
| - Manages readiness signaling deployment | ||
| - Waits for `HaEnabled=False` (set by HypervisorInstanceHaController) before allowing node deletion during offboarding | ||
|
|
||
| ### Auxiliary Controllers |
|
|
||
| ### Auxiliary Controllers | ||
|
|
||
| 8. **HypervisorInstanceHaController** (`hypervisor_instance_ha_controller.go`) |
There was a problem hiding this comment.
this controller should not be in here
| - Creates cert-manager `Certificate` CRs for libvirt TLS | ||
| - Generates RSA 4096-bit certificates covering all node IPs/DNS names | ||
|
|
||
| ## Spec Fields |
There was a problem hiding this comment.
We should generate the spec documentation with go-docs directly from the spec. This is doomed to be outdated quickly
|
|
||
| 11. **NodeCertificateController** (`node_certificate_controller.go`) | ||
| - Creates cert-manager `Certificate` CRs for libvirt TLS | ||
| - Generates RSA 4096-bit certificates covering all node IPs/DNS names |
| | `TraitsUpdated` | Placement API traits synced | TraitsController | `True` — required for onboarding to complete (which sets Ready=True); not directly re-checked at runtime | | ||
| | `HaEnabled` | Instance HA enabled/disabled state | HypervisorInstanceHaController | `True` — required for onboarding to complete when `Spec.HighAvailability=true`; `False` required before GardenerNodeLifecycleController allows node deletion during offboarding | | ||
|
|
||
| ### Condition Reasons |
There was a problem hiding this comment.
not sure if this section is helpful at all. Conditions have human-readable messages for a reason, which should explain the same.
2 participants
This is a documentation of the various controllers and their interactions.