feat: add mgmt endpoint & cli command for infra machine reboot #826
Conversation
Introduce a new mgmt endpoint to reboot an infra machine and wait for the reboot request to be processed by the respective infra provider. It does the heavy lifting of modifying/waiting for the resources. Additionally: - In the `InfraMachineController`, to verify that the machine info was collected before powering it off, check the `TalosVersion` on the `MachineStatus` instead of `SecureBootStatus`. It is a more suitable field to check this. - Rename the infra provider label prefix format from being `omni.sidero.dev/infra-provider[provider-id]/key=value` to be `omni.sidero.dev/infra-provider/provider-id/key=value`. Signed-off-by: Utku Ozdemir <[email protected]>
utkuozdemir
force-pushed
the
reboot-mgmt-api
branch
from
2b76852 to
6866536
Compare
|
|
||
| ctx = actor.MarkContextAsInternalActor(ctx) | ||
|
|
||
| if _, err := safe.StateGetByID[*infra.Machine](ctx, s.omniState, req.MachineId); err != nil { |
There was a problem hiding this comment.
Maybe a bit crazy idea, but we could make reboot work seamlessly for all kind of machines.
If the infra machine is not found we can fall back to the talos API call.
What do you think?
There was a problem hiding this comment.
Sure, we can do that. I'm just concerned with the fact that this endpoint does not block or report any status (like stream the phases back or something), so not sure how much value it actually brings.
There was a problem hiding this comment.
Does it block in case of omnictl power reboot?
There was a problem hiding this comment.
nope, none. It blocks until reboot is issued, not until machine comes back.
There was a problem hiding this comment.
so it will be more or less consistent I guess
There was a problem hiding this comment.
another crazy idea is to intercept talosctl reboot command and make it go through the infra provider 🙃
| @@ -55,7 +55,7 @@ func (suite *InfraMachineControllerSuite) TestReconcile() { | |||
| }) | |||
|
|
|||
| machineStatus := omni.NewMachineStatus(resources.DefaultNamespace, "machine-1") | |||
| machineStatus.TypedSpec().Value.SecureBootStatus = &specs.SecureBootStatus{} | |||
| machineStatus.TypedSpec().Value.TalosVersion = "v1.9.1" | |||
| var ( | ||
| machineStatus *infra.MachineStatus | ||
| event safe.WrappedStateEvent[*infra.MachineStatus] | ||
| ) |
There was a problem hiding this comment.
No reason do declare those outside of for { ... } since they are not used outside. Actually if machineStatus, err = event.Resource(); err != nil { can be split and there will be no need to declare machineStatus.
|
|
||
| if machineConfig == nil { | ||
| machineConfig = omnires.NewInfraMachineConfig(resources.DefaultNamespace, req.MachineId) | ||
|
|
|
This PR is stale because it has been open 45 days with no activity. |
Introduce a new mgmt endpoint to reboot an infra machine and wait for the reboot request to be processed by the respective infra provider. It does the heavy lifting of modifying/waiting for the resources.
Additionally:
InfraMachineController, to verify that the machine info was collected before powering it off, check theTalosVersionon theMachineStatusinstead ofSecureBootStatus. It is a more suitable field to check this.omni.sidero.dev/infra-provider[provider-id]/key=valueto beomni.sidero.dev/infra-provider/provider-id/key=value.