NodeRef is missing from machines #1062
Comments
|
This should work as long as CAPI has access to the workload cluster. What you should see is that Node resources have a label set by Sidero, and Sidero should set |
|
I do see that Sidero has set a label on the node, but I do not see a The machine configuration shows the correct IP address and there is no indication of communication issues between the management cluster and the workload cluster otherwise. |
|
I still haven't been able to get this to work properly. "conditions": [
{
"lastTransitionTime": "2023-03-18T20:10:06Z",
"status": "True",
"type": "Ready"
},
{
"lastTransitionTime": "2023-03-18T20:09:46Z",
"status": "True",
"type": "BootstrapReady"
},
{
"lastTransitionTime": "2023-03-18T20:10:06Z",
"status": "True",
"type": "InfrastructureReady"
},
{
"lastTransitionTime": "2023-03-18T20:09:46Z",
"reason": "WaitingForNodeRef",
"severity": "Info",
"status": "False",
"type": "NodeHealthy"
}
],The IP addresses don't seem to match. Could this be the problem? I don't think there's really much I can do about it as the machines are assigned a DHCP IP when networking booting, but have a static IP in the config. Even if I leave DHCP enabled the address is different when the Kubernetes cluster actually comes up, likely because of bonding? |
|
@uhthomas I was having the same problem, namely no After some investigation, I noticed the following spammed errors in my So the root cause was that Now all of my In my case, the issue was that I'm using the new KubePrism feature in Talos v1.5.0, so the Given when you filed this issue, it seems unlikely that my specific issue is the same as the one you're seeing, but my advice would be to check your (By the way, as it took me awhile to figure out what |
|
@dhess thanks for looking into it. I believe the original issue is unrelated, as it happened before Talos 1.5.0, but I will take a look into why KubePrism endpoint got into the |
|
@smira To be clear, I set |
|
@dhess I'm more confused then. When KubePrism is enabled, Talos uses that automatically for Kubernetes components running on the host network. You don't need to do anything specifically for it. Your external endpoint should still be "external". |
|
@smira OK, so to be clear, when using |
|
@dhess yes, KubePrism only needs to be enabled. if you're installing e.g. Cilium, you can point it to the KubePrism endpoint (and same for any other host pods which need access to the Kubernetes API). |
|
Here's another issue I encountered where This is because I added the DNS name ( I've tested this with multiple configs, and the only way I can get the Talos CAPI providers to properly resolve the cluster configuration is if I use an IP address in |
I'm running into an issue with
clusterctl movewhere it thinks the nodes are not provisioned. This is not the case as I am the Kubernetes cluster is running and healthy.This error message leads to here which suggests the machines are missing a
NodeRef? Why? What information can I give you to help understand this? The logs don't show anything interesting from what I can see.Originally raised here.
The text was updated successfully, but these errors were encountered: