BaremetalMachine stuck in Pending, InventoryAllocated=False / Reason=PoolMissing | The machineInventoryPoolRef.name of the template points to a pool that does not exist, for example because the pool was deleted after create. | The pool name and namespace. |
BaremetalMachine stuck in Pending, InventoryAllocated=False / Reason=PoolExhausted | The pool has no Available inventory left. | MachineInventoryPool.status.available, and the allocation annotations on each member. |
BaremetalMachine stuck after allocation, BootstrapReady=False / Reason=BootstrapWaiting | The KubeadmConfig has not produced a bootstrap data Secret yet. | Machine.spec.bootstrap.dataSecretName. |
BaremetalMachine waits in Allocated, ImageResolved=False / Reason=ImageCatalogMiss | elemental-image-catalog has no key for Machine.spec.version on the architecture and OS of the host. | ac -n cpaas-system get cm elemental-image-catalog -o yaml. Add the key as described in Add the Image Catalog Keys. The BaremetalMachine then continues by itself. |
BaremetalMachine waits in Allocated, ImageResolved=False / Reason=RegistryConfigPending | The registry configuration of the cluster is not ready yet. The platform creates it with the cluster. | If the reason does not change within a few minutes, contact technical support. |
BaremetalMachine waits in Allocated, ImageResolved=False / Reason=ImagePullUnauthorized | The registry of the cluster refused its registry account. | If the username or password is wrong, contact technical support. Otherwise, allow the account to pull the OS images; see Image Registry. The BaremetalMachine then continues by itself. |
BaremetalMachine waits in Allocated, ImageResolved=False / Reason=ImageNotFound | The registry of the cluster does not have the OS image. | Push it as described in Push the OS Images. The BaremetalMachine then continues by itself. |
BaremetalMachine waits in Allocated, ImageResolved=False / Reason=RegistryUnavailable | The global cluster cannot reach the registry of the cluster. The message gives the error. | The network from the global cluster to the registry, and that the registry answers over HTTPS. The BaremetalMachine continues once the registry answers. |
BaremetalMachine waits in Allocated, ImageResolved=False / Reason=ImageRegistryMissing | The provider finds no registry address for the cluster. | Contact technical support. |
BaremetalMachine is Failed, Ready=False / Reason=OSFamilyUnknown | The os.baremetal.alauda.io/id label of the MachineInventory is not of the form alaudaos.m.<version> or alaudaos.k.<version>, so no plan is written. | The label on the MachineInventory and in machineInventoryLabels of the MachineRegistration; see Create the Registration and the ISO. Then contact technical support. |
A host does not register or install: no MachineInventory appears, it stays in status.install.phase=AwaitingTarget, or it lists a disk in status.install.staleSystemDisks with action: Blocked | The host cannot reach the global cluster, it waits for you to select its system disk, or one of its disks has system partitions from an earlier installation that cannot be wiped safely. | Troubleshooting in Register Hosts. |
An inventory is registered but not counted as Available | The inventory has no plan Secret or is not Ready. | MachineInventory.status.conditions and status.plan.secretRef. |
A pool reports MembersValid=False | An inventory listed in the pool does not exist or belongs to another pool. | The MembersValid message, which names the inventory. |
A MachineInventory is Ready=False / Reason=IdentityConflict | An installed host registered with the identity of this inventory but from different hardware, for example a host started from a copy of the system disk of another host. The inventory is not allocated to a new machine, and no system disk can be selected for it, until the conflict is settled. The elemental.cattle.io/identity-conflict annotation names the system UUID and the address of the host that registered. | If the main board of a host with elemental.cattle.io/auth: emulated-tpm was replaced, see the next row. Otherwise, decide which host keeps the inventory. Shut down the other host, or reinstall it from the ISO so that it registers with its own identity. If the host that keeps the inventory is not the one named in the annotation, remove the elemental.cattle.io/identity-conflict annotation. Otherwise, annotate the inventory with elemental.cattle.io/allow-identity-rebind=<system-uuid> of that host, then restart it; both annotations are removed when it registers. |
The main board of a host whose MachineInventory has the annotation elemental.cattle.io/auth: emulated-tpm was replaced, and the inventory is Ready=False / Reason=IdentityConflict | Its hardware no longer matches the MachineInventory. | Annotate the MachineInventory with elemental.cattle.io/allow-identity-rebind=true, then restart the host. The annotation is removed once it has been used. |
The TPM or the main board of a host whose MachineInventory has the annotation elemental.cattle.io/auth: tpm was replaced | The installed system can no longer register with its MachineInventory. | Contact technical support. |
Reprovisioning of a host never completes, and MachineInventory.status.plan.state=Failed | Writing the OS image failed: the registry is unreachable, TLS fails, or the disk is full. | BaremetalMachine.status.conditions, journalctl -u elemental-system-agent on the host, and the serial console of the host. On Alauda OS KU hosts, the failed-output key of the plan Secret also holds the output of the disk space check. Read these before you delete the Machine; see Replacing a Single Failed Node. |
BaremetalCluster uses Internal, but the cluster cannot be reached through the VIP (healthz does not print ok), or no Alive Pod runs on a control plane node | Another cluster already uses the same vrid, the VIP is not in the control plane Layer 2 domain, or VRRP or gratuitous ARP is blocked. | The Alive Pods in kube-system of the workload cluster, and ip addr show on a control plane host. |
BaremetalCluster uses External, but the endpoint is not reachable | The listener, backend membership, TLS passthrough, health check, route, DNS, or certificate SAN is wrong. | The load balancer configuration, and HTTPS /healthz on each backend. |