In a dual-stack OpenShift cluster, ironic-proxy pods on 2 out of 3 ACM hub master nodes only have IPv4 addresses assigned, while the third master node has both IPv4 and IPv6 addresses. This causes failures when customers attempt to discover IPv6-only BMC hosts, as requests originating from nodes without IPv6 addresses fail with error messages indicating inability to locate ISO/IMG image files.

ACM 2.13/ocp 4.18

Consistently reproducible when:

Cluster is configured as dual-stack

Ironic-proxy pods land on master nodes without IPv6 addresses

Attempting to discover IPv6-only BMC hosts

Deploy a dual-stack OpenShift cluster with ACM

Configure IPv6-only BMC hosts

Attempt to discover the BMC hosts through metal3/ironic

Observe failures when requests originate from nodes without IPv6 addresses

Ironic-proxy pods on some master nodes only have IPv4 addresses

BMC discovery fails with errors when using IPv6 addresses

Error messages indicate inability to locate ISO/IMG image files

Node status shows inconsistent IP address assignments across masters

All ironic-proxy pods in a dual-stack cluster should have both IPv4 and IPv6 addresses

BMC discovery should work consistently regardless of which master node handles the request

Cluster should maintain consistent dual-stack networking configuration across all nodes

Customer observations:

IPv4 addresses are issued by DHCP (Infoblox)

Underlying network uses Cisco ACI with RA messages and SLAAC (but using DHCPv6 due to OpenShift limitations)

DHCPv6 service running on a dedicated server

NodeIP configuration seems to initialize before IPv6 address is available

Cluster was installed as dual-stack from beginning (not converted)

Diagnostic data available:

ACM must-gather

OCP must-gather

sosreport from affected nodes

Detailed network configuration information

Relevant error from logs (sensitive data redacted):
{"level":"info","ts":1751443199.2524848,"logger":"controllers.BareMetalHost","msg":"using PreprovisioningImage","baremetalhost":{"name":"test","namespace":"baremetal"},"provisioningState":"provisioning","Image":{"ImageURL":"https://assisted-image-service-multicluster-engine.apps.<redacted-domain>/byapikey/<redacted>","KernelURL":"","ExtraKernelParams":"","Format":"iso"}}
{"level":"info","ts":1751443199.2841182,"logger":"provisioner.ironic","msg":"current provision state","host":"baremetal~test","lastError":"Failed to prepare to deploy. Exception: HTTP POST https://[<redacted-ipv6>]/redfish/v1/Systems/System.Embedded.1/VirtualMedia/1/Actions/VirtualMedia.InsertMedia returned code 500. Base.1.12.GeneralError: A general error has occurred. See ExtendedInfo for more information Extended information: [{'Message': 'Unable to locate the ISO or IMG image file or folder in the network share location because the file or folder path or the user credentials entered are incorrect.', 'MessageArgs': ['https://<redacted-ipv4>:6183/redfish/boot-<redacted-uuid>.iso'], 'MessageArgs@odata.count': 1, 'MessageId': 'IDRAC.2.9.RAC0720', 'RelatedProperties': ['#/Image'], 'RelatedProperties@odata.count': 1, 'Resolution': 'Enter the correct file or folder path and credentials, and then retry the operation.', 'Severity': 'Informational'}]","current":"deploy failed","target":"active"}

Note:

Hostnames (e.g., masterXX.cluster.example.com) replaced with generic terms

IPv4/IPv6 addresses redacted (e.g., 159.216.9.19 → <redacted-ipv4>, 2a13:6203:... → <redacted-ipv6>)

MAC addresses, API keys, and UUIDs removed

Domain names anonymized