During cluster deployment with FCoE storage, the task fails with the error: "Checking the Datastore availability".
Reviewing the /var/run/log/vmkernel.log on the affected ESXi host reveals attempts to set an explicit VLAN ID, followed by FIP (FCoE Initialization Protocol) discovery timeouts:
In(182) vmkernel: cpu0:2098488)ql_fcoe:vmhba4:CreateFabric:112:Info: Fabric created: 0x431f7a571bb0In(182) vmkernel: cpu0:2098488)ql_fcoe_spin_lock_init: lock = RportEvListLockIn(182) vmkernel: cpu0:2098488)ql_fcoe:vmhba4:SendFCoEDiscoverySolicitation:1204:Info: Sending FIP discovery for vlan_id = 0x3ea (0x431f7a571bb0In(182) vmkernel: cpu64:2098487)ql_fcoe:vmhba3:SendFCoEVlanSolicitation:1477:Info: Sending FCoEVlanSolicitation request (0x3)In(182) vmkernel: cpu0:2098488)ql_fcoe:vmhba4:SendFCoEDiscoverySolicitation:1204:Info: Sending FIP discovery for vlan_id = 0x3ea (0x431f7a571bb0Wa(180) vmkwarning: cpu64:2098487)WARNING: ql_fcoe:vmhba3:FipVlanTimeoutWork:254: FIP VLAN Max Retries reached, cur vlan and pri: <VLAN_TBD>In(182) vmkernel: cpu64:2098487)ql_fcoe_spin_lock_init: lock = SessionListLockIn(182) vmkernel: cpu64:2098487)ql_fcoe:vmhba3:CreateFabric:112:Info: Fabric created: 0x431f7a59d830In(182) vmkernel: cpu64:2098487)ql_fcoe_spin_lock_init: lock = RportEvListLockIn(182) vmkernel: cpu64:2098487)ql_fcoe:vmhba3:SendFCoEDiscoverySolicitation:1204:Info: Sending FIP discovery for vlan_id = 0x3ea (0x431f7a59d830
The operation eventually fails and destroys the fabric after exhausting the maximum number of retry attempts:
Wa(180) vmkwarning: cpu0:2098488)WARNING: ql_fcoe:vmhba4:FipDiscoveryTimeoutWork:125: Max retry exhausted for Fabric = 0x431f7a571bb0 vlan_id = 0x3eaIn(182) vmkernel: cpu0:2098488)ql_fcoe:vmhba4:StartPortLogout:1236:Info: Sess 0x431f7a58c740 port_id fffffeIn(182) vmkernel: cpu0:2098488)ql_fcoe:vmhba4:CancelExchangeHandling:196:Info: Enter for Sess = 0x431f7a58c740In(182) vmkernel: cpu2:2098489)ql_fcoe:vmhba4:DeleteFabric:123:Info: Fabric 0x431f7a571bb0 000000 destroyed
Although the cluster is successfully created in the inventory, the deployment task cannot complete because the datastore remains inaccessible from the ESXi host and vCenter Server perspective.
This issue can also occur during the initial deployment via the VCF Installer, as there is no native configuration option to define the LLDP listening mode during greenfield bring-up.
VMware Cloud Foundation (VCF) 5.x, 9.x
VMware Cloud Foundation Installer 5.x, 9.x
VMware SDDC Manager 5.x
VMware vCenter Server 8.x
As per Broadcom engineering, the ESXi storage stack does not impact the VLAN configuration for the FCoE solution.
This should be configured either on the switch side or within the vendor-owned FCoE firmware.
The default deployment sets the LLDP value to Both (from the available options: Listen, Advertise, Both), which appears to cause the failure.
A fix for this issue is included in VMware SDDC Manager versions 5.2.4 and 9.1.1. This resolves the issue for Day-2 workflows only. A fix for the VCF Installer will be introduced in a future release.
Workaround:
To mitigate this issue and successfully complete the cluster deployment, perform the following steps:
Initiate the cluster deployment with FCoE storage.
Allow the task to fail during the "Checking the datastore availability" phase. The task details will display the following progress messages:
Description: Checking the datastore availability
Progress Message: Datastore <Datastore_name> of datacenter datacenter_<TBD> is inaccessible
Following the task failure, once the hosts are successfully deployed and joined to the cluster, the Link Layer Discovery Protocol (LLDP) setting on the Distributed Virtual Switch (DVS) must be modified.
In the vSphere Client, navigate to Networking > Select the affected DVS > Configure > Edit > Advanced > Discovery Protocol.
Change the Link Layer Discovery Protocol setting to Listen.
Reboot the affected ESXi hosts. Upon initialization, the Host Bus Adapters (HBAs) will be brought online.
Once the host reboots are complete, return to the SDDC Manager interface and retry the cluster deployment task to finalize the deployment.
Note: If storage remains inaccessible after configuring LLDP to "listen" mode, ensure that physical uplink configurations (such as Cisco VIC profiles) match across all fabric ports, as physical configuration mismatches can trigger similar datastore availability failures.
Alternatively, a proactive workaround can be applied on the SDDC Manager appliance to set the default LLDP Operation Type prior to VDS creation during addCluster or addVI workflows:
Establish an SSH session to the SDDC Manager appliance and switch to the root user.
Append the desired LLDP property to the Domain Manager application properties file by executing the following command:
echo vcf.vds.lldp.operationtype="listen" >> /etc/vmware/vcf/domainmanager/application.properties
Note: The accepted values for this property are both, listen, or advertise.
Restart the Domain Manager service to apply the changes:
systemctl restart domainmanager
Caution: The property value is case-sensitive. You must use the exact lowercase value (i.e., listen, not LISTEN). If an incorrect value is set, the Domain Manager service will fail to restart. If a restart failure occurs, correct the value or completely remove the newly added entry from /etc/vmware/vcf/domainmanager/application.properties, and then restart the Domain Manager service again.