VMware Cloud Director hosted applications managed by third party orchestrators experience continuous reboot loops during rebuild operations.

Products

• VMware Cloud Director

Issue/Introduction

• VCD hosted Huawei applications experience continuous reboot loops during LCM rebuildOS operations.
• VMs are assigned duplicate IP addresses by vCenter or fail to present vNIC hardware to the Guest OS after reboot.
• High frequency of recomposeVApp requests observed in VMware Cloud Director debug logs.
• Infrastructure stress indicated by applyDVPortgroup task failures and DvsOperationBulkFault errors in vpxd.log.
• Host agent latencies and RPC timeouts (status code 3) returning Null responses for MoDvSwitch calls.

Environment

• 10.3
• 10.4
• 10.5
• 10.6

Cause

The root cause is a configuration or password mismatch between the tenant application layer and the VCD infrastructure. An administrative change triggers an automated reconciliation loop from the Huawei LCM, which attempts to force the network state back to a legacy configuration. This high frequency API activity creates a race condition that overwhelms the management plane, preventing network changes from being synchronised across cluster members.

Resolution

To stabilise the environment and prevent management plane saturation:

• Ensure the application team verifies the alignment between the LCM database and the VCD actual state before maintenance activities.
• Temporarily disable LCM self-healing logic during major infrastructure changes to prevent automated reconciliation loops.
• Synchronise the LCM database with the VCD actual state prior to triggering rebuild operations.
• If an active conflict exists, manually realign IP addresses in the tenant layer to match the LCM expected state to restore immediate stability.