Hello Proxmox team and community,
We are encountering a critical SDN issue on a production Proxmox VE cluster (9.1.2) involving VNETs backed by a Linux bond interface. The behavior appears to be a regression or edge-case bug triggered by SDN changes on an existing setup.
Environment Overview:
As a temporary mitigation, the issue was resolved using the following steps:
Thank you for your time and support.
We are encountering a critical SDN issue on a production Proxmox VE cluster (9.1.2) involving VNETs backed by a Linux bond interface. The behavior appears to be a regression or edge-case bug triggered by SDN changes on an existing setup.
Environment Overview:
- Proxmox VE: Production cluster (9.1.2 recent stable release)
- NW Configs: Phy NICs -> Bond -> Bridge -> SDN -> ZONE -> VNETS
- SDN enabled
- Linux bonding (bond interface used as SDN uplink)
- Multiple existing VNETs mapped to the same bridge (Approx 90 VNETs)
- VMs actively running on these VNETs
- The bond interface was correctly detected, stable, and fully functional.
- SDN VNETs were already associated with the bond, and:
- VNETs were visible in the UI
- VM connectivity was stable
- No packet loss or SDN-related issues were observed
- Recently, we added an additional VNET to the same existing SDN zone / bond uplink
- Immediately after this change:
- VNETs associated with the bond disappeared
- VMs attached to the affected VNETs lost network connectivity and went down
- SDN configuration became inconsistent/unusable
- Rolling back or restarting services did not restore connectivity
- This is not an initial setup issue — the configuration was working correctly for a significant period
- The failure was triggered by adding a new VNET to an already working bonded SDN configuration
- Impact was immediate and affected running production workloads
- Behavior suggests a state inconsistency or SDN recalculation issue when modifying VNETs on bonded uplinks.
As a temporary mitigation, the issue was resolved using the following steps:
- Performed an apt upgrade to bring Proxmox VE and related SDN packages to the latest available versions (9.1.4).
- After the upgrade, re-applied the SDN configuration via the Proxmox UI (without changing the existing design).
- Once the SDN configuration was re-applied:
- VNETs reappeared correctly
- VM network connectivity was restored
- Bonded uplink functioned as expected
- Is this a known SDN bug related to bonded uplinks, especially when modifying existing VNET configurations? ( PS: I did search the bug list for 9.1.2 could'nt find any)
- Has this issue been addressed or fixed in the latest Proxmox VE releases?
- Are there specific bonding modes (e.g., active-backup vs LACP) recommended or required for SDN?
- Are there any mandatory configuration steps or limitations when:
- Adding new VNETs
- Modifying SDN objects
- Using bonds as SDN uplinks in production
- Is there a safe procedure for extending SDN VNETs on bonded interfaces without disrupting live workloads?
- This is an SDN design limitation.
- A known bug.
- Or a configuration requirement that needs to be adjusted.
Thank you for your time and support.