| Product | Versa VOS (FlexVNF) |
| Component | BGP / Routing |
| Affected Build | 22.1.4 builds Feb 24, 2026 |
| Fixed In | Next public hotfix 22.1.4 release. |
| Bug References | Bug-ID #144599 |
BGP sessions between Hub-Controller / Controller and downstream VOS nodes flap continuously. Session never reaches Established, stays in Idle.
On show bgp neighbor (spoke side):
Impact: route convergence broken, on control-plane dependent paths.

Two-stage issue in BGP Extended Message Length handling (RFC 8654, 65535 byte cap).
BGP packed UPDATE messages up to full 65535 bytes. Under certain conditions, on-wire frames exceeded transport limits. Peer rejected with Bad Length, flap loop ensued. Not reported from the field; reproduced internally in EVPN scenarios only.
Capped both send AND receive at 65335 (200-byte headroom). Resolved the original issue but introduced backward incompatibility. Older peers still send up to 65535; newer peers reject anything between 65336 and 65535. Bug-ID #144599.
Split into asymmetric caps. Send stays at 65335. Receive restored to 65535. Backward-compatible with older peers.
Affected: any deployment running a Feb 24, 2026 build of 22.1.4 as Hub-Controller / Controller while spokes run older builds. Symptom is fabric-wide BGP flap, control-plane outage.
show bgp neighbor <hub-controller-ip>Upgrade Hub-Controller / Controller to a 22.1.4 build [Build on or after Aprl 6th 2026]. Spoke nodes can stay on current version, fix is receiver-side tolerance.
Disable extended-message-length capability on BGP neighbor. Forces 4096-byte cap on both sides. Functional but slows convergence at scale.
CLI:
set routing-instances <Tenant-Name> protocols bgp <ID> disable-extended-message-length-capability true