Description

When performing a Site WAN edit for a system upgraded from CSO 6.1 to 6.2, the WAN interface became unstable upon editing the WAN.

This article discusses the cause and troubleshooting steps to narrow down the issue and how to fix it.

Symptoms

Upon WAN Edit workflow in CSO, WAN interface becomes unusable. NFX running software version contains vSRX as a VM with JDM as component handling underlying Linux and JCP managing Broadcom chip. During Site WAN Edit workflow, CSO  pushed MTU changes to the WAN interface for vSRX VM in Linux.

The following configuration is seen pushed from CSO to NFX JDM:
 
set virtual-network-function vSRX.JNPR0111-5000.Juniper interfaces eth4 mtu 2048

Along with the changes performed by the user during Site WAN Edit.

From the VNF (vSRX in this case) XML dump, there are 2 PCI addressed assigned to the same interface:
 
VNF Interfaces
---------------------------------------------------------------------------------------------
VNF Interface Type Source Model MAC VLAN-ID
-------------------- --------- --------- ------------ ---------- -------------------- -------
vSRX.JNPR0111-5000.Juniper vnet6 network default virtio f4:bf:a8:57:92:2c --
vSRX.JNPR0111-5000.Juniper vnet7 bridge eth0br virtio f4:bf:a8:57:92:2d --
vSRX.JNPR0111-5000.Juniper vSRX.JNPR0111-5000.Juniperia_eth2 vhostuser -- virtio f4:bf:a8:57:92:2e 4010
vSRX.JNPR0111-5000.Juniper vSRX.JNPR0111-5000.Juniperia_eth3 vhostuser -- virtio f4:bf:a8:57:92:2f 4050
vSRX.JNPR0111-5000.Juniper vSRX.JNPR0111-5000.Juniperia_eth4 vhostuser -- virtio f4:bf:a8:57:92:30 4051 <----
vSRX.JNPR0111-5000.Juniper vSRX.JNPR0111-5000.Juniperia_eth5 vhostuser -- virtio f4:bf:a8:57:92:31 --
vSRX.JNPR0111-5000.Juniper vSRX.JNPR0111-5000.Juniperia_eth6 vhostuser -- virtio f4:bf:a8:57:92:32 4002
vSRX.JNPR0111-5000.Juniper vSRX.JNPR0111-5000.Juniperia_eth7 vhostuser -- virtio f4:bf:a8:57:92:33 4003
vSRX.JNPR0111-5000.Juniper vSRX.JNPR0111-5000.Juniperia_eth8 vhostuser -- virtio f4:bf:a8:57:92:34 4052
vSRX.JNPR0111-5000.Juniper vSRX.JNPR0111-5000.Juniperia_eth9 vhostuser -- virtio f4:bf:a8:57:92:35 4053
vSRX.JNPR0111-5000.Juniper vSRX.JNPR0111-5000.Juniperia_eth4 vhostuser -- virtio f4:bf:a8:57:92:30 4051 <---

From the /var/log/syslog, the following messages are seen:
 
Oct 13 18:08:24 vjunos0-node ovs-vswitchd[4202]: ovs|00540|netdev_dpdk|ERR|Removing port 'vSRX.JNPR0111-5000.Juniper_eth4' while vhost device still attached.
Oct 13 18:08:24 vjunos0-node ovs-vswitchd[4202]: ovs|00541|netdev_dpdk|ERR|To restore connectivity after re-adding of port, VM on socket '/var/run/openvswitch/vSRX.JNPR0111-5000.Juniper_eth4' must be restarted.

The vSRX version running on NFX does not support a host swap of MTU config on the virtual interfaces and the changes performed requires a reboot.
 

Solution

Rebooting the NFX will complete the ask of ovs-vswitchd to correctly make the change and assign one unique port to the interface.

Note: This issue is specific to devices that were upgraded from previous version into 6.2. All newly onboarded NFX devices in version 6.2 will have the MTU set to 2048 from the time of onboard and would not encounter this issue.

Modification History

.