MX 240 platform. Wide Area VPLS instance for Cisco IP phones. lt-1/0/10 interface with a vpls leg in the VPLS instance and an inet leg in the VRF instance facing the DHCP server. The lt inet interface is the dhcp relay interface under the vrf forwarding options. Cisco IP phones do not pull an IP address from the DHCP server, but Windows machines in the same vlan pull an IP every time.
I'd also like to point out that this problem appeared out of no where a few weeks ago and was working fine for 2 years. No apparent change to the Juniper devices and according to the phone person, no changes to the phones either.
What makes this bizarre is I can actually see dhcp relay bindings, but the phone is unreachable. When I check the lldp neighbor details from the 4200 switch, the phone shows 0.0.0.0. To add to the oddity, if I static IP the phone I can not ping from an SVI on the 4200 and Wireshark show no ARP response from the phone. If I bounce the phone, it will respond to ping but will quickly stop responding to ping and ARP. If I delete the vlan from the trunk facing the MX commit and bounce the phone again, the phone will respond to ping from the 4200 SVI all day long. Once I add the vlan to trunk, it will stop responding the ping and ARP again! A Windows machine does not exibit this behavior in the same vlan.
I have tried multiple phones.
Does anyone have any ideas on this one? I'm sure anyone reading this is like...hmm....that's a new one!