Troubleshooting Epipe Flaps in MPLS: A Deep Dive into T-LDP Session Instability
Overview:
Frequent T-LDP session flaps between two associated PE routers configured with an Epipe service may not only be a symptom of control-plane or transport-layer instability but also can have significant effects on L2 VPN service reliability. As the targeted version of the Label Distribution Protocol, T-LDP creates sessions across directed TCP connections (often over port 646) between non directly connected peers to share labels for Epipe and others. When these sessions flap it typically is indicative of problems such as intermittent IP connectivity, physical interface flaps or routing divergences along the path. Common cause of this issue is link layer issues (such as interface flaps caused by faulty cables or transceivers), Routing protocol instability is causing temporary reachability loss between peers, Failed Adaptive Link Restoration where a connection at the Hardware level to not sit idle for long else will remove from routing table, and misconfigured LDP hello and hold timers in early session timeout. Session flapping can also be caused by network firewalls or intermediate devices that dynamically block (through Access Control Lists, for example) or filter TCP port 646. Control-plane CPU starvation on PE routers, invalid MPLS MTU configurations or LDP label exhaustion may result in dropping or late-processing T-LDP keepalives and hellos causing continuous recreations of sessions. The effect of this instability is non-negligible — every round of flap will drop and re-establish the Epipe pseudo wire, resulting in temporary service outages, MAC learning events, convergence times and general misery for the customer base because poor service delivery. These disruptions affect delivery of critical Layer 2 services and can potentially cascade into more significant control-plane problems without immediate mitigation. Therefore T-LDP session stability can largely be influenced by all appropriate configuration as well as the physical link integrity checks in place, IP routing, CPU and resource monitoring and is key to a high-availability MPLS-based Epipe infrastructure.
How Epipe is working?
The specific behavior of the Epipe service is that it doesn't do any MAC learning. Similarly, a local Epipe service is defined as two SAPs on the same node whereas a distributed Epipe service is defined as two SAPs on separate nodes. SDPs aren't used with Epipe services at local locations. All of the SAP configuration has a dedicated port/channel on which that service traffic comes from customer side (access side) into the router as opposed to going out. Each port is assigned an encapsulation type. If the port is configured to use IEEE 802.1Q (dot1q) encapsulation, an encapsulation value (ID) must be configured.
Let's deep and dive by one scenario ---
Scenario: David is working as a NOC Engineer at BT Telecom. As per a request from Federal Bank, an Epipe service was configured between two sites—Cardiff and Manchester. Although the SDP and LSP are up and stable, the Epipe service is continuously flapping.
After performing detailed troubleshooting, it was identified that the root cause of the issue is a flapping TLDP session between the devices. Since Epipe services rely on a stable targeted LDP (TLDP) session for pseudo wire establishment, any instability in TLDP directly impacts the service state.
.png)
.png)
Step-1 configure router interface at Cardiff-R1.
Cardiff-R1>configure router
#interface "to-Machester"
address 10.1.1.1/30
port 1/1/1
no shutdown
exitStep-2 configure router port towards Manchester.
Cardiff-R1>
config>port 1/1/1# ethernet mode access
config>port 1/1/1# ethernet encapsulation dot1q
config>port 1/1/1# no shutdownStep-3 configure IS-IS on routers in MPLS backbone at Cardiff-R1
Cardiff-R1>configure>router# isis
net 49.0000.0000.0000.0001.00
level-capability l2
interface "system"
no shutdown
exit
interface "to-Manchester"
no shutdown
exit
no shutdownStep-4 configure MPLS and LDP on routers in MPLS backbone network Cardiff-R1
Cardiff-R1>configure
router
ldp
interface-parameters
interface "to-Manchester"
no shutdown
exit
exit
no shutdown
exit
exitStep-5 configure SDP at Cardiff-R1
*A:Cardiff-R1>config>service# sdp 101 mpls create
*A:Cardiff-R1>config>service>sdp$ description "To_Manchester-R1"
*A:Cardiff-R1>config>service>sdp$ far-end 10.1.1.2
*A:Cardiff-R1>config>service>sdp$ lsp " Cardiff-R1_to_Manchester-R1"
*A:Cardiff-R1>config>service>sdp$
no shutdown
*A:Cardiff-R1>config>service>sdp$
exitStep-6 configure E-pipe service at Cardiff-R1
A:cardiff-R1# configure service epipe 89110
A:cardiff-R1>config>service>epipe# info
----------------------------------------------
sap 1/1/1:100 create
ingress
qos 10002
exit
egress
qos 10002
exit
collect-stats
no shutdown
exit
spoke-sdp 101:89110 create
collect-stats
no shutdown
exit
no shutdown=================================================================================================
Step-7 configure router interface at Manchester-R1
Manchester-R1> configure router
interface "to-Cardiff-R1"
address 10.1.1.2/30
port 1/1/1
no shutdown
exit
Step-8 configure router port towards Cardiff-R1
Manchester-R1>
config>port 1/1/1# ethernet mode access
config>port 1/1/1# ethernet encapsulation dot1q
config>port 1/1/1# no shutdownStep-9 configure IS-IS on router in MPLS backbone at Manchester-R1
Manchester-R1> configure>router# isis
net 49.0000.0000.0000.0001.00
level-capability l2
interface "system"
no shutdown
exit
interface "to-Cardiff-R1"
no shutdown
exit
no shutdownStep-10 configure MPLS and LDP on routers in MPLS backbone network Manchester-R1
Manchester-R1> configure
router
ldp
interface-parameters
interface "to-cardiff-R1"
no shutdown
exit
exit
no shutdown
exit
exitStep-11 configure SDP on Manchester-R1
*B:Manchester-R1>config>service# sdp 101 mpls create
*B:Manchester-R1>config>service>sdp$ description "To_Cardiff-R1"
*B:Manchester-R1>config>service>sdp$far-end 10.1.1.1
*B:Manchester-R1>config>service>sdp$ lsp " Manchester-R1_to_ Cardiff-R1"
*B:Manchester-R1>config>service>sdp$
no shutdown
*B:Manchester-R1>config>service>sdp$
exitStep-12 configure E-pipe service at Manchester-R1
B:Manchester-R1# configure service epipe 89110
B:Manchester-R1>config>service>epipe# info
----------------------------------------------
sap 1/1/1:100 create
ingress
qos 10002
exit
egress
qos 10002
exit
collect-stats
no shutdown
exit
spoke-sdp 101:89110 create
collect-stats
no shutdown
exit
no shutdown
How to Troubleshoot this issue?
Check all Basic protocol status at cardiff-R1.
- Check LDP status between two routers adjacencies.
show router ldp session
show router ldp discovery
- Check RSVP Status
Show router rsvp neighbor
Show router rsvp session
Show router rsvp interface 3.Check LSP status
Show router mpls lsp
Show router mpls lsp " Manchester-R1_to_ Cardiff-R1" detail
- Check TLDP session
Show router tldp session
Show log log-id 99
Show log log-id 101
Show router tldp events
show router tldp session detail
- Check SDP of router
Show service sdp 101 detail
Check SDP should be up and far-end IP should be correct
Ping 10.1.1.1
- Check Filter/ TCP blocking of firewall
Show filter ip- MTU of interface
Show router interface 8.No route to loopback – check IGP reachability
Show router route
Show router isis adjacency detail
Show router isis database
===============================================================================
Check all Basic protocol status at Manchester-R1.
- Check LDP status between two routers adjacencies.
show router ldp session
show router ldp session detail
show router ldp discovery- Check RSVP Status
Show router rsvp neighbor
Show router rsvp session
Show router rsvp interface - Check LSP status
Show router mpls lsp
Show router mpls lsp " Cardiff-R1_to_ Manchester-R1" detail
- Check TLDP session
Show router tldp session
Show log log-id 99
Show log log-id 101
Show router tldp events
show router tldp session detail
- MTU of interface
Show router interface
- Check SDP of router
Show service sdp 101 detail
Check SDP should be up and far-end IP should be correct
Ping 10.1.1.2
7 Check Filter/ TCP blocking of firewall
Show filter ip
8 No route to loopback – check IGP reachability
Show router route
Show router isis adjacency detail
Show router isis database
=============================================================================================
1.TLDP session detail at Cardiff-R1
A:Cardiff-R1 # show router tldp session detail
===============================================================================
TLDP Session Detail
===============================================================================
Peer Address : 10.1.1.2
Peer Transport Addr: 10.1.1.2
Local Transport Addr: 10.1.1.1
Session State : Down
Last Up Time : 2026/05/01 19:45:12
Last Down Time : 2026/05/01 19:46:05
Last Down Reason : TCP connection reset by peer
Up/Down Count : 12
Hello Interval : 5 seconds
Hello Hold Time : 15 seconds
Keepalive Timer : Expired
TCP Port : 646
Admin State : Enabled
===============================================================================
2. TLDP session detail at Manchester-R1
B:Manchester-R1 # show router tldp session detail
===============================================================================
TLDP Session Detail
===============================================================================
Peer Address : 10.1.1.1
Peer Transport Addr: 10.1.1.1
Local Transport Addr: 10.1.1.2
Session State : Down
Last Up Time : 2026/05/01 19:45:12
Last Down Time : 2026/05/01 19:46:05
Last Down Reason : TCP connection reset by peer
Up/Down Count : 12
Hello Interval : 5 seconds
Hello Hold Time : 15 seconds
Keepalive Timer : Expired
TCP Port : 646Admin State : Enabled==============================================================================
Conclusion: The SDP and LSP are up, but the T-LDP session is flapping as the underlay network cannot fast converge—the routes to the peer loopback appear intermittently in the routing table, and they are dropped again almost instantly if all packets flow from these routes via a flapped physical interface. This directly affects TCP-based T-LDP (port 646) by causing the resets of sessions. The physical layer should be stabilized by checking the interface error, SFP cleanup (Reseat/Clean optics), and JOJI (if required). Replace SFP with the defective one if flapping continues. Optical Fiber with Low loss between two router sites, because if the attenuation is high, you will not have a link constant, and it drops regularly. Once the physical layer is up and working in a steady state, the IGP routes will be stable and TLDP session will be normalized.

Comments (3)
Great introduction! Looking forward to more HTML5 articles.
Thanks Jane! We have more articles coming soon 🚀
This helped me understand semantic tags better. Thanks!
Could you also write about Canvas API in detail?
Leave a Comment