A branch that is up but invisible
BR-01 forwards traffic normally, yet the Manager lists it as unreachable and a template change does not reach it. Another branch on the same transports looks healthy. The difference is in the control connections: BR-01 has the one to the Controller on both transports, but the one to the Manager is missing. This is why the control connection output is the first command in most SD-WAN troubleshooting: it shows, in a few lines, which parts of the control plane a router can actually reach.
Who connects to whom
| From | To | Used for |
|---|---|---|
| WAN Edge | Validator | First contact and introduction; see the Validator lesson |
| WAN Edge | Controller | OMP: routes, TLOCs, services and policy |
| WAN Edge | Manager | Configuration, monitoring and statistics |
| Controller, Manager | Validator, and each other | Controllers learn about devices and share what they need |
Edges do not build control connections to each other. They build IPsec tunnels, which is a separate mechanism. The Validator connection is mainly a start-up step: after that, an edge normally works through its Controller and Manager connections.
On small screens, swipe sideways or focus the diagram and use the arrow keys.
One per transport
A WAN Edge can connect to the controllers over each of its transports. BR-01 has an MPLS circuit and an internet line, so it can have a Controller connection on each, and one to the Manager. If one transport fails, the control connection on the other keeps the router in the fabric. The settings of the tunnel interface decide which transports are allowed to build them, so a transport can be used for user tunnels without carrying control connections, or the reverse. The next lesson in this unit puts the roles together: control component workflow.
On small screens, swipe sideways or focus the diagram and use the arrow keys.
DTLS and TLS
| DTLS | TLS | |
|---|---|---|
| Runs over | UDP | TCP |
| Reliability | UDP does not retransmit; the protocol copes with loss itself | TCP retransmits and orders data |
| Firewall view | A UDP flow; some networks restrict or time out UDP faster | A TCP connection; often easier to pass or inspect |
| Chosen by | A setting on the devices; both ends must use the same protocol. Check your release for the default and the exact option. | |
Both protect the connection in the same way: each end proves its identity with a certificate, the session is encrypted, and the content cannot be read or altered on the way. If the certificates are not accepted, no connection comes up, whatever the network allows. How certificates are issued and trusted is the topic of a later lesson.
On small screens, swipe sideways or focus the diagram and use the arrow keys.
How a connection is built
On small screens, swipe sideways or focus the diagram and use the arrow keys.
- The edge has a path. The transport interface has an address and a route to the controller's address.
- A DTLS or TLS session starts. The edge connects to the controller on the chosen protocol, and the usual handshake begins.
- Each side shows a certificate. The edge and the controller check each other against the trust they hold. A device that fails this is refused.
- The device is checked against the fabric's device list. A valid certificate for a device that the fabric does not list is also not accepted.
- The connection comes up. The edge then starts OMP with the Controller and exchanges management data with the Manager over their connections.
Reading the output
show sdwan control connectionsOn an IOS XE edge: one line per connection. Look at the peer type, the colour it uses, and the state.
BR-01#show sdwan control connections PEER TYPE PROT PEER SYSTEM IP SITE ID LOCAL COLOR STATE vsmart dtls 10.255.0.11 1 mpls up vsmart dtls 10.255.0.11 1 biz-internet up vmanage dtls 10.255.0.10 1 biz-internet connect
Read it in three passes. Is there a line for each component the router should reach? Is each line up? Is it up on each transport that should carry it? The first line that fails the test is where to start.
On small screens, swipe sideways or focus the diagram and use the arrow keys.
Verify the missing connection and the user path separately
show sdwan control connection-history
show sdwan omp peers
show sdwan bfd sessionsThese read-only checks use IOS XE Catalyst SD-WAN 17.x command style. In connection history, correlate the peer and local colour with the recent failure reason. Check OMP peer state separately, then inspect BFD for the intended edge-to-edge path. An up BFD session verifies that monitored tunnel; test the actual application destination and its service-VPN route as well. Check the selected release for exact output fields. Do not clear healthy connections to obtain evidence.
Technical basis: Cisco control and data-plane troubleshooting and Cisco data-plane monitoring.
Common reasons a connection does not come up
On small screens, swipe sideways or focus the diagram and use the arrow keys.
Compare a failing router with a working one on the same transport, but do not treat that comparison as proof of an identical path. The routers can have different source addresses, NAT mappings or firewall treatment. Use the failing router's history and endpoint reachability to narrow the cause.
On small screens, swipe sideways or focus the diagram and use the arrow keys.
Mistakes to avoid
- Reading a healthy control connection as proof that user traffic works. The data plane has its own tunnels.
- Checking one transport and calling the router healthy. A line can be up on one colour and missing on another.
- Changing the protocol on one end only. DTLS and TLS do not interoperate.
- Ignoring the clock on a new device. A wrong date is a common reason for certificate errors.
Test yourself on control connections
Which pair of devices does NOT normally build a control connection?
What is the main practical difference between DTLS and TLS for a control connection?
In the output above, BR-01's Manager line is in connect while OMP to the Controller is up. What is the effect?
A new edge has a valid route and firewall rules are open, yet no control connection comes up on any transport. The date on the router is two years wrong. What is the best next step?