- Version 15.x
- Version 14.x
- Version 13.x
- Version 12.x
- Older Versions
- Available for:
Support for TLS Routing behind layer 7 (HTTP/HTTPS) load balancers and
reverse proxies is available in Preview starting from Teleport
In TLS routing mode Teleport proxy multiplexes all client connections on a single TLS port.
With TLS routing, cluster administrators can simplify network configurations since proxy only listens on one port. All connections are authenticated with mutual TLS and users are able to tunnel protocols that may be blocked on the network such as SSH.
Starting from version
8.0 Teleport proxy listens for all client connections
web_listen_addr by default:
proxy_service: web_listen_addr: "0.0.0.0:443"
All Teleport clients including SSH, web browser, kubectl, database and reverse tunnel clients establish a TLS tunnel to the proxy's web port and indicate the protocol they're requesting using SNI and ALPN TLS extensions.
Upon accepting a new connection, the proxy inspects the SNI/ALPN value in the TLS handshake and forwards the connection to appropriate backend service.
mysql implement TLS handshake as a part of their
protocol-specific connection negotiation phase (aka STARTTLS).
To support these clients, as well as clients that do support TLS but don't allow
setting custom ALPN values, Teleport's
tsh client includes ability to start
a local TLS routing aware proxy.
Such clients connect to the local proxy instead of Teleport proxy directly. The local proxy establishes a TLS connection to the Teleport proxy with the proper SNI/ALPN values set and tunnels the client's connection over it.
In most cases, clients handle TLS routing transparently when establishing
connection. For example,
tsh client starts local proxy and sets appropriate
SNI/ALPN values automatically. For some clients, like native/GUI database
clients instead of
tsh db connect, the user needs to start the local proxy so
these clients can connect to it.
Let's take a look at how each protocol Teleport supports implements TLS routing.
tsh, when connecting to an SSH node, first dials Teleport
proxy over TLS and requests
teleport-proxy-ssh ALPN protocol.
No local proxy is started in this case as
tsh uses this TLS connection as a
transport to establish the SSH connection.
To support standard OpenSSH client, Teleport provides a
tsh proxy ssh command
which can be used as a
tsh proxy ssh establishes a TLS tunnel to Teleport
teleport-proxy-ssh ALPN protocol, which
ssh then connects over.
See the OpenSSH client guide for details on how it's configured.
Reverse tunnel workers within the Teleport Node, Application and Database
Services, as well as for Trusted Clusters, open a TLS tunnel to the cluster's
Proxy Service with the
teleport-reversetunnel ALPN protocol. The workers then
dial SSH over the tunnel, establishing a secure connection.
kubectl uses HTTPS API and TLS handshake to talk to the API
As such, it is not possible to request a custom ALPN protocol using
Instead, Teleport leverages SNI and sets a
ServerName prefixed with
kube-teleport-proxy-alpn. when generating a kubeconfig file during
tsh kube login:
apiVersion: v1 kind: Config clusters: - cluster: certificate-authority-data: ... server: https://proxy.example.com:443 tls-server-name: kube-teleport-proxy-alpn.proxy.example.com name: example
tsh db connect command executes an appropriate database client for the
database you're connecting to.
In TLS routing mode,
tsh starts a local proxy which database client
connections are tunneled through. The local proxy uses ALPN values like
teleport-mysql depending on the database. The proxy is shut down when the
database session ends.
For the native or graphical database clients to work with TLS routing, they must be connecting to the local proxy instead of Teleport proxy directly.
Teleport provides a
tsh proxy db command to launch a local database proxy:
tsh proxy db example-db
See GUI clients guide for a usage example.
Application access, desktop access and Teleport web UI are served by the
Teleport proxy's web listener and don't require a local proxy or any special
ALPN/SNI negotiation. These web connections use standard
protocols for ALPN.
Starting from version
13.0, TLS routing can now be enabled allowing the
Teleport Proxy Service to serve a single port behind a layer 7 load balancer or
It is expected that the layer 7 load balancer or reverse proxy will terminate
TLS with a public certificate, such as using ACM for AWS ALB. This means that
the Proxy Service does not require a Web TLS certificate using
Teleport clients automatically detect whether the Teleport Proxy Service is
behind a layer 7 load balancer or a reverse proxy. In such cases, the client
initiates a connection upgrade (the same principle used by
WebSocket) and then
sends the TLS routing request through the upgraded connection.
Non-Teleport clients should require local proxies that can perform the special connection upgrades.
Let's take a closer look at how each protocol functions in this configuration.
When transporting the SSH protocol over TLS routing,
tsh make connection
upgrades seamlessly. This is applicable to
tsh ssh/scp commands, as well as
tsh proxy ssh when connected via
ProxyCommand using OpenSSH clients.
tsh proxy kube command creates a local proxy and an ephemeral kubeconfig
for Kubernetes clients like
kubectl. The local proxy generates a self-signed
certificate to secure local communications with the Kubernetes clients.
When forwarding the requests to the Proxy Service, the local proxy performs necessary connection upgrades and sets the required SNI for the TLS handshake.
tsh kubectl and
tsh kube exec commands also start a local proxy
automatically when connection upgrades are required.
In TLS routing mode,
tsh db connect starts a local proxy which database
client connection is tunneled through. The local proxy initiates a connection
to the Proxy Service with the connection upgrade then uses database specific
ALPN values for the TLS handshake.
Similarly, native and GUI clients can connect through
tsh proxy db which
starts a local proxy that handles the connection upgrades.
The Teleport Web UI is fully functional with standard browsers, without any special ALPN/SNI values or connection upgrades.
For both HTTP and TCP apps,
tsh proxy app can launch a local proxy that
handles the connection upgrades and sets appropriate ALPN value for TLS
tsh CLI commands for accessing Cloud APIs, e.g.,
tsh aws, transparently
start a local proxy that performs connection upgrades for TLS routing. To start
local proxies for native applications, you can use
tsh proxy aws.
proxy_service.trust_x_forwarded_for is set to
true, the Proxy service
will take the client source IPs from the "X-Forwarded-For" headers set by the
load balancer or the reverse proxy. This also applies to TLS routing requests
that utilize connection upgrades, since they are essentially HTTP requests.
To prevent IP spoofing, only a single IP address is expected in the "X-Forwarded-For" headers per request. Any requests with multiple IP addresses will be rejected.
Yes. When the Teleport Proxy Service is behind a layer 4 load balancer that terminates TLS, Teleport clients handle the situation similarly to when a layer 7 load balancer is present, performing the connection upgrades.
Note that the load balancer must forward the TLS layer to the Teleport Proxy Service. For example, an AWS Network Load Balancer (NLB) must use "TLS" protocol for the target group.
TLS routing does support the Teleport Proxy Service behind layer 7 reverse proxies. The reverse proxy must support custom connection upgrades and long-lived connections.
You can set up a Teleport Proxy Service behind the reverse proxy and run the following command to test the connection:
$ curl -v https://<proxy.example.com>/webapi/connectionupgrade -H "Connection: Upgrade" -H "Upgrade: alpn-ping" --no-alpn
For a successful upgrade, the server should return "101 Switching Protocols"
with the requested upgrade type and you should also get some binary output in
about 30 seconds running the
< HTTP/1.1 101 Switching Protocols < Connection: Upgrade < Upgrade: alpn-ping < X-Teleport-Upgrade: alpn-ping < Warning: Binary output can mess up your terminal.
Local proxies are required when the Teleport Proxy Service is behind layer 7 load balancers or reverse proxies.
One alternative to
tsh proxy kube is to use
tsh kubectl which starts a
local proxy automatically when connection upgrades are required.
To completely avoid local proxies, either use layer 4 load balancers in TCP
mode instead of layer 7 load balancers, or configure your Teleport Proxy
separate port mode and use a separate Kubernetes port that is not
behind layer 7 load balancers.