Routers¶
Connecting Requests to Services
A router is in charge of connecting incoming requests to the services that can handle them. In the process, routers may use pieces of middleware to update the request, or act before forwarding the request to the service.
Configuration Example¶
Requests /foo are Handled by service-foo -- Using the File Provider
## Dynamic configuration
http:
routers:
my-router:
rule: "Path(`/foo`)"
service: service-foo
## Dynamic configuration
[http.routers]
[http.routers.my-router]
rule = "Path(`/foo`)"
service = "service-foo"
Forwarding all (non-tls) requests on port 3306 to a database service
Dynamic Configuration
## Dynamic configuration
tcp:
routers:
to-database:
entryPoints:
- "mysql"
# Catch every request (only available rule for non-tls routers. See below.)
rule: "HostSNI(`*`)"
service: database
## Dynamic configuration
[tcp]
[tcp.routers]
[tcp.routers.to-database]
entryPoints = ["mysql"]
# Catch every request (only available rule for non-tls routers. See below.)
rule = "HostSNI(`*`)"
service = "database"
Static Configuration
## Static configuration
entryPoints:
web:
address: ":80"
mysql:
address: ":3306"
## Static configuration
[entryPoints]
[entryPoints.web]
address = ":80"
[entryPoints.mysql]
address = ":3306"
## Static configuration
--entryPoints.web.address=:80
--entryPoints.mysql.address=:3306
Configuring HTTP Routers¶
The character @
is not authorized in the router name
EntryPoints¶
If not specified, HTTP routers will accept requests from all EntryPoints in the list of default EntryPoints.
If you want to limit the router scope to a set of entry points, set the entryPoints
option.
Listens to Every EntryPoint
Dynamic Configuration
## Dynamic configuration
http:
routers:
Router-1:
# By default, routers listen to every EntryPoints.
rule: "Host(`example.com`)"
service: "service-1"
## Dynamic configuration
[http.routers]
[http.routers.Router-1]
# By default, routers listen to every EntryPoints.
rule = "Host(`example.com`)"
service = "service-1"
Static Configuration
## Static configuration
entryPoints:
web:
address: ":80"
websecure:
address: ":443"
other:
address: ":9090"
## Static configuration
[entryPoints]
[entryPoints.web]
address = ":80"
[entryPoints.websecure]
address = ":443"
[entryPoints.other]
address = ":9090"
## Static configuration
--entryPoints.web.address=:80
--entryPoints.websecure.address=:443
--entryPoints.other.address=:9090
Listens to Specific EntryPoints
Dynamic Configuration
## Dynamic configuration
http:
routers:
Router-1:
# won't listen to entry point web
entryPoints:
- "websecure"
- "other"
rule: "Host(`example.com`)"
service: "service-1"
## Dynamic configuration
[http.routers]
[http.routers.Router-1]
# won't listen to entry point web
entryPoints = ["websecure", "other"]
rule = "Host(`example.com`)"
service = "service-1"
Static Configuration
## Static configuration
entryPoints:
web:
address: ":80"
websecure:
address: ":443"
other:
address: ":9090"
## Static configuration
[entryPoints]
[entryPoints.web]
address = ":80"
[entryPoints.websecure]
address = ":443"
[entryPoints.other]
address = ":9090"
## Static configuration
--entryPoints.web.address=:80
--entryPoints.websecure.address=:443
--entryPoints.other.address=:9090
Rule¶
Rules are a set of matchers configured with values, that determine if a particular request matches specific criteria. If the rule is verified, the router becomes active, calls middlewares, and then forwards the request to the service.
The table below lists all the available matchers:
Rule | Description |
---|---|
Header(`key`, `value`) |
Matches requests containing a header named key set to value . |
HeaderRegexp(`key`, `regexp`) |
Matches requests containing a header named key matching regexp . |
Host(`domain`) |
Matches requests host set to domain . |
HostRegexp(`regexp`) |
Matches requests host matching regexp . |
Method(`method`) |
Matches requests method set to method . |
Path(`path`) |
Matches requests path set to path . |
PathPrefix(`prefix`) |
Matches requests path prefix set to prefix . |
PathRegexp(`regexp`) |
Matches request path using regexp . |
Query(`key`, `value`) |
Matches requests query parameters named key set to value . |
QueryRegexp(`key`, `regexp`) |
Matches requests query parameters named key matching regexp . |
ClientIP(`ip`) |
Matches requests client IP using ip . It accepts IPv4, IPv6 and CIDR formats. |
Backticks or Quotes?
To set the value of a rule, use backticks `
or escaped double-quotes \"
.
Single quotes '
are not accepted since the values are Go's String Literals.
Regexp Syntax
Matchers that accept a regexp as their value use a Go flavored syntax.
Expressing Complex Rules Using Operators and Parenthesis
The usual AND (&&
) and OR (||
) logical operators can be used, with the expected precedence rules,
as well as parentheses.
One can invert a matcher by using the NOT (!
) operator.
The following rule matches requests where:
- either host is
example.com
OR, - host is
example.org
AND path is NOT/traefik
Host(`example.com`) || (Host(`example.org`) && !Path(`/traefik`))
Header and HeaderRegexp¶
The Header
and HeaderRegexp
matchers allow to match requests that contain specific header.
Examples
Match requests with a Content-Type
header set to application/yaml
:
Header(`Content-Type`, `application/yaml`)
Match requests with a Content-Type
header set to either application/json
or application/yaml
:
HeaderRegexp(`Content-Type`, `^application/(json|yaml)$`)
To match headers case-insensitively, use the (?i)
option:
HeaderRegexp(`Content-Type`, `(?i)^application/(json|yaml)$`)
Host and HostRegexp¶
The Host
and HostRegexp
matchers allow to match requests that are targeted to a given host.
These matchers do not support non-ASCII characters, use punycode encoded values (rfc 3492) to match such domains.
If no Host is set in the request URL (e.g., it's an IP address), these matchers will look at the Host
header.
These matchers will match the request's host in lowercase.
Examples
Match requests with Host
set to example.com
:
Host(`example.com`)
Match requests sent to any subdomain of example.com
:
HostRegexp(`^.+\.example\.com$`)
Match requests with Host
set to either example.com
or example.org
:
HostRegexp(`^example\.(com|org)$`)
To match domains case-insensitively, use the (?i)
option:
HostRegexp(`(?i)^example\.(com|org)$`)
Method¶
The Method
matchers allows to match requests sent with the given method.
Example
Match OPTIONS
requests:
Method(`OPTIONS`)
Path, PathPrefix, and PathRegexp¶
These matchers allow matching requests based on their URL path.
For exact matches, use Path
and its prefixed alternative PathPrefix
, for regexp matches, use PathRegexp
.
Path are always starting with a /
, except for PathRegexp
.
Examples
Match /products
but neither /products/shoes
nor /products/
:
Path(`/products`)
Match /products
as well as everything under /products
,
such as /products/shoes
, /products/
but also /products-for-sale
:
PathPrefix(`/products`)
Match both /products/shoes
and /products/socks
with and ID like /products/shoes/57
:
PathRegexp(`^/products/(shoes|socks)/[0-9]+$`)
Match requests with a path ending in either .jpeg
, .jpg
or .png
:
PathRegexp(`\.(jpeg|jpg|png)$`)
Match /products
as well as everything under /products
,
such as /products/shoes
, /products/
but also /products-for-sale
,
case-insensitively:
PathRegexp(`(?i)^/products`)
Query and QueryRegexp¶
The Query
and QueryRegexp
matchers allow to match requests based on query parameters.
Examples
Match requests with a mobile
query parameter set to true
, such as in /search?mobile=true
:
Query(`mobile`, `true`)
To match requests with a query parameter mobile
that has no value, such as in /search?mobile
, use:
Query(`mobile`)
Match requests with a mobile
query parameter set to either true
or yes
:
QueryRegexp(`mobile`, `^(true|yes)$`)
Match requests with a mobile
query parameter set to any value (including the empty value):
QueryRegexp(`mobile`, `^.*$`)
To match query parameters case-insensitively, use the (?i)
option:
QueryRegexp(`mobile`, `(?i)^(true|yes)$`)
ClientIP¶
The ClientIP
matcher allows matching requests sent from the given client IP.
It only matches the request client IP and does not use the X-Forwarded-For
header for matching.
Examples
Match requests coming from a given IP:
ClientIP(`10.76.105.11`)
ClientIP(`::1`)
Match requests coming from a given subnet:
ClientIP(`192.168.1.0/24`)
ClientIP(`fe80::/10`)
Priority¶
To avoid path overlap, routes are sorted, by default, in descending order using rules length. The priority is directly equal to the length of the rule, and so the longest length has the highest priority.
A value of 0
for the priority is ignored: priority = 0
means that the default rules length sorting is used.
Maximum Value
Traefik reserves a range of priorities for its internal routers, the maximum user-defined router priority value is:
(MaxInt32 - 1000)
for 32-bit platforms,(MaxInt64 - 1000)
for 64-bit platforms.
How default priorities are computed
## Dynamic configuration
http:
routers:
Router-1:
rule: "HostRegexp(`[a-z]+\.traefik\.com`)"
# ...
Router-2:
rule: "Host(`foobar.traefik.com`)"
# ...
## Dynamic configuration
[http.routers]
[http.routers.Router-1]
rule = "HostRegexp(`[a-z]+\\.traefik\\.com`)"
# ...
[http.routers.Router-2]
rule = "Host(`foobar.traefik.com`)"
# ...
In this case, all requests with host foobar.traefik.com
will be routed through Router-1
instead of Router-2
.
Name | Rule | Priority |
---|---|---|
Router-1 | HostRegexp(`[a-z]+\.traefik\.com`) |
34 |
Router-2 | Host(`foobar.traefik.com`) |
26 |
The previous table shows that Router-1
has a higher priority than Router-2
.
To solve this issue, the priority must be set.
Set priorities -- using the File Provider
## Dynamic configuration
http:
routers:
Router-1:
rule: "HostRegexp(`[a-z]+\\.traefik\\.com`)"
entryPoints:
- "web"
service: service-1
priority: 1
Router-2:
rule: "Host(`foobar.traefik.com`)"
entryPoints:
- "web"
priority: 2
service: service-2
## Dynamic configuration
[http.routers]
[http.routers.Router-1]
rule = "HostRegexp(`[a-z]+\\.traefik\\.com`)"
entryPoints = ["web"]
service = "service-1"
priority = 1
[http.routers.Router-2]
rule = "Host(`foobar.traefik.com`)"
entryPoints = ["web"]
priority = 2
service = "service-2"
In this configuration, the priority is configured to allow Router-2
to handle requests with the foobar.traefik.com
host.
RuleSyntax¶
Optional, Default=""
In Traefik v3 a new rule syntax has been introduced (migration guide).
ruleSyntax
option allows to configure the rule syntax to be used for parsing the rule on a per-router basis.
This allows to have heterogeneous router configurations and ease migration.
The default value of the ruleSyntax
option is inherited from the defaultRuleSyntax
option in the static configuration.
By default, the defaultRuleSyntax
static option is v3
, meaning that the default rule syntax is also v3
.
Set rule syntax -- using the File Provider
## Dynamic configuration
http:
routers:
Router-v3:
rule: HostRegexp(`[a-z]+\\.traefik\\.com`)
ruleSyntax: v3
Router-v2:
rule: HostRegexp(`{subdomain:[a-z]+}.traefik.com`)
ruleSyntax: v2
## Dynamic configuration
[http.routers]
[http.routers.Router-v3]
rule = "HostRegexp(`[a-z]+\\.traefik\\.com`)"
ruleSyntax = v3
[http.routers.Router-v2]
rule = "HostRegexp(`{subdomain:[a-z]+}.traefik.com`)"
ruleSyntax = v2
apiVersion: traefik.io/v1alpha1
kind: IngressRoute
metadata:
name: test.route
namespace: default
spec:
routes:
# route v3
- match: HostRegexp(`[a-z]+\\.traefik\\.com`)
syntax: v3
kind: Rule
# route v2
- match: HostRegexp(`{subdomain:[a-z]+}.traefik.com`)
syntax: v2
kind: Rule
In this configuration, the ruleSyntax is configured to allow Router-v2
to use v2 syntax,
while for Router-v3
it is configured to use v3 syntax.
Middlewares¶
You can attach a list of middlewares to each HTTP router. The middlewares will take effect only if the rule matches, and before forwarding the request to the service.
The character @
is not authorized in the middleware name.
Middlewares order
Middlewares are applied in the same order as their declaration in router.
With a middleware -- using the File Provider
## Dynamic configuration
http:
routers:
my-router:
rule: "Path(`/foo`)"
# declared elsewhere
middlewares:
- authentication
service: service-foo
## Dynamic configuration
[http.routers]
[http.routers.my-router]
rule = "Path(`/foo`)"
# declared elsewhere
middlewares = ["authentication"]
service = "service-foo"
Service¶
Each request must eventually be handled by a service, which is why each router definition should include a service target, which is basically where the request will be passed along to.
In general, a service assigned to a router should have been defined, but there are exceptions for label-based providers. See the specific docker documentation.
The character @
is not authorized in the service name.
HTTP routers can only target HTTP services (not TCP services).
TLS¶
General¶
When a TLS section is specified, it instructs Traefik that the current router is dedicated to HTTPS requests only (and that the router should ignore HTTP (non TLS) requests). Traefik will terminate the SSL connections (meaning that it will send decrypted data to the services).
Configuring the router to accept HTTPS requests only
## Dynamic configuration
http:
routers:
Router-1:
rule: "Host(`foo-domain`) && Path(`/foo-path/`)"
service: service-id
# will terminate the TLS request
tls: {}
## Dynamic configuration
[http.routers]
[http.routers.Router-1]
rule = "Host(`foo-domain`) && Path(`/foo-path/`)"
service = "service-id"
# will terminate the TLS request
[http.routers.Router-1.tls]
Routers for HTTP & HTTPS
If you need to define the same route for both HTTP and HTTPS requests, you will need to define two different routers: one with the tls section, one without.
HTTP & HTTPS routes
## Dynamic configuration
http:
routers:
my-https-router:
rule: "Host(`foo-domain`) && Path(`/foo-path/`)"
service: service-id
# will terminate the TLS request
tls: {}
my-http-router:
rule: "Host(`foo-domain`) && Path(`/foo-path/`)"
service: service-id
## Dynamic configuration
[http.routers]
[http.routers.my-https-router]
rule = "Host(`foo-domain`) && Path(`/foo-path/`)"
service = "service-id"
# will terminate the TLS request
[http.routers.my-https-router.tls]
[http.routers.my-http-router]
rule = "Host(`foo-domain`) && Path(`/foo-path/`)"
service = "service-id"
options
¶
The options
field enables fine-grained control of the TLS parameters.
It refers to a TLS Options and will be applied only if a Host
rule is defined.
Server Name Association
Even though one might get the impression that a TLS options reference is mapped to a router, or a router rule,
one should realize that it is actually mapped only to the host name found in the Host
part of the rule.
Of course, there could also be several Host
parts in a rule, in which case the TLS options reference would be mapped to as many host names.
Another thing to keep in mind is: the TLS option is picked from the mapping mentioned above and based on the server name provided during the TLS handshake, and it all happens before routing actually occurs.
Domain Fronting
In the case of domain fronting,
if the TLS options associated with the Host Header and the SNI are different then Traefik will respond with a status code 421
.
Configuring the TLS options
## Dynamic configuration
http:
routers:
Router-1:
rule: "Host(`foo-domain`) && Path(`/foo-path/`)"
service: service-id
# will terminate the TLS request
tls:
options: foo
tls:
options:
foo:
minVersion: VersionTLS12
cipherSuites:
- TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384
- TLS_ECDHE_ECDSA_WITH_CHACHA20_POLY1305_SHA256
- TLS_ECDHE_RSA_WITH_CHACHA20_POLY1305_SHA256
- TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256
- TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256
## Dynamic configuration
[http.routers]
[http.routers.Router-1]
rule = "Host(`foo-domain`) && Path(`/foo-path/`)"
service = "service-id"
# will terminate the TLS request
[http.routers.Router-1.tls]
options = "foo"
[tls.options]
[tls.options.foo]
minVersion = "VersionTLS12"
cipherSuites = [
"TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384",
"TLS_ECDHE_ECDSA_WITH_CHACHA20_POLY1305_SHA256",
"TLS_ECDHE_RSA_WITH_CHACHA20_POLY1305_SHA256",
"TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256",
"TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256",
]
Conflicting TLS Options
Since a TLS options reference is mapped to a host name,
if a configuration introduces a situation where the same host name (from a Host
rule) gets matched with two TLS options references,
a conflict occurs, such as in the example below:
## Dynamic configuration
http:
routers:
routerfoo:
rule: "Host(`snitest.com`) && Path(`/foo`)"
tls:
options: foo
routerbar:
rule: "Host(`snitest.com`) && Path(`/bar`)"
tls:
options: bar
## Dynamic configuration
[http.routers]
[http.routers.routerfoo]
rule = "Host(`snitest.com`) && Path(`/foo`)"
[http.routers.routerfoo.tls]
options = "foo"
[http.routers]
[http.routers.routerbar]
rule = "Host(`snitest.com`) && Path(`/bar`)"
[http.routers.routerbar.tls]
options = "bar"
If that happens, both mappings are discarded, and the host name (snitest.com
in this case) for these routers gets associated with the default TLS options instead.
certResolver
¶
If certResolver
is defined, Traefik will try to generate certificates based on routers Host
& HostSNI
rules.
## Dynamic configuration
http:
routers:
routerfoo:
rule: "Host(`snitest.com`) && Path(`/foo`)"
tls:
certResolver: foo
## Dynamic configuration
[http.routers]
[http.routers.routerfoo]
rule = "Host(`snitest.com`) && Path(`/foo`)"
[http.routers.routerfoo.tls]
certResolver = "foo"
Multiple Hosts in a Rule
The rule Host(`test1.example.com`) || Host(`test2.example.com`)
will request a certificate with the main domain test1.example.com
and SAN test2.example.com
.
domains
¶
You can set SANs (alternative domains) for each main domain. Every domain must have A/AAAA records pointing to Traefik. Each domain & SAN will lead to a certificate request.
## Dynamic configuration
http:
routers:
routerbar:
rule: "Host(`snitest.com`) && Path(`/bar`)"
tls:
certResolver: "bar"
domains:
- main: "snitest.com"
sans:
- "*.snitest.com"
## Dynamic configuration
[http.routers]
[http.routers.routerbar]
rule = "Host(`snitest.com`) && Path(`/bar`)"
[http.routers.routerbar.tls]
certResolver = "bar"
[[http.routers.routerbar.tls.domains]]
main = "snitest.com"
sans = ["*.snitest.com"]
ACME v2 supports wildcard certificates.
As described in Let's Encrypt's post wildcard certificates can only be generated through a DNS-01
challenge.
Most likely the root domain should receive a certificate too, so it needs to be specified as SAN and 2 DNS-01
challenges are executed.
In this case the generated DNS TXT record for both domains is the same.
Even though this behavior is DNS RFC compliant,
it can lead to problems as all DNS providers keep DNS records cached for a given time (TTL) and this TTL can be greater than the challenge timeout making the DNS-01
challenge fail.
The Traefik ACME client library lego supports some but not all DNS providers to work around this issue.
The supported provider
table indicates if they allow generating certificates for a wildcard domain and its root domain.
Wildcard certificates can only be verified through a DNS-01
challenge.
Double Wildcard Certificates
It is not possible to request a double wildcard certificate for a domain (for example *.*.local.com
).
Observability¶
The Observability section defines a per router behavior regarding access-logs, metrics or tracing.
The default router observability configuration is inherited from the attached EntryPoints and can be configured with the observability options. However, a router defining its own observability configuration will opt-out from these defaults.
Note that to enable router-level observability, you must first enable access-logs, tracing, and/or metrics.
AddInternals option
By default, and for any type of signals (access-logs, metrics and tracing),
Traefik disables observability for internal resources.
The observability options described below cannot interfere with the AddInternals
ones,
and will be ignored.
For instance, if a router exposes the api@internal
service and metrics.AddInternals
is false,
it will never produces metrics, even if the router observability configuration enables metrics.
accessLogs
¶
Optional
The accessLogs
option controls whether the router will produce access-logs.
Disable access-logs for a router using the File Provider
## Dynamic configuration
http:
routers:
my-router:
rule: "Path(`/foo`)"
service: service-foo
observability:
accessLogs: false
## Dynamic configuration
[http.routers]
[http.routers.my-router]
rule = "Path(`/foo`)"
service = "service-foo"
[http.routers.my-router.observability]
accessLogs = false
metrics
¶
Optional
The metrics
option controls whether the router will produce metrics.
Metrics layers
When metrics layers are not enabled with the addEntryPointsLabels
, addRoutersLabels
and/or addServicesLabels
options,
enabling metrics for a router will not enable them.
Disable metrics for a router using the File Provider
## Dynamic configuration
http:
routers:
my-router:
rule: "Path(`/foo`)"
service: service-foo
observability:
metrics: false
## Dynamic configuration
[http.routers]
[http.routers.my-router]
rule = "Path(`/foo`)"
service = "service-foo"
[http.routers.my-router.observability]
metrics = false
tracing
¶
Optional
The tracing
option controls whether the router will produce traces.
Disable tracing for a router using the File Provider
## Dynamic configuration
http:
routers:
my-router:
rule: "Path(`/foo`)"
service: service-foo
observability:
tracing: false
## Dynamic configuration
[http.routers]
[http.routers.my-router]
rule = "Path(`/foo`)"
service = "service-foo"
[http.routers.my-router.observability]
tracing = false
Configuring TCP Routers¶
The character @
is not authorized in the router name
General¶
If both HTTP routers and TCP routers listen to the same EntryPoint, the TCP routers will apply before the HTTP routers. If no matching route is found for the TCP routers, then the HTTP routers will take over.
EntryPoints¶
If not specified, TCP routers will accept requests from all EntryPoints in the list of default EntryPoints. If you want to limit the router scope to a set of entry points, set the entry points option.
How to handle Server First protocols?
To correctly handle a request, Traefik needs to wait for the first few bytes to arrive before it can decide what to do with it.
For protocols where the server is expected to send first, such as SMTP, if no specific setup is in place, we could end up in a situation where both sides are waiting for data and the connection appears to have hanged.
The only way that Traefik can deal with such a case, is to make sure that on the concerned entry point, there is no TLS router whatsoever (neither TCP nor HTTP), and there is at least one non-TLS TCP router that leads to the server in question.
Listens to Every Entry Point
Dynamic Configuration
## Dynamic configuration
tcp:
routers:
Router-1:
# By default, routers listen to every EntryPoints.
rule: "HostSNI(`example.com`)"
service: "service-1"
# will route TLS requests (and ignore non tls requests)
tls: {}
## Dynamic configuration
[tcp.routers]
[tcp.routers.Router-1]
# By default, routers listen to every EntryPoints.
rule = "HostSNI(`example.com`)"
service = "service-1"
# will route TLS requests (and ignore non tls requests)
[tcp.routers.Router-1.tls]
Static Configuration
## Static configuration
entryPoints:
web:
address: ":80"
websecure:
address: ":443"
other:
address: ":9090"
## Static configuration
[entryPoints]
[entryPoints.web]
address = ":80"
[entryPoints.websecure]
address = ":443"
[entryPoints.other]
address = ":9090"
## Static configuration
--entryPoints.web.address=:80
--entryPoints.websecure.address=:443
--entryPoints.other.address=:9090
Listens to Specific EntryPoints
Dynamic Configuration
## Dynamic configuration
tcp:
routers:
Router-1:
# won't listen to entry point web
entryPoints:
- "websecure"
- "other"
rule: "HostSNI(`example.com`)"
service: "service-1"
# will route TLS requests (and ignore non tls requests)
tls: {}
## Dynamic configuration
[tcp.routers]
[tcp.routers.Router-1]
# won't listen to entry point web
entryPoints = ["websecure", "other"]
rule = "HostSNI(`example.com`)"
service = "service-1"
# will route TLS requests (and ignore non tls requests)
[tcp.routers.Router-1.tls]
Static Configuration
## Static configuration
entryPoints:
web:
address: ":80"
websecure:
address: ":443"
other:
address: ":9090"
## Static configuration
[entryPoints]
[entryPoints.web]
address = ":80"
[entryPoints.websecure]
address = ":443"
[entryPoints.other]
address = ":9090"
## Static configuration
--entryPoints.web.address=:80
--entryPoints.websecure.address=:443
--entryPoints.other.address=:9090
Rule¶
Rules are a set of matchers configured with values, that determine if a particular connection matches specific criteria. If the rule is verified, the router becomes active, calls middlewares, and then forwards the request to the service.
The table below lists all the available matchers:
Rule | Description |
---|---|
HostSNI(`domain`) |
Checks if the connection's Server Name Indication is equal to domain . |
HostSNIRegexp(`regexp`) |
Checks if the connection's Server Name Indication matches regexp . |
ClientIP(`ip`) |
Checks if the connection's client IP correspond to ip . It accepts IPv4, IPv6 and CIDR formats. |
ALPN(`protocol`) |
Checks if the connection's ALPN protocol equals protocol . |
Backticks or Quotes?
To set the value of a rule, use backticks `
or escaped double-quotes \"
.
Single quotes '
are not accepted since the values are Go's String Literals.
Regexp Syntax
Matchers that accept a regexp as their value use a Go flavored syntax.
Expressing Complex Rules Using Operators and Parenthesis
The usual AND (&&
) and OR (||
) logical operators can be used, with the expected precedence rules,
as well as parentheses.
One can invert a matcher by using the NOT (!
) operator.
The following rule matches connections where:
- either Server Name Indication is
example.com
OR, - Server Name Indication is
example.org
AND ALPN protocol is NOTh2
HostSNI(`example.com`) || (HostSNI(`example.org`) && !ALPN(`h2`))
HostSNI and HostSNIRegexp¶
HostSNI
and HostSNIRegexp
matchers allow to match connections targeted to a given domain.
These matchers do not support non-ASCII characters, use punycode encoded values (rfc 3492) to match such domains.
HostSNI & TLS
It is important to note that the Server Name Indication is an extension of the TLS protocol.
Hence, only TLS routers will be able to specify a domain name with that rule.
However, there is one special use case for HostSNI with non-TLS routers:
when one wants a non-TLS router that matches all (non-TLS) requests,
one should use the specific HostSNI(`*`)
syntax.
Examples
Match all connections:
HostSNI(`*`)
HostSNIRegexp(`^.*$`)
Match TCP connections sent to example.com
:
HostSNI(`example.com`)
Match TCP connections openned on any subdomain of example.com
:
HostSNIRegexp(`^.+\.example\.com$`)
ClientIP¶
The ClientIP
matcher allows matching connections opened by a client with the given IP.
Examples
Match connections opened by a given IP:
ClientIP(`10.76.105.11`)
ClientIP(`::1`)
Match connections coming from a given subnet:
ClientIP(`192.168.1.0/24`)
ClientIP(`fe80::/10`)
ALPN¶
The ALPN
matcher allows matching connections the given protocol.
It would be a security issue to let a user-defined router catch the response to
an ACME TLS challenge previously initiated by Traefik.
For this reason, the ALPN
matcher is not allowed to match the ACME-TLS/1
protocol, and Traefik returns an error if this is attempted.
Example
Match connections using the ALPN protocol h2
:
ALPN(`h2`)
Priority¶
To avoid path overlap, routes are sorted, by default, in descending order using rules length. The priority is directly equal to the length of the rule, and so the longest length has the highest priority.
A value of 0
for the priority is ignored: priority = 0
means that the default rules length sorting is used.
Maximum Value
Traefik reserves a range of priorities for its internal routers, the maximum user-defined router priority value is:
(MaxInt32 - 1000)
for 32-bit platforms,(MaxInt64 - 1000)
for 64-bit platforms.
How default priorities are computed
## Dynamic configuration
tcp:
routers:
Router-1:
rule: "ClientIP(`192.168.0.12`)"
# ...
Router-2:
rule: "ClientIP(`192.168.0.0/24`)"
# ...
## Dynamic configuration
[tcp.routers]
[tcp.routers.Router-1]
rule = "ClientIP(`192.168.0.12`)"
# ...
[tcp.routers.Router-2]
rule = "ClientIP(`192.168.0.0/24`)"
# ...
The table below shows that Router-2
has a higher computed priority than Router-1
.
Name | Rule | Priority |
---|---|---|
Router-1 | ClientIP(`192.168.0.12`) |
24 |
Router-2 | ClientIP(`192.168.0.0/24`) |
26 |
Which means that requests from 192.168.0.12
would go to Router-2 even though Router-1 is intended to specifically handle them.
To achieve this intention, a priority (greater than 26) should be set on Router-1.
Setting priorities -- using the File Provider
## Dynamic configuration
tcp:
routers:
Router-1:
rule: "ClientIP(`192.168.0.12`)"
entryPoints:
- "web"
service: service-1
priority: 2
Router-2:
rule: "ClientIP(`192.168.0.0/24`)"
entryPoints:
- "web"
priority: 1
service: service-2
## Dynamic configuration
[tcp.routers]
[tcp.routers.Router-1]
rule = "ClientIP(`192.168.0.12`)"
entryPoints = ["web"]
service = "service-1"
priority = 2
[tcp.routers.Router-2]
rule = "ClientIP(`192.168.0.0/24`)"
entryPoints = ["web"]
priority = 1
service = "service-2"
In this configuration, the priority is configured so that Router-1
will handle requests from 192.168.0.12
.
RuleSyntax¶
Optional, Default=""
In Traefik v3 a new rule syntax has been introduced (migration guide).
ruleSyntax
option allows to configure the rule syntax to be used for parsing the rule on a per-router basis.
This allows to have heterogeneous router configurations and ease migration.
The default value of the ruleSyntax
option is inherited from the defaultRuleSyntax
option in the static configuration.
By default, the defaultRuleSyntax
static option is v3
, meaning that the default rule syntax is also v3
.
Set rule syntax -- using the File Provider
## Dynamic configuration
tcp:
routers:
Router-v3:
rule: ClientIP(`192.168.0.11`) || ClientIP(`192.168.0.12`)
ruleSyntax: v3
Router-v2:
rule: ClientIP(`192.168.0.11`, `192.168.0.12`)
ruleSyntax: v2
## Dynamic configuration
[tcp.routers]
[tcp.routers.Router-v3]
rule = "ClientIP(`192.168.0.11`) || ClientIP(`192.168.0.12`)"
ruleSyntax = v3
[tcp.routers.Router-v2]
rule = "ClientIP(`192.168.0.11`, `192.168.0.12`)"
ruleSyntax = v2
apiVersion: traefik.io/v1alpha1
kind: IngressRouteTCP
metadata:
name: test.route
namespace: default
spec:
routes:
# route v3
- match: ClientIP(`192.168.0.11`) || ClientIP(`192.168.0.12`)
syntax: v3
kind: Rule
# route v2
- match: ClientIP(`192.168.0.11`, `192.168.0.12`)
syntax: v2
kind: Rule
In this configuration, the ruleSyntax is configured to allow Router-v2
to use v2 syntax,
while for Router-v3
it is configured to use v3 syntax.
Middlewares¶
You can attach a list of middlewares to each TCP router. The middlewares will take effect only if the rule matches, and before connecting to the service.
The character @
is not allowed to be used in the middleware name.
Middlewares order
Middlewares are applied in the same order as their declaration in router.
With a middleware -- using the File Provider
## Dynamic configuration
[tcp.routers]
[tcp.routers.my-router]
rule = "HostSNI(`*`)"
# declared elsewhere
middlewares = ["ipallowlist"]
service = "service-foo"
## Dynamic configuration
tcp:
routers:
my-router:
rule: "HostSNI(`*`)"
# declared elsewhere
middlewares:
- ipallowlist
service: service-foo
Services¶
You must attach a TCP service per TCP router. Services are the target for the router.
TCP routers can only target TCP services (not HTTP services).
TLS¶
General¶
When a TLS section is specified, it instructs Traefik that the current router is dedicated to TLS requests only (and that the router should ignore non-TLS requests).
By default, a router with a TLS section will terminate the TLS connections, meaning that it will send decrypted data to the services.
Router for TLS requests
## Dynamic configuration
tcp:
routers:
Router-1:
rule: "HostSNI(`foo-domain`)"
service: service-id
# will terminate the TLS request by default
tls: {}
## Dynamic configuration
[tcp.routers]
[tcp.routers.Router-1]
rule = "HostSNI(`foo-domain`)"
service = "service-id"
# will terminate the TLS request by default
[tcp.routers.Router-1.tls]
Postgres STARTTLS
Traefik supports the Postgres STARTTLS protocol, which allows TLS routing for Postgres connections.
To do so, Traefik reads the first bytes sent by a Postgres client, identifies if they correspond to the message of a STARTTLS negotiation, and, if so, acknowledges and signals the client that it can start the TLS handshake.
Please note/remember that there are subtleties inherent to STARTTLS in whether the connection ends up being a TLS one or not.
These subtleties depend on the sslmode
value in the client configuration (and on the server authentication rules).
Therefore, it is recommended to use the require
value for the sslmode
.
Afterwards, the TLS handshake, and routing based on TLS, can proceed as expected.
Postgres STARTTLS with TCP TLS PassThrough routers
As mentioned above, the sslmode
configuration parameter does have an impact on whether a STARTTLS session will succeed.
In particular in the context of TCP TLS PassThrough, some of the values (such as allow
) do not even make sense.
Which is why, once more it is recommended to use the require
value.
passthrough
¶
As seen above, a TLS router will terminate the TLS connection by default.
However, the passthrough
option can be specified to set whether the requests should be forwarded "as is", keeping all data encrypted.
It defaults to false
.
Configuring passthrough
## Dynamic configuration
tcp:
routers:
Router-1:
rule: "HostSNI(`foo-domain`)"
service: service-id
tls:
passthrough: true
## Dynamic configuration
[tcp.routers]
[tcp.routers.Router-1]
rule = "HostSNI(`foo-domain`)"
service = "service-id"
[tcp.routers.Router-1.tls]
passthrough = true
options
¶
The options
field enables fine-grained control of the TLS parameters.
It refers to a TLS Options and will be applied only if a HostSNI
rule is defined.
Configuring the tls options
## Dynamic configuration
tcp:
routers:
Router-1:
rule: "HostSNI(`foo-domain`)"
service: service-id
# will terminate the TLS request
tls:
options: foo
tls:
options:
foo:
minVersion: VersionTLS12
cipherSuites:
- TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384
- TLS_ECDHE_ECDSA_WITH_CHACHA20_POLY1305_SHA256
- TLS_ECDHE_RSA_WITH_CHACHA20_POLY1305_SHA256
- TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256
- TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256
## Dynamic configuration
[tcp.routers]
[tcp.routers.Router-1]
rule = "HostSNI(`foo-domain`)"
service = "service-id"
# will terminate the TLS request
[tcp.routers.Router-1.tls]
options = "foo"
[tls.options]
[tls.options.foo]
minVersion = "VersionTLS12"
cipherSuites = [
"TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384",
"TLS_ECDHE_ECDSA_WITH_CHACHA20_POLY1305_SHA256",
"TLS_ECDHE_RSA_WITH_CHACHA20_POLY1305_SHA256",
"TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256",
"TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256",
]
certResolver
¶
See certResolver
for HTTP router for more information.
## Dynamic configuration
tcp:
routers:
routerfoo:
rule: "HostSNI(`snitest.com`)"
tls:
certResolver: foo
## Dynamic configuration
[tcp.routers]
[tcp.routers.routerfoo]
rule = "HostSNI(`snitest.com`)"
[tcp.routers.routerfoo.tls]
certResolver = "foo"
domains
¶
See domains
for HTTP router for more information.
## Dynamic configuration
tcp:
routers:
routerbar:
rule: "HostSNI(`snitest.com`)"
tls:
certResolver: "bar"
domains:
- main: "snitest.com"
sans:
- "*.snitest.com"
## Dynamic configuration
[tcp.routers]
[tcp.routers.routerbar]
rule = "HostSNI(`snitest.com`)"
[tcp.routers.routerbar.tls]
certResolver = "bar"
[[tcp.routers.routerbar.tls.domains]]
main = "snitest.com"
sans = ["*.snitest.com"]
Configuring UDP Routers¶
The character @
is not allowed in the router name
General¶
Similarly to TCP, as UDP is the transport layer, there is no concept of a request, so there is no notion of an URL path prefix to match an incoming UDP packet with. Furthermore, as there is no good TLS support at the moment for multiple hosts, there is no Host SNI notion to match against either. Therefore, there is no criterion that could be used as a rule to match incoming packets in order to route them. So UDP "routers" at this time are pretty much only load-balancers in one form or another.
Sessions and timeout
Even though UDP is connectionless (and because of that),
the implementation of an UDP router in Traefik relies on what we (and a couple of other implementations) call a session
.
It basically means that some state is kept about an ongoing communication between a client and a backend,
notably so that the proxy knows where to forward a response packet from a backend.
As expected, a timeout
is associated to each of these sessions,
so that they get cleaned out if they go through a period of inactivity longer than a given duration.
Timeout can be configured using the entryPoints.name.udp.timeout
option as described under EntryPoints.
EntryPoints¶
If not specified, UDP routers will accept packets from all defined (UDP) EntryPoints.
If one wants to limit the router scope to a set of EntryPoints, one should set the entryPoints
option.
Listens to Every Entry Point
Dynamic Configuration
## Dynamic configuration
udp:
routers:
Router-1:
# By default, routers listen to all UDP entrypoints
# i.e. "other", and "streaming".
service: "service-1"
## Dynamic configuration
[udp.routers]
[udp.routers.Router-1]
# By default, routers listen to all UDP entrypoints,
# i.e. "other", and "streaming".
service = "service-1"
Static Configuration
## Static configuration
entryPoints:
# not used by UDP routers
web:
address: ":80"
# used by UDP routers
other:
address: ":9090/udp"
streaming:
address: ":9191/udp"
## Static configuration
[entryPoints]
# not used by UDP routers
[entryPoints.web]
address = ":80"
# used by UDP routers
[entryPoints.other]
address = ":9090/udp"
[entryPoints.streaming]
address = ":9191/udp"
## Static configuration
--entryPoints.web.address=":80"
--entryPoints.other.address=":9090/udp"
--entryPoints.streaming.address=":9191/udp"
Listens to Specific EntryPoints
Dynamic Configuration
## Dynamic configuration
udp:
routers:
Router-1:
# does not listen on "other" entry point
entryPoints:
- "streaming"
service: "service-1"
## Dynamic configuration
[udp.routers]
[udp.routers.Router-1]
# does not listen on "other" entry point
entryPoints = ["streaming"]
service = "service-1"
Static Configuration
## Static configuration
entryPoints:
web:
address: ":80"
other:
address: ":9090/udp"
streaming:
address: ":9191/udp"
## Static configuration
[entryPoints]
[entryPoints.web]
address = ":80"
[entryPoints.other]
address = ":9090/udp"
[entryPoints.streaming]
address = ":9191/udp"
## Static configuration
--entryPoints.web.address=":80"
--entryPoints.other.address=":9090/udp"
--entryPoints.streaming.address=":9191/udp"
Services¶
There must be one (and only one) UDP service referenced per UDP router. Services are the target for the router.
UDP routers can only target UDP services (and not HTTP or TCP services).
Using Traefik OSS in Production?
If you are using Traefik at work, consider adding enterprise-grade API gateway capabilities or commercial support for Traefik OSS.
Adding API Gateway capabilities to Traefik OSS is fast and seamless. There's no rip and replace and all configurations remain intact. See it in action via this short video.