Clixon supports two RESTCONF compile-time variants: FCGI and Native.



The restconf deamon provides a http/https RESTCONF interface to the Clixon backend. It comes in two variants, as shown by (1) and (2) in the figure above:

  1. Native http using libevhtp, which combines a web server and restconf handler.
  2. A reverse proxy (such as NGINX) and fastCGI where web and restconf function is separated

The restconf daemon communicates with the backend using internal netconf over the CLIXON_SOCK. If FCGI is used, there is also a FCGI socket specified by CLICON_RESTCONF_PATH.

The restconf daemon reads its initial config options from the configuration file on startup. The native http variant can read config options from the backend as an alternative to reading everything from clixon options.

You can add plugins to the restconf daemon, where the primary usecase is authentication, using the ca_auth callback.


The RESTCONF daemon can be configured for compile-time (by autotools) as follows:
 RESTCONF using fcgi/ reverse proxy. This is default.
 RESTCONF using native http with libevhtp
 Set www user different from www-data

After that perform system-wide compilation:

make && sudo make install

Command-line options

The restconf daemon have the following command-line options:
-h Help
-D <level> Debug level
-f <file> Clixon config file
-E <dir> Extra configuration directory
-l <option> Log on (s)yslog, std(e)rr, std(o)ut or (f)ile. Syslog is default. If foreground, then syslog and stderr is default. Filename is given after -f: -lf<file>.
-p <dir> Yang directory path (see CLICON_YANG_DIR)
-b <dir> Specify XMLDB database directory
-a <family> Internal backend socket family: UNIX|IPv4|IPv6
-u <path|addr> Internal socket domain path or IP addr (see -a)(default: /usr/var/hello.sock)
-r Do not drop privileges if run as root
-o <option=value>
 Give configuration option overriding config file (see clixon-config.yang)

Note that the restconf daemon can be started as root, drops privileges to wwwuser, unless the -r command-line option is used.


The following RESTCONF configuration options can be defined in the clixon configuration file:

Location of restconf .so plugins. Load all .so plugins in this dir as restconf code plugins.
Start restconf daemon internally from backend daemon.
If RESTCONF authentication auth-type=none then use this user

The remaining options are defined in clixon-restconf.yang where they can be defined locally within the clixon config file:

<clixon-config xmlns="">
      ...  <!-- Add restconf options here -->

Alternatively, the restconf configuration can be defined in the regular datastore by adding, as follows:

<restconf xmlns="">
   ...   <!--  Add restconf options here -->

In the latter case, the restconf daemon reads its config from the running datastore on startup.

Those restconf options are as follows:

Enable the RESTCONF daemon. If disabled, the restconf daemon will not start
Authentication method (see auth types)
Enable to debug on syslog
Restconf vallues are pretty printed by default. Disable to turn this off

Auth types

The RESTCONF daemon uses the following authentication types:

Messages are not authenticated and set to the value of CLICON_ANONYMOUS_USER. A callback can revise this behavior.
Set to authenticated and extract the username from the SSL_CN parameter. A callback can revise this behavior.
User-defined behaviour as implemented by the auth callback. Typically done by basic auth, eg HTTP_AUTHORIZATION header, and verify password


Nginx specific config options are:

Path to FCGI unix socket. This path should be the same as specific in fcgi reverse proxy


Evhtp specific config options are:

Path to server certificate file
Path to server key file
Path to server CA cert file
List of server sockets that the restconf daemon listens to with the following fields:
socket namespace
Network namespace
socket address
IP address to bind to
socket port
TCP port to bind to
socket ssl
If true: HTTPS; if false: HTTP protocol


Configure a single HTTP on port 80 in the default config file:

<clixon-config xmlns="">

Configure two HTTPS listeners in two different namespaces:

<restconf xmlns="">

SSL Certificates

If you use native RESTCONF you may want to have server/client certs. If you use FCGI, certs are configured according to the reverse proxy documentation, such as NGINX. The rest of this section applies to native restconf only.

If you already have certified server certs, ensure CLICON_SSL_SERVER_CERT and CLICON_SSL_SERVER_KEY points to them.

If you do not have them, you can generate self-signed certs, for example as follows:

openssl req -x509 -nodes -newkey rsa:4096 -keyout /etc/ssl/private/clixon-server-key.pem -out /etc/ssl/certs/clixon-server-crt.pem -days 365

You can also generate client certs (not shown here) using CLICON_SSL_CA_CERT. Example using client certs and curl for client andy:

curl -Ssik --key andy.key --cert andy.crt -X GET https://localhost/restconf/data/example:x


You can start the RESTCONF daemon in several ways:

  1. systemd as described in Installation
  2. internally using the process-control RPC (see below)
  3. docker mechanisms, see the docker container docs

Internal start

For starting restconf internally, you need to enable CLICON_BACKEND_RESTCONF_PROCESS option

Thereafter, you can either use the clixon-restconf.yang configuration or use the clixon-lib.yang process control RPC:s to start/stop/restart the daemon or query status.

The algorithm for starting and stopping the clixon-restconf internally is as follows:

  1. on RPC start, if enable is true, start the service, if false, error or ignore it
  2. on RPC stop, stop the service
  3. on backend start make the state as configured
  4. on enable change, make the state as configured

Example 1, using netconf edit-config to start the process:

<rpc xmlns="urn:ietf:params:xml:ns:netconf:base:1.0" message-id="10">
         <restconf xmlns="">
<rpc-reply xmlns="urn:ietf:params:xml:ns:netconf:base:1.0" message-id="10">
<rpc xmlns="urn:ietf:params:xml:ns:netconf:base:1.0" message-id="11">
<rpc-reply xmlns="urn:ietf:params:xml:ns:netconf:base:1.0" message-id="10">

Example 2, using netconf RPC to restart the process:

<rpc xmlns="urn:ietf:params:xml:ns:netconf:base:1.0" message-id="10">
   <process-control xmlns="">
<rpc-reply xmlns="urn:ietf:params:xml:ns:netconf:base:1.0" message-id="10">
   <pid xmlns="">1029</pid>

Note that the backend daemon must run as root (no lowering of privileges) to use this feature.

Plugin callbacks

Restconf plugins implement callbacks, some are same as for backend plugins.

Clixon plugin init function, called immediately after plugin is loaded into the restconf daemon.
Called when application is started and initialization is complete, and after drop privileges.
Called just before plugin is unloaded
Called at parsing of yang modules containing an extension statement.
See auth callback

Auth callback

The role of the authentication callback is, given a message (its headers) and authentication type, determine if the message passes authentication and return an associated user.

The auth callback is invoked after incoming processing, including cert validation, if any, but before relaying the message to the backend for NACM checks and datastore processing.

If the message is not authenticated, an error message is returned with tag: access denied and HTTP error code 401 Unauthorized.

There are default handlers for TLS client certs and for “none” authentication. But other variants, such as http basic authentication, oauth2 or the remapping of client certs to NACM usernames, can be implemented by this callback

If the message is authenticated, a user is associated with the message. This user can be derived from the headers or mapped in an application-dependent way. This user is used internally in Clixon and sent via the IPC protocol to the backend where it may be used for NACM authorization.

The signature of the auth callback is as follows:

int ca_auth(clicon_handle h, void *req, clixon_auth_type_t auth_type, int *authp, char **userp);


Clixon handle
Per-message request www handle to use with restconf_api.h
Specifies how the authentication is made and what default value
Boolean result value: 0 for not authenticated, 1: authenticated
The associated user, malloced by plugin. Only if retval is 1/OK and authp=1,

The return value is one of:

  • -1: Fatal error
  • 0: Ignore, undecided, not handled, same as no callback
  • 1: OK see authp and userp parameter for result.

If there are multiple callbacks, the first result which is not “ignore” is returned. This is to allow for different callbacks registering different classes, or grouping of authentication.

The main example contains example code.


This section describes the RESTCONF FCGI mode using NGINX.

If you use FCGI, you need to configure a reverse-proxy, such as NGINX. A typical configuration is as follows:

server {
  location / {
    fastcgi_pass unix:/www-data/fastcgi_restconf.sock;
    include fastcgi_params;

where fastcgi_pass setting must match CLICON_RESTCONF_PATH.

RESTCONF streams

Clixon has an experimental RESTCONF event stream implementations following RFC8040 Section 6 using Server-Sent Events (SSE). Currently this is implemented in FCGI/Nginx only (not evhtp).


RESTCONF streams are experimental and only implemented for FCGI.

Example: set the Clixon configuration options:


In this example, the stream example is accessed with

Clixon defines an internal in-memory (not persistent) replay function controlled by the configure option above. In this example, the retention is configured to 1 hour, i.e., the stream replay function will only save timeseries one hour, but if the restconf daemon is restarted, the history will be lost.

In the Nginx configuration, add the following to extend the nginx configuration file with the following statements (for example):

location /streams {
    fastcgi_pass unix:/www-data/fastcgi_restconf.sock;
    include fastcgi_params;
    proxy_http_version 1.1;
    proxy_set_header Connection "";

An example of a stream access is as follows:

curl -H "Accept: text/event-stream" -s -X GET http://localhost/streams/EXAMPLE
data: <notification xmlns="urn:ietf:params:xml:ns:netconf:notification:1.0"><eventTime>2018-11-04T14:47:11.373124</eventTime><event><event-class>fault</event-class><reportingEntity><card>Ethernet0</card></reportingEntity><severity>major</severity></event></notification>
data: <notification xmlns="urn:ietf:params:xml:ns:netconf:notification:1.0"><eventTime>2018-11-04T14:47:16.375265</eventTime><event><event-class>fault</event-class><reportingEntity><card>Ethernet0</card></reportingEntity><severity>major</severity></event></notification>

You can also specify start and stop time. Start-time enables replay of existing samples, while stop-time is used both for replay, but also for stopping a stream at some future time:

curl -H "Accept: text/event-stream" -s -X GET http://localhost/streams/EXAMPLE?start-time=2014-10-25T10:02:00&stop-time=2014-10-25T12:31:00

Fcgi stream options

The following options apply only for fcgi mode and notification streams:

Enable monitoring information for the RESTCONF protocol from RFC 804 (only fcgi)
Stream path appended to CLICON_STREAM_URL to form stream subscription URL (only fcgi)