USERNAME

username daemon

Specify the user sniproxy will run as. When sniproxy is launched as super user, it will drop permissions to this user.

PIDFILE

pidfile /var/run/sniproxy.pid

Specify the path to the pid file, the directory much be writeable by the user sniproxy runs as.

ERROR_LOG

error_log {
    syslog daemon
    priority notice
}

Specify how error messages should be handled. Messages can be either logged to syslog using the syslog directive specifies that logs should to a given syslog facility. Alternatively the filename directive may be specified to log to file, these two options are mutually exclusive. The priority directive indicates what severity of messages should be logged. Accepted priorities the standard syslog priorities, in increasing verbosity: emergency, alert, critical, error, warning, notice, info, and debug.

ACCESS_LOG

access_log {
    filename /var/log/sniproxy/access.log
}

Specify how connections should be logged, may be overridden in a specific listener. Connections are logged after both the client and server have sockets have been closed. The syslog and priority directive may be used here as in error_log.

RESOLVER

resolver {
    nameserver 127.0.0.1
    mode ipv6_first
}

Specify how DNS queries should be resolved, this is only required if using hostnames as addresses in the configuration or using wildcard backends. If not specified the IPv4 only queries will be preformed using the system default name servers.

Four modes are supported:

ipv4_only: query for any A records, use the first A record returned (following CNAME records).

ipv6_only: query for any AAAA records, use the first AAAA record returned (following CNAME records).

ipv4_first: query for both A and AAAA records, wait for both queries to complete, use the first A record if any, otherwise use the first AAAA record.

ipv6_first: query for both A and AAAA records, wait for both queries to complete, use the first AAAA record if any, otherwise use the first A record.

It is strongly recommended to use a local name server, since a single socket is reused for all DNS queries and thus the UDP port number is predictable leaving the query only protected from spoofed replies by the 16 bit query ID. Additionally since no internal DNS caching is performed a local resolver can improve performance.

LISTENER

listener 192.0.2.10:80 {
    protocol http
    reuseport yes
    table http_hosts
    fallback 192.0.2.100:80
    bad_requests log
    source 192.0.2.10
    access_log {
        filename /var/log/sniproxy/http_access.log
    }
}
listener [::]:80 {
    protocol http
    ipv6_v6only yes
    table http_hosts
    fallback unix:/var/run/http_fallback_unix.sock
}

Define a listening address to accept new connections from clients on. Addresses may be specified as an IPv4 literal followed by a TCP port number, and IPv6 literal followed by a TCP port number, a bare TCP port number or a unix socket path prefixed with 'unix:'.

Protocol defines how the client request should be parsed to obtain the requested hostname, two protocols are supported http and tls.

Reuseport directive controls if the port is opened in SO_REUSEPORT mode, which allows to run several sniproxy instances on the same ip:port pair. This enables us to evenly load-balance incoming connections between these instances without the use of any external load-balancing proxy. Requires Linux kernel 3.9+. Setting reuseport to "yes" enables this functionality.

The ipv6_v6only directive controls if listening on the IPv6 any address '::' will accepting connections to any IPv4 address as well as an IPv6 address. This is useful if the user wants different configurations for IPv4 and IPv6 or wishes to handle IPv4 traffic with another server/proxy entirely. This is only applicable to IPv6 listeners, and is ignored on other listeners.

Table specifies the name of the table used to lookup which server to forward the connection to based on the hostname extracted from the initial client request. If no table directive is specified the default, unnamed, table will be used.

The fallback directive specifies a server to be used if the client request can not be parsed, a server can not be found in the table for the hostname specified or the hostname can not be resolved.. This should be an IP address and port or unix socket path.

The bad_requests directive allows logging the contents of the client request if it is not parsable, this is useful for debugging.

The source directive allows specifying a specified address to bind to before connecting to the backend server. In most cases it is better to omit this option and allow the operating system to select the outgoing address automatically. Do not include a port number in this address, doing so will limit the proxy to one simultaneous to each server at time.

The access log configuration may be overridden on each listener.

TABLE

table http_hosts {
    ^example\.com$ 192.0.2.101
    ^example\.net$ 192.0.2.102
    ^example\.org$ 192.0.2.103 proxy_protocol
}

Tables define how to map each hostname to a backend server. Each request's hostname is matched against entries in the table in order, until a match is found and that server is used. The server address may be either IP, an IP and port, a unix socket path, a hostname or '*'. If no port is specified, the port of the listener which connection was received on will be used.

The optional proxy_protocol option will prepend a HAProxy PROXY v1 protocol header to the proxied connection allowing supporting webservers to obtain the source and destination IP and port of the original incoming TCP connection.