Run your own caching, non-censoring, non-logging, DNSSEC-capable, DNSCrypt-enabled DNS resolver virtually anywhere!
If you are already familiar with Docker, it shouldn't take more than 5 minutes to get your resolver up and running.
Table of contents:
Think about a name. This is going to be part of your DNSCrypt provider name.
If you are planning to make your resolver publicly accessible, this name will
be public.
By convention, it has to look like a domain name (example.com
), but it doesn't
have to be an actual, registered domain.
Let's pick example.com
here.
You probably need to perform the following steps as root
.
Download, create and initialize the container:
docker run --name=dnscrypt-server -p 443:443/udp -p 443:443/tcp \
--restart=unless-stopped \
-v /etc/dnscrypt-server/keys:/opt/encrypted-dns/etc/keys \
jedisct1/dnscrypt-server init -N example.com -E '192.168.1.1:443'
This will only accept connections via DNSCrypt on the standard port (443). Replace
192.168.1.1
with the actual external IP address (not the internal Docker one)
clients will connect to.
IPv6 addresses should be enclosed in brackets; for example: [2001:0db8::412f]:443
.
Multiple comma-separated IPs and ports can be specified, as in -E '192.168.1.1:443,[2001:0db8::412f]:443'
.
If you want to use a different port, replace all occurrences of 443
with the alternative port in the
command above (including -p ...
). But if you have an existing website that should be accessible on
port 443
, the server can transparently relay non-DNS traffic to it (see below).
-v /etc/dnscrypt-server:/opt/encrypted-dns/etc/keys
means that the path /opt/encrypted-dns/etc/keys
, internal to the container, is mapped to /etc/dnscrypt-server/keys
, the directory we just created before. Do not change /opt/encrypted-dns/etc/keys
. But if you created a directory in a different location, replace /etc/dnscrypt-server/keys
accordingly in the command above.
Note: on MacOS, don't use -v ...:...
. Remove that part from the command-line, as current versions of MacOS and Docker don't seem to work well with shared directories.
The init
command will print the DNS stamp of your server.
Done.
You can verify that the server is running with:
docker ps
Note: if you previously created a container with the same name, and Docker complains that the name is already in use, remove it and try again:
docker rm --force dnscrypt-server
In order to install the latest version of the image, or change parameters, use the following steps:
docker pull jedisct1/dnscrypt-server
state
directory):ls -l /etc/dnscrypt-server/keys
If you have some content here, skip to step 3.
Nothing here? Maybe you didn't use the -v
option to map container files to a local directory when creating the container.
In that case, copy the data directly from the container:
docker cp dnscrypt-server:/opt/encrypted-dns/etc/keys ~/keys
docker stop dnscrypt-server
docker ps # Check that it's not running
docker rename dnscrypt-server dnscrypt-server-old
init
command again and start the new container:docker run --name=dnscrypt-server -p 443:443/udp -p 443:443/tcp \
--restart=unless-stopped \
-v /etc/dnscrypt-server/keys:/opt/encrypted-dns/etc/keys \
jedisct1/dnscrypt-server init -N example.com -E '192.168.1.1:443'
# (adjust accordingly)
docker ps # Check that it's running
docker rm dnscrypt-server-old
Parameters differ from the ones used in the previous container.
For example, if you originally didn't activate relaying
but want to enable it, append -A
to the command. Or if you want to enable
metrics, append -M 0.0.0.0:9100
to the end, and -p 9100:9100/tcp
after
-p 443:443/tcp
(see below).
The server can be configured as a relay for the Anonymized DNSCrypt protocol by adding the -A
switch to the init
command.
The relay DNS stamp will be printed right after the regular stamp.
Metrics are accessible inside the container as http://127.0.0.1:9100/metrics.
They can be made accessible outside of the container by adding the -M
option followed by the listening IP and port (for example: -M 0.0.0.0:9100
).
These metrics can be indexed with Prometheus and dashboards can be created with Grafana.
If the DNS server is listening to port 443
, but you still want to have a web (or DoH) service accessible on that port, add the -T
switch followed by the backend server IP and port to the init
command (for example: -T 10.0.0.1:4443
).
The backend server must support the HTTP/2 protocol.
The server can be used block domains. For example, the sfw.scaleway-fr
server uses that feature to provide a service that blocks websites possibly not suitable for children.
In order to do so, create a directory that will contain the blacklists:
mkdir -p /etc/dnscrypt-server/lists
And put the list of domains to block in a file named /etc/dnscrypt-server/lists/blacklist.txt
, one domain per line.
Then, follow the upgrade procedure, adding the following option to the docker run
command: -v /etc/dnscrypt-server/lists:/opt/encrypted-dns/etc/lists
.
If you want to help against DNS centralization and surveillance, announce your server and/or relay on the list of public DNS DoH and DNSCrypt servers.
The best way to do so is to send a pull request to the dnscrypt-resolvers repository.
Kubernetes configurations are located in the kube
directory. Currently these assume
a persistent disk named dnscrypt-keys
on GCE. You will need to adjust the volumes
definition on other platforms. Once that is setup, you can have a dnscrypt server up
in minutes.
kube/dnscrypt-init-job.yml
. Change example.com
to your desired hostname
and 127.0.0.1
to your static IP.kube/dnscrypt-srv.yml
and change loadBalancerIP
to your static IP in both locations.kubectl create -f kube/dnscrypt-init-job.yml
to setup your keys.kubectl create -f kube/dnscrypt-deployment.yml
to deploy the dnscrypt server.kubectl create -f kube/dnscrypt-srv.yml
to expose your server to the world.To get your public key just view the logs for the dnscrypt-init
job. The public
IP for your server is merely the dnscrypt
service address.
You can setup a server very quickly with Docker Compose.
docker-compose up
docker-compose.yml
to use the start command.docker-compose up
Now you will have a local server running on port 5443.
To add new configuration to Unbound, add files to the /opt/unbound/etc/unbound/zones
directory. All files ending in .conf
will be processed. In this manner, you
can add any directives to the server:
section of the Unbound configuration.
While Unbound is not a full authoritative name server, it supports resolving
custom entries in a way that is serviceable on a small, private LAN. You can use
unbound to resolve private hostnames such as my-computer.example.com
within
your LAN.
To support such custom entries using this image, first map a volume to the zones
directory. Add this to your docker run
line:
-v /etc/dnscrypt-server/zones:/opt/unbound/etc/unbound/zones
The whole command to create and initialize a container would look something like this:
docker run --name=dnscrypt-server \
-v /etc/dnscrypt-server/zones:/opt/unbound/etc/unbound/zones \
-p 443:443/udp -p 443:443/tcp --net=host \
jedisct1/dnscrypt-server init -N example.com -E '192.168.1.1:443'
Create a new .conf
file:
touch /etc/dnscrypt-server/zones/example.conf
Now, add one or more unbound directives to the file, such as:
local-zone: "example.com." static
local-data: "my-computer.example.com. IN A 10.0.0.1"
local-data: "other-computer.example.com. IN A 10.0.0.2"
If Unbound doesn't like one of the newly added directives, it will probably not respond over the network. In that case, here are some commands to work out what is wrong:
docker logs dnscrypt-server
docker exec dnscrypt-server /opt/unbound/sbin/unbound-checkconf
In order to delete everything (containers and images), type:
docker rm --force dnscrypt-server ||:
docker rmi --force jedisct1/dnscrypt-server ||:
Keys and certificates are automatically rotated every 8 hour.