Replace cloudflared with dnscrypt-proxy

[pedropombeiro]

The cloudflared command is written in Go and is not very lightweight. In my experience, it's not made for long-term running. Instead, the project DoTe has a tiny memory footprint and operates on an event loop with some major optimisations for connection caching. It allows you to forward traffic to any DNS-over-TLS provider.

Have you considered using https://github.com/DNSCrypt/dnscrypt-proxy instead of cloudflared? When I was running Pi-hole on a Raspberry Pi, I also quickly ran into issues with cloudflared. I ended up replacing it with dnscrypt-proxy and it has ran solid for the last couple of years. As an added benefit, dnscrypt-proxy can automatically select the best DNS upstream based on criteria like DoH or DoT support.

This is my full working dnscrypt-proxy.toml configuration:

##############################################
#                                            #
#        dnscrypt-proxy configuration        #
#                                            #
##############################################

## This is an example configuration file.
## You should adjust it to your needs, and save it as "dnscrypt-proxy.toml"
##
## Online documentation is available here: https://dnscrypt.info/doc



##################################
#         Global settings        #
##################################

## List of servers to use
##
## Servers from the "public-resolvers" source (see down below) can
## be viewed here: https://dnscrypt.info/public-servers
##
## If this line is commented, all registered servers matching the require_* filters
## will be used.
##
## The proxy will automatically pick the fastest, working servers from the list.
## Remove the leading # first to enable this; lines starting with # are ignored.

server_names = ['cloudflare']


## List of local addresses and ports to listen to. Can be IPv4 and/or IPv6.
## Example with both IPv4 and IPv6:
## listen_addresses = ['127.0.0.1:53', '[::1]:53']

listen_addresses = ['127.0.0.1:5300']


## Maximum number of simultaneous client connections to accept

max_clients = 32

## Switch to a different system user after listening sockets have been created.
## Note (1): this feature is currently unsupported on Windows.
## Note (2): this feature is not compatible with systemd socket activation.
## Note (3): when using -pidfile, the PID file directory must be writable by the new user

# user_name = 'nobody'


## Require servers (from static + remote sources) to satisfy specific properties

# Use servers reachable over IPv4
ipv4_servers = true

# Use servers reachable over IPv6 -- Do not enable if you don't have IPv6 connectivity
ipv6_servers = false

# Use servers implementing the DNSCrypt protocol
dnscrypt_servers = true

# Use servers implementing the DNS-over-HTTPS protocol
doh_servers = true


## Require servers defined by remote sources to satisfy specific properties

# Server must support DNS security extensions (DNSSEC)
require_dnssec = true

# Server must not log user queries (declarative)
require_nolog = true

# Server must not enforce its own blacklist (for parental control, ads blocking...)
require_nofilter = true

# Server names to avoid even if they match all criteria
disabled_server_names = []


## Always use TCP to connect to upstream servers.
## This can be useful if you need to route everything through Tor.
## Otherwise, leave this to `false`, as it doesn't improve security
## (dnscrypt-proxy will always encrypt everything even using UDP), and can
## only increase latency.

force_tcp = false


## SOCKS proxy
## Uncomment the following line to route all TCP connections to a local Tor node
## Tor doesn't support UDP, so set `force_tcp` to `true` as well.

# proxy = 'socks5://127.0.0.1:9050'


## HTTP/HTTPS proxy
## Only for DoH servers

# http_proxy = 'http://127.0.0.1:8888'


## How long a DNS query will wait for a response, in milliseconds.
## If you have a network with *a lot* of latency, you may need to
## increase this. Startup may be slower if you do so.
## Don't increase it too much. 10000 is the highest reasonable value.

timeout = 5000


## Keepalive for HTTP (HTTPS, HTTP/2) queries, in seconds

keepalive = 30


## Response for blocked queries.  Options are `refused`, `hinfo` (default) or
## an IP response.  To give an IP response, use the format `a:<IPv4>,aaaa:<IPv6>`.
## Using the `hinfo` option means that some responses will be lies.
## Unfortunately, the `hinfo` option appears to be required for Android 8+

# blocked_query_response = 'refused'


## Load-balancing strategy: 'p2' (default), 'ph', 'first' or 'random'

# lb_strategy = 'p2'

## Set to `true` to constantly try to estimate the latency of all the resolvers
## and adjust the load-balancing parameters accordingly, or to `false` to disable.

# lb_estimator = true


## Log level (0-6, default: 2 - 0 is very verbose, 6 only contains fatal errors)

# log_level = 2


## log file for the application

# log_file = 'dnscrypt-proxy.log'


## Use the system logger (syslog on Unix, Event Log on Windows)

# use_syslog = true


## Delay, in minutes, after which certificates are reloaded

cert_refresh_delay = 240


## DNSCrypt: Create a new, unique key for every single DNS query
## This may improve privacy but can also have a significant impact on CPU usage
## Only enable if you don't have a lot of network load

# dnscrypt_ephemeral_keys = false


## DoH: Disable TLS session tickets - increases privacy but also latency

# tls_disable_session_tickets = false


## DoH: Use a specific cipher suite instead of the server preference
## 49199 = TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256
## 49195 = TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256
## 52392 = TLS_ECDHE_RSA_WITH_CHACHA20_POLY1305
## 52393 = TLS_ECDHE_ECDSA_WITH_CHACHA20_POLY1305
##  4865 = TLS_AES_128_GCM_SHA256
##  4867 = TLS_CHACHA20_POLY1305_SHA256
##
## On non-Intel CPUs such as MIPS routers and ARM systems (Android, Raspberry Pi...),
## the following suite improves performance.
## This may also help on Intel CPUs running 32-bit operating systems.
##
## Keep tls_cipher_suite empty if you have issues fetching sources or
## connecting to some DoH servers. Google and Cloudflare are fine with it.

# tls_cipher_suite = [52392, 49199]


## Fallback resolvers
## These are normal, non-encrypted DNS resolvers, that will be only used
## for one-shot queries when retrieving the initial resolvers list, and
## only if the system DNS configuration doesn't work.
## No user application queries will ever be leaked through these resolvers,
## and they will not be used after IP addresses of resolvers URLs have been found.
## They will never be used if lists have already been cached, and if stamps
## don't include host names without IP addresses.
## They will not be used if the configured system DNS works.
## Resolver supporting DNSSEC are recommended.
##
## People in China may need to use 114.114.114.114:53 here.
## Other popular options include 8.8.8.8 and 1.1.1.1.
##
## If more than one resolver are specified, they will be tried in sequence.

fallback_resolvers = ['9.9.9.9:53', '8.8.8.8:53']


## Always use the fallback resolver before the system DNS settings.

ignore_system_dns = true


## Maximum time (in seconds) to wait for network connectivity before
## initializing the proxy.
## Useful if the proxy is automatically started at boot, and network
## connectivity is not guaranteed to be immediately available.
## Use 0 to not test for connectivity at all (not recommended),
## and -1 to wait as much as possible.

netprobe_timeout = 60

## Address and port to try initializing a connection to, just to check
## if the network is up. It can be any address and any port, even if
## there is nothing answering these on the other side. Just don't use
## a local address, as the goal is to check for Internet connectivity.
## On Windows, a datagram with a single, nul byte will be sent, only
## when the system starts.
## On other operating systems, the connection will be initialized
## but nothing will be sent at all.

netprobe_address = '9.9.9.9:53'


## Offline mode - Do not use any remote encrypted servers.
## The proxy will remain fully functional to respond to queries that
## plugins can handle directly (forwarding, cloaking, ...)

# offline_mode = false


## Additional data to attach to outgoing queries.
## These strings will be added as TXT records to queries.
## Do not use, except on servers explicitly asking for extra data
## to be present.

# query_meta = ["key1:value1", "key2:value2", "key3:value3"]


## Automatic log files rotation

# Maximum log files size in MB - Set to 0 for unlimited.
log_files_max_size = 10

# How long to keep backup files, in days
log_files_max_age = 7

# Maximum log files backups to keep (or 0 to keep all backups)
log_files_max_backups = 1



#########################
#        Filters        #
#########################

## Note: if you are using dnsmasq, disable the `dnssec` option in dnsmasq if you
## configure dnscrypt-proxy to do any kind of filtering (including the filters
## below and blacklists).
## But you can still choose resolvers that do DNSSEC validation.


## Immediately respond to IPv6-related queries with an empty response
## This makes things faster when there is no IPv6 connectivity, but can
## also cause reliability issues with some stub resolvers.

block_ipv6 = false


## Immediately respond to A and AAAA queries for host names without a domain name

block_unqualified = true


## Immediately respond to queries for local zones instead of leaking them to
## upstream resolvers (always causing errors or timeouts).

block_undelegated = true


## TTL for synthetic responses sent when a request has been blocked (due to
## IPv6 or blacklists).

reject_ttl = 600



##################################################################################
#        Route queries for specific domains to a dedicated set of servers        #
##################################################################################

## Example map entries (one entry per line):
## example.com 9.9.9.9
## example.net 9.9.9.9,8.8.8.8,1.1.1.1

# forwarding_rules = 'forwarding-rules.txt'



###############################
#        Cloaking rules       #
###############################

## Cloaking returns a predefined address for a specific name.
## In addition to acting as a HOSTS file, it can also return the IP address
## of a different name. It will also do CNAME flattening.
##
## Example map entries (one entry per line)
## example.com     10.1.1.1
## www.google.com  forcesafesearch.google.com

# cloaking_rules = 'cloaking-rules.txt'

## TTL used when serving entries in cloaking-rules.txt

# cloak_ttl = 600


###########################
#        DNS cache        #
###########################

## Enable a DNS cache to reduce latency and outgoing traffic

cache = false


## Cache size

cache_size = 4096


## Minimum TTL for cached entries

cache_min_ttl = 2400


## Maximum TTL for cached entries

cache_max_ttl = 86400


## Minimum TTL for negatively cached entries

cache_neg_min_ttl = 60


## Maximum TTL for negatively cached entries

cache_neg_max_ttl = 600



##################################
#        Local DoH server        #
##################################

[local_doh]

## dnscrypt-proxy can act as a local DoH server. By doing so, web browsers
## requiring a direct connection to a DoH server in order to enable some
## features will enable these, without bypassing your DNS proxy.

## Addresses that the local DoH server should listen to

# listen_addresses = ['127.0.0.1:3000']


## Path of the DoH URL. This is not a file, but the part after the hostname
## in the URL. By convention, `/dns-query` is frequently chosen.
## For each `listen_address` the complete URL to access the server will be:
## `https://<listen_address><path>` (ex: `https://127.0.0.1/dns-query`)

# path = "/dns-query"


## Certificate file and key - Note that the certificate has to be trusted.
## See the documentation (wiki) for more information.

# cert_file = "localhost.pem"
# cert_key_file = "localhost.pem"



###############################
#        Query logging        #
###############################

## Log client queries to a file

[query_log]

  ## Path to the query log file (absolute, or relative to the same directory as the config file)
  ## Non non-Windows systems, can be /dev/stdout to log to the standard output (and set log_files_max_size to 0)

  # file = 'query.log'


  ## Query log format (currently supported: tsv and ltsv)

  format = 'tsv'


  ## Do not log these query types, to reduce verbosity. Keep empty to log everything.

  # ignored_qtypes = ['DNSKEY', 'NS']



############################################
#        Suspicious queries logging        #
############################################

## Log queries for nonexistent zones
## These queries can reveal the presence of malware, broken/obsolete applications,
## and devices signaling their presence to 3rd parties.

[nx_log]

  ## Path to the query log file (absolute, or relative to the same directory as the config file)

  # file = 'nx.log'


  ## Query log format (currently supported: tsv and ltsv)

  format = 'tsv'



######################################################
#        Pattern-based blocking (blacklists)        #
######################################################

## Blacklists are made of one pattern per line. Example of valid patterns:
##
##   example.com
##   =example.com
##   *sex*
##   ads.*
##   ads*.example.*
##   ads*.example[0-9]*.com
##
## Example blacklist files can be found at https://download.dnscrypt.info/blacklists/
## A script to build blacklists from public feeds can be found in the
## `utils/generate-domains-blacklists` directory of the dnscrypt-proxy source code.

[blacklist]

  ## Path to the file of blocking rules (absolute, or relative to the same directory as the config file)

  # blacklist_file = 'blacklist.txt'


  ## Optional path to a file logging blocked queries

  # log_file = 'blocked.log'


  ## Optional log format: tsv or ltsv (default: tsv)

  # log_format = 'tsv'



###########################################################
#        Pattern-based IP blocking (IP blacklists)        #
###########################################################

## IP blacklists are made of one pattern per line. Example of valid patterns:
##
##   127.*
##   fe80:abcd:*
##   192.168.1.4

[ip_blacklist]

  ## Path to the file of blocking rules (absolute, or relative to the same directory as the config file)

  # blacklist_file = 'ip-blacklist.txt'


  ## Optional path to a file logging blocked queries

  # log_file = 'ip-blocked.log'


  ## Optional log format: tsv or ltsv (default: tsv)

  # log_format = 'tsv'



######################################################
#   Pattern-based whitelisting (blacklists bypass)   #
######################################################

## Whitelists support the same patterns as blacklists
## If a name matches a whitelist entry, the corresponding session
## will bypass names and IP filters.
##
## Time-based rules are also supported to make some websites only accessible at specific times of the day.

[whitelist]

  ## Path to the file of whitelisting rules (absolute, or relative to the same directory as the config file)

  # whitelist_file = 'whitelist.txt'


  ## Optional path to a file logging whitelisted queries

  # log_file = 'whitelisted.log'


  ## Optional log format: tsv or ltsv (default: tsv)

  # log_format = 'tsv'



##########################################
#        Time access restrictions        #
##########################################

## One or more weekly schedules can be defined here.
## Patterns in the name-based blocklist can optionally be followed with @schedule_name
## to apply the pattern 'schedule_name' only when it matches a time range of that schedule.
##
## For example, the following rule in a blacklist file:
## *.youtube.* @time-to-sleep
## would block access to YouTube only during the days, and period of the days
## define by the 'time-to-sleep' schedule.
##
## {after='21:00', before= '7:00'} matches 0:00-7:00 and 21:00-0:00
## {after= '9:00', before='18:00'} matches 9:00-18:00

[schedules]

  # [schedules.'time-to-sleep']
  # mon = [{after='21:00', before='7:00'}]
  # tue = [{after='21:00', before='7:00'}]
  # wed = [{after='21:00', before='7:00'}]
  # thu = [{after='21:00', before='7:00'}]
  # fri = [{after='23:00', before='7:00'}]
  # sat = [{after='23:00', before='7:00'}]
  # sun = [{after='21:00', before='7:00'}]

  # [schedules.'work']
  # mon = [{after='9:00', before='18:00'}]
  # tue = [{after='9:00', before='18:00'}]
  # wed = [{after='9:00', before='18:00'}]
  # thu = [{after='9:00', before='18:00'}]
  # fri = [{after='9:00', before='17:00'}]



#########################
#        Servers        #
#########################

## Remote lists of available servers
## Multiple sources can be used simultaneously, but every source
## requires a dedicated cache file.
##
## Refer to the documentation for URLs of public sources.
##
## A prefix can be prepended to server names in order to
## avoid collisions if different sources share the same for
## different servers. In that case, names listed in `server_names`
## must include the prefixes.
##
## If the `urls` property is missing, cache files and valid signatures
## must be already present; This doesn't prevent these cache files from
## expiring after `refresh_delay` hours.

[sources]

  ## An example of a remote source from https://github.com/DNSCrypt/dnscrypt-resolvers

  [sources.'public-resolvers']
  urls = ['https://raw.githubusercontent.com/DNSCrypt/dnscrypt-resolvers/master/v2/public-resolvers.md', 'https://download.dnscrypt.info/resolvers-list/v2/public-resolvers.md']
  cache_file = 'public-resolvers.md'
  minisign_key = 'RWQf6LRCGA9i53mlYecO4IzT51TGPpvWucNSCh1CBM0QTaLn73Y7GFO3'
  prefix = ''

  ## Anonymized DNS relays

  [sources.'relays']
  urls = ['https://raw.githubusercontent.com/DNSCrypt/dnscrypt-resolvers/master/v2/relays.md', 'https://download.dnscrypt.info/resolvers-list/v2/relays.md']
  cache_file = 'relays.md'
  minisign_key = 'RWQf6LRCGA9i53mlYecO4IzT51TGPpvWucNSCh1CBM0QTaLn73Y7GFO3'
  refresh_delay = 72
  prefix = ''

  ## Quad9 over DNSCrypt - https://quad9.net/

  # [sources.quad9-resolvers]
  # urls = ['https://www.quad9.net/quad9-resolvers.md']
  # minisign_key = 'RWQBphd2+f6eiAqBsvDZEBXBGHQBJfeG6G+wJPPKxCZMoEQYpmoysKUN'
  # cache_file = 'quad9-resolvers.md'
  # prefix = 'quad9-'

  ## Another example source, with resolvers censoring some websites not appropriate for children
  ## This is a subset of the `public-resolvers` list, so enabling both is useless

  #  [sources.'parental-control']
  #  urls = ['https://raw.githubusercontent.com/DNSCrypt/dnscrypt-resolvers/master/v2/parental-control.md', 'https://download.dnscrypt.info/resolvers-list/v2/parental-control.md']
  #  cache_file = 'parental-control.md'
  #  minisign_key = 'RWQf6LRCGA9i53mlYecO4IzT51TGPpvWucNSCh1CBM0QTaLn73Y7GFO3'




#########################################
#        Servers with known bugs        #
#########################################

[broken_implementations]

# Cisco servers currently cannot handle queries larger than 1472 bytes, and don't
# truncate reponses larger than questions as expected by the DNSCrypt protocol.
# This prevents large responses from being received, and breaks relaying.
# A workaround for the first issue will be applied to servers in list below.
# Do not change that list until the bugs are fixed server-side.

broken_query_padding = ['cisco', 'cisco-ipv6', 'cisco-familyshield']




################################
#        Anonymized DNS        #
################################

[anonymized_dns]

## Routes are indirect ways to reach DNSCrypt servers.
##
## A route maps a server name ("server_name") to one or more relays that will be
## used to connect to that server.
##
## A relay can be specified as a DNS Stamp (either a relay stamp, or a
## DNSCrypt stamp), an IP:port, a hostname:port, or a server name.
##
## The following example routes "example-server-1" via `anon-example-1` or `anon-example-2`,
## and "example-server-2" via the relay whose relay DNS stamp
## is "sdns://gRIxMzcuNzQuMjIzLjIzNDo0NDM".
##
## !!! THESE ARE JUST EXAMPLES !!!
##
## Review the list of available relays from the "relays.md` file, and, for each
## server you want to use, define the relays you want connections to go through.
##
## Carefully choose relays and servers so that the are run by different entities.
##
## "server_name" can also be set to "*" to define a default route, but this is not
## recommended. if you do so, keep "server_names" short and distinct from relays.

# routes = [
#    { server_name='example-server-1', via=['anon-example-1', 'anon-example-2'] },
#    { server_name='example-server-2', via=['sdns://gRIxMzcuNzQuMjIzLjIzNDo0NDM'] }
# ]


## Optional, local, static list of additional servers
## Mostly useful for testing your own servers.

[static]

  # [static.'myserver']
  # stamp = 'sdns:AQcAAAAAAAAAAAAQMi5kbnNjcnlwdC1jZXJ0Lg'

[boostchicken]

I am not up to speed on dnscrypt-proxy. What does it do that cloudflared doesn't? It seems more complex / configurable, but whats the tangible benefit?

[pedropombeiro]

It does the same as cloudflared. The most tangible benefits are the reliability and the ability to use other services, not just cloudflare. You can whitelist providers (or a single one), and you can define the criteria that the selected DNS provider should have. dnscrypt-proxy will then select the fastest provider from the list of acceptable providers based on the criteria. So for people arriving at this project, they could see a minimal version of the the configuration file above allowing them to use cloudflare, instead of having to decide between the cloudflared and DoTE approaches.

[boostchicken]

I will see if I can make a variant

[gatesry]

@boostchicken - I just got dnscrypt-proxy working great on my UDM-P w/ PiHole as the primary. Will look to submit a PR soon.

[pedropombeiro]

I made an initial attempt to use https://hub.docker.com/r/klutchell/dnscrypt-proxy, but I realized I'd need to allocate a new IP on the dns network so that the pihole container can see it. I'm not too familiar with networking configuration in general, so I didn't proceed with it. I imagine it would be a case of allocating an additional IP in 20-dns.conflist, and to explicitly assign an IP to each of the containers, but apparently there is a bug in podman 1.6.0 which prevents that from working.

[gatesry]

I was able to get my stuff working. You need a new .conflist for every container.

[pedropombeiro]

That's awesome @gatesry! Would you care to open a pull request for the examples/ folder or maybe https://github.com/boostchicken-dev/udm-utilities/tree/master/run-pihole, so that others can easily find instructions on how to set that up?

[pedropombeiro]

@gatesry did you need to create a separate network, or are you able to create the additional IP on the same dns subnet?

[gatesry]

Yeah, I will create a PR. Just need time to think about the best place here to put it.

For the network, I had to create a separate network in podman, but the VLAN and gateway/subnet can be the same.

See below.

pihole - https://github.com/gatesry/udm-lab/blob/main/packages/dns/pihole/20-pihole.conflist
unbound - https://github.com/gatesry/udm-lab/blob/main/packages/dns/unbound/20-unbound.conflist

[pedropombeiro]

Gotcha, thanks!