DD-WRT: Connecting to the Optimal ProtonVPN Server with OpenVPN

In a recent post, I outlined a solution that I was testing for configuring my DD-WRT router to connect to an optimal ProtonVPN server. In short, I configured a domain name that I own to return a round-robin IP address from a set of generally optimal servers for my location with Cloudflare.

This solution partially mimicked ProtonVPN’s country configuration option that allows you to connect to us.protonvpn.com which should resolve to the most ideal US IP for your location and the current server load. In my testing, however, this domain often connected me to servers that were physically far from my location with sub-par speed and latency.

My custom domain solution worked alright but did not take into account the servers’ current loads. Server loads seem to be fluctuating more wildly during the pandemic, and performance degradation can be quite noticeable during peak loads. ProtonVPN exposes an API endpoint here that provides up-to-date Load and Score metrics for each server. In this post, I outline how I automated selecting a more optimal server IP via a shell script on DD-WRT. This solution assumes that you already have the DD-WRT OpenVPN Client enabled and connected to ProtonVPN (ProtonVPN docs).

Entware

To parse the JSON results from the ProtonVPN API, I wanted to use the popular jq command-line tool. DD-WRT does not ship with this tool out of the box, however, so I first had to install the Entware package manager.

The DD-WRT wiki page provides fairly straightforward instructions on how to format a USB flash drive and install the package manager. I connected to my router via telnet. The installation went smoothly for me following that guide, but there could certainly be room for errors or hiccups, especially amongst various routers, chipsets, and DD-WRT builds.

Once Entware was installed, I installed jq with:

opkg install jq

curl

The next step of the process was curling the ProtonVPN API and piping the results through a jq query to select the currently optimal IP address.

I am not sure if their API respects the no-cache header, but I decided to pass it along regardless in the hopes of always receiving the freshest results. The API returns JSON by default, but I prefer to be explicit and request a response of application/json.

curl -s -H "Cache-Control: no-cache" -H "Accept: application/json" https://api.protonmail.ch/vpn/logicals

I have noticed that sometimes the ssl negotiation fails when calling the API with curl on DD-WRT. I suspect this to be an issue with my DD-WRT configuration. The workaround, although undoubtedly a bit insecure, is passing the -k (--insecure) flag to curl.

jq

Below, I filter the JSON response with jq. I select a Status of 1 which means to only consider servers that are currently online. Since I pay for a Plus subscription, I only want to connect to Tier 2 servers which theoretically have more available capacity as they are reserved for higher-paying subscribers. I only want to connect to a server in one of the two nearest metros to my current location to reduce latency. I specified that I do not want one of ProtonVPN’s tor nodes due to the much higher performance penalty that incurs.

With that subset of servers selected, I applied a sort by Score and then by Load. Lastly, I selected the top IP address from the list.

echo "$LOGICALS" | jq '.LogicalServers | map(select(.Status == 1 and .Tier == 2 and (.Name | (startswith("US-TX#") or startswith("US-GA#")) and (endswith("TOR") | not)))) | [sort_by(.Score, .Load)[]][0] | .Servers[0].EntryIP'

Shell Script

Here is the final script I am using (also available as a gist). Since I run the script from cron, I needed to define the PATH so the script knows where to locate tools like jq and curl. I then kill the running instance of openvpn and wait for it to be terminated. Since ProtonVPN’s API factors in my current location in its Score values, the call to their API is the one network request that I do want to execute on my raw WAN connection.

After fetching and filtering to the optimal server IP, I update the openvpn.conf and then restart the openvpn service.

#!/bin/sh

# specify PATH to run from cron
PATH=/bin:/usr/bin:/sbin:/usr/sbin:/jffs/sbin:/jffs/bin:/jffs/usr/sbin:/jffs/usr/bin:/mmc/sbin:/mmc/bin:/mmc/usr/sbin:/mmc/usr/bin:/opt/sbin:/opt/bin:/opt/usr/sbin:/opt/usr/bin

# kill openvpn
PID=$(pidof openvpn)
kill -s SIGTERM "$PID"
while $(kill -0 "$PID" 2>/dev/null); do
  sleep 1
done

# fetch ProtonVPN server info
LOGICALS=$(curl -s -H "Cache-Control: no-cache" -H "Accept: application/json" https://api.protonmail.ch/vpn/logicals)

# query for optimal server
IPSTRING=$(echo "$LOGICALS" | jq '.LogicalServers | map(select(.Status == 1 and .Tier == 2 and (.Name | (startswith("US-TX#") or startswith("US-GA#")) and (endswith("TOR") | not)))) | [sort_by(.Score, .Load)[]][0] | .Servers[0].EntryIP')

# update openvpn config
if [ -n "$IPSTRING" ]
then
  sed -i '/^remote/d' /tmp/openvpncl/openvpn.conf
  IPSTRING="${IPSTRING%\"}"
  IP="${IPSTRING#\"}"
  REMOTE="remote ${IP} 443"
  echo $REMOTE >> /tmp/openvpncl/openvpn.conf
fi

# restart openvpn
openvpn --config /tmp/openvpncl/openvpn.conf --route-up /tmp/openvpncl/route-up.sh --route-pre-down /tmp/openvpncl/route-down.sh --daemon

Cron

The final step of the process was setting up cron jobs to execute this script. I just used the DD-WRT GUI to refresh the connection twice a day (7:25 am and 5 pm, the beginning and end of my typical workday).

Conclusion

I have been using this solution for about a week, and I have been running speed tests at various times of the day. Every time I have tested, I have been able to max out my ISP plan’s bandwidth while connected to a ProtonVPN server without any issue. I also have been able to connect to teleconferencing tools without any noticeable latency or quality degradation.