refactor(ci): move DNS control method up to docker-compose.yml (#4341)

This is part of a yak shave towards CI testing of #3812 

Moving the DNS control method out of `docker-compose.yml` and up to the
integration tests themselves allows us to test these scenarios:

- `systemd-resolved`
- `etc-resolv-conf`
- `systemd-resolved` but we're in a container where that won't work, so
we should gracefully degrade to just allowing IP/CIDR resources

.github/workflows/_integration_tests.yml

@@ -106,8 +106,14 @@ jobs:
           direct-download-roaming-network,
           dns-etc-resolvconf,
           dns-nm,
+          dns-failsafe, # Uses the default DNS control method
           systemd/dns-systemd-resolved,
         ]
+        include:
+          - test: direct-download-roaming-network
+            dns-control: etc-resolv-conf
+          - test: dns-etc-resolvconf
+            dns-control: etc-resolv-conf
     steps:
       - uses: actions/checkout@v4
       - uses: ./.github/actions/gcp-docker-login
@@ -118,12 +124,16 @@ jobs:
         run: docker compose run elixir /bin/sh -c 'cd apps/domain && mix ecto.seed'
       - name: Start docker compose in the background
         run: |
+          # Overrides the Client's env var in docker-compose.yml
+          echo "FIREZONE_DNS_CONTROL=${{ matrix.dns-control }}" >> .env
+
           # Start one-by-one to avoid variability in service startup order
           docker compose up -d dns.httpbin httpbin download.httpbin
           docker compose up -d api web domain --no-build
           docker compose up -d relay --no-build
           docker compose up -d gateway --no-build
           docker compose up -d client --no-build
+          docker compose exec -it client env
 
       - run: ./scripts/tests/${{ matrix.test }}.sh
 

docker-compose.yml

@@ -283,7 +283,7 @@ services:
 
   client:
     environment:
-      FIREZONE_DNS_CONTROL: "etc-resolv-conf"
+      FIREZONE_DNS_CONTROL: "${FIREZONE_DNS_CONTROL}"
       FIREZONE_TOKEN: "n.SFMyNTY.g2gDaANtAAAAJGM4OWJjYzhjLTkzOTItNGRhZS1hNDBkLTg4OGFlZjZkMjhlMG0AAAAkN2RhN2QxY2QtMTExYy00NGE3LWI1YWMtNDAyN2I5ZDIzMGU1bQAAACtBaUl5XzZwQmstV0xlUkFQenprQ0ZYTnFJWktXQnMyRGR3XzJ2Z0lRdkZnbgYAGUmu74wBYgABUYA.UN3vSLLcAMkHeEh5VHumPOutkuue8JA6wlxM9JxJEPE"
       RUST_LOG: firezone_linux_client=trace,wire=trace,connlib_client_shared=trace,firezone_tunnel=trace,connlib_shared=trace,boringtun=debug,snownet=debug,str0m=debug,info
       FIREZONE_API_URL: ws://api:8081

scripts/tests/dns-failsafe.sh

@@ -0,0 +1,41 @@
+#!/usr/bin/env bash
+
+# If we set the DNS control to `systemd-resolved` but that's not available,
+# we should still boot up and allow IP / CIDR resources to work
+
+set -euo pipefail
+
+source "./scripts/tests/lib.sh"
+
+function client() {
+    docker compose exec -it client "$@"
+}
+
+function client_nslookup() {
+    # Skip the first 3 lines so that grep won't see the DNS server IP
+    # `tee` here copies stdout to stderr
+    client timeout 30 sh -c "nslookup $1 | tee >(cat 1>&2) | tail -n +4"
+}
+
+function gateway() {
+    docker compose exec -it gateway "$@"
+}
+
+# Re-up the gateway since a local dev setup may run this back-to-back
+docker compose up -d gateway --no-build
+
+echo "# make sure resolv.conf was not changed"
+client sh -c "cat /etc/resolv.conf"
+
+echo "# Make sure gateway can reach httpbin by DNS"
+gateway sh -c "curl --fail dns.httpbin/get"
+
+echo "# Access httpbin by IP"
+client_curl_resource
+
+echo "# Stop the gateway and make sure the resource is inaccessible"
+docker compose stop gateway
+client sh -c "curl --connect-timeout 15 --fail 172.20.0.100/get" && exit 1
+
+# Needed so that the previous failure doesn't bail out of the whole script
+exit 0