test: in-process Redis fake instead of skipping Redis-backed tests

Replaces the previous :requires_redis skip with a real test double so
the suite stays self-contained — no external Redis service needed in
CI, but the call paths through RedisHealthCheck and FourOhFourTracker
still get exercised end-to-end.

- Towerops.Redix.Fake: GenServer that speaks PING / SADD / SCARD /
  SISMEMBER / SMEMBERS / EXPIRE / DEL / FLUSHDB. start_link refuses
  anything other than localhost:6379 so failure-path tests still fail
  the way they did against real Redix.
- Application supervisor starts the fake under name Towerops.Redix in
  test env (instead of the dummy Agent stub).
- RedisHealthCheck and FourOhFourTracker now route Redix calls through
  Application.get_env(:towerops, :redix_module, Redix). Test config
  points it at the fake.
- Drops the test_helper :requires_redis exclusion and the @moduletag
  on the two test files; FourOhFourTrackerTest setup just resets the
  fake between runs.

config/test.exs

@@ -127,5 +127,8 @@ config :towerops,
   device_deletion_delay_ms: 10,
   # Reduce DiscoveryWorker poll cadence (25ms vs 500ms in prod)
   discovery_wait_interval_ms: 25,
+  # Use the in-process Redis fake so tests don't depend on a real
+  # Redis service (CI has none).
+  redix_module: Towerops.Redix.Fake,
   # Stripe test configuration
   stripe_product_id: "prod_test"

lib/towerops/application.ex

@@ -214,26 +214,33 @@ defmodule Towerops.Application do
   defp redis_spec do
     redis_config = Application.get_env(:towerops, :redis, [])
 
-    if Keyword.has_key?(redis_config, :host) do
-      opts = [
-        host: Keyword.get(redis_config, :host, "localhost"),
-        port: Keyword.get(redis_config, :port, 6379),
-        name: Towerops.Redix
-      ]
+    cond do
+      # In test, run an in-process Redis fake so the suite stays
+      # self-contained — CI doesn't ship a Redis service.
+      Application.get_env(:towerops, :env) == :test ->
+        {Towerops.Redix.Fake, name: Towerops.Redix}
 
-      # Add password if configured
-      opts =
-        case Keyword.get(redis_config, :password) do
-          nil -> opts
-          "" -> opts
-          password -> Keyword.put(opts, :password, password)
-        end
+      Keyword.has_key?(redis_config, :host) ->
+        opts = [
+          host: Keyword.get(redis_config, :host, "localhost"),
+          port: Keyword.get(redis_config, :port, 6379),
+          name: Towerops.Redix
+        ]
 
-      {Redix, opts}
-    else
-      # No Redis configured - start a stub process that will fail gracefully
-      # This allows the app to start in environments without Redis
-      Supervisor.child_spec({Agent, fn -> nil end}, id: Towerops.Redix)
+        # Add password if configured
+        opts =
+          case Keyword.get(redis_config, :password) do
+            nil -> opts
+            "" -> opts
+            password -> Keyword.put(opts, :password, password)
+          end
+
+        {Redix, opts}
+
+      true ->
+        # No Redis configured - start a stub process that will fail gracefully
+        # This allows the app to start in environments without Redis
+        Supervisor.child_spec({Agent, fn -> nil end}, id: Towerops.Redix)
     end
   end
 

lib/towerops/redis_health_check.ex

@@ -90,16 +90,18 @@ defmodule Towerops.RedisHealthCheck do
         conn_opts
       end
 
+    redix = redix_module()
+
     # Start a temporary connection for health check
-    case Redix.start_link(conn_opts) do
+    case redix.start_link(conn_opts) do
       {:ok, conn} ->
         result =
-          case Redix.command(conn, ["PING"], timeout: timeout_ms) do
+          case redix.command(conn, ["PING"], timeout: timeout_ms) do
             {:ok, "PONG"} -> :ok
             {:error, reason} -> {:error, reason}
           end
 
-        Redix.stop(conn)
+        redix.stop(conn)
         result
 
       {:error, reason} ->
@@ -115,4 +117,9 @@ defmodule Towerops.RedisHealthCheck do
       Logger.debug("Redis health check exit: #{inspect(reason)}")
       {:error, reason}
   end
+
+  # Allows the test suite to swap in `Towerops.Redix.Fake` so health
+  # checks don't reach out to a real Redis. Defaults to `Redix` in
+  # every other environment.
+  defp redix_module, do: Application.get_env(:towerops, :redix_module, Redix)
 end

lib/towerops/redix/fake.ex

@@ -0,0 +1,116 @@
+defmodule Towerops.Redix.Fake do
+  @moduledoc """
+  In-process stand-in for Redix that speaks just enough of the Redis
+  command protocol for our test suite. CI doesn't run a Redis service,
+  and we don't want our unit tests to depend on one — anything in the
+  Towerops codebase that talks to Redis goes through a configurable
+  client module (`:towerops, :redix_module`), and tests point that at
+  this fake.
+
+  Supported commands: `PING`, `SADD`, `SCARD`, `SISMEMBER`, `SMEMBERS`,
+  `EXPIRE`, `DEL`, `FLUSHDB`. Anything else returns `{:error,
+  :unsupported_command}` — extend as new commands appear in the codebase.
+
+  The signature mirrors `Redix.start_link/1`, `command/3`, and `stop/1`
+  so the modules using it stay agnostic of which client is wired up.
+  """
+
+  use GenServer
+
+  @reachable_hosts ["localhost", "127.0.0.1", "::1"]
+
+  # ── Redix-compatible API ────────────────────────────────────────────
+  def start_link(opts \\ []) do
+    # Simulate connection refusal when the caller asks for anything that
+    # isn't standard localhost:6379. RedisHealthCheck failure-path tests
+    # dial unreachable hosts/ports specifically to drive error branches,
+    # and we'd silently return `{:ok, _}` if we accepted everything.
+    host = Keyword.get(opts, :host, "localhost")
+    port = Keyword.get(opts, :port, 6379)
+
+    if host in @reachable_hosts and port == 6379 do
+      {name, _opts} = Keyword.pop(opts, :name)
+      GenServer.start_link(__MODULE__, %{}, if(name, do: [name: name], else: []))
+    else
+      {:error, %{__struct__: Redix.ConnectionError, reason: :econnrefused}}
+    end
+  end
+
+  def command(conn, cmd, _opts \\ []) do
+    GenServer.call(conn, {:command, cmd})
+  end
+
+  def stop(conn) do
+    if Process.alive?(conn), do: GenServer.stop(conn, :normal), else: :ok
+  end
+
+  @doc """
+  Resets the fake Redis state. Use in test setup blocks to keep tests
+  isolated.
+  """
+  def reset(conn \\ __MODULE__) do
+    GenServer.call(conn, :reset)
+  end
+
+  # ── GenServer ────────────────────────────────────────────────────────
+  @impl true
+  def init(_), do: {:ok, %{sets: %{}}}
+
+  @impl true
+  def handle_call({:command, cmd}, _from, state) do
+    {reply, state} = run(cmd, state)
+    {:reply, reply, state}
+  end
+
+  def handle_call(:reset, _from, _state), do: {:reply, :ok, %{sets: %{}}}
+
+  # ── Command dispatch ────────────────────────────────────────────────
+  defp run(["PING"], state), do: {{:ok, "PONG"}, state}
+  defp run(["PING" | _], state), do: {{:ok, "PONG"}, state}
+
+  defp run(["SADD", key, value | rest], state) do
+    set = Map.get(state.sets, key, MapSet.new())
+    members = [value | rest]
+
+    {new_set, added} =
+      Enum.reduce(members, {set, 0}, fn m, {acc, count} ->
+        if MapSet.member?(acc, m), do: {acc, count}, else: {MapSet.put(acc, m), count + 1}
+      end)
+
+    {{:ok, added}, %{state | sets: Map.put(state.sets, key, new_set)}}
+  end
+
+  defp run(["SCARD", key], state) do
+    size = state.sets |> Map.get(key, MapSet.new()) |> MapSet.size()
+    {{:ok, size}, state}
+  end
+
+  defp run(["SISMEMBER", key, value], state) do
+    in? = state.sets |> Map.get(key, MapSet.new()) |> MapSet.member?(value)
+    {{:ok, if(in?, do: 1, else: 0)}, state}
+  end
+
+  defp run(["SMEMBERS", key], state) do
+    members = state.sets |> Map.get(key, MapSet.new()) |> MapSet.to_list()
+    {{:ok, members}, state}
+  end
+
+  # EXPIRE is a no-op — tests don't need the TTL semantics, just the
+  # success ack.
+  defp run(["EXPIRE", _key, _ttl], state), do: {{:ok, 1}, state}
+
+  defp run(["DEL" | keys], state) do
+    {sets, removed} =
+      Enum.reduce(keys, {state.sets, 0}, fn k, {acc, count} ->
+        if Map.has_key?(acc, k),
+          do: {Map.delete(acc, k), count + 1},
+          else: {acc, count}
+      end)
+
+    {{:ok, removed}, %{state | sets: sets}}
+  end
+
+  defp run(["FLUSHDB"], _state), do: {{:ok, "OK"}, %{sets: %{}}}
+
+  defp run(_unknown, state), do: {{:error, :unsupported_command}, state}
+end

lib/towerops/security/four_oh_four_tracker.ex

@@ -23,15 +23,16 @@ defmodule Towerops.Security.FourOhFourTracker do
   def record_404(ip_address, path) do
     key = redis_key(ip_address)
     conn = Towerops.Redix
+    redix = redix_module()
 
     try do
       # Add path to set
-      _ = Redix.command(conn, ["SADD", key, path])
+      _ = redix.command(conn, ["SADD", key, path])
       # Set TTL if this is the first path
-      _ = Redix.command(conn, ["EXPIRE", key, @window_seconds])
+      _ = redix.command(conn, ["EXPIRE", key, @window_seconds])
 
       # Get unique count
-      case Redix.command(conn, ["SCARD", key]) do
+      case redix.command(conn, ["SCARD", key]) do
         {:ok, count} when count >= @threshold ->
           Logger.warning("IP #{ip_address} hit threshold with #{count} unique 404s, triggering ban")
           BruteForce.create_or_escalate_ban(ip_address)
@@ -58,9 +59,10 @@ defmodule Towerops.Security.FourOhFourTracker do
   def get_count(ip_address) do
     key = redis_key(ip_address)
     conn = Towerops.Redix
+    redix = redix_module()
 
     try do
-      Result.unwrap_or(Redix.command(conn, ["SCARD", key]), 0)
+      Result.unwrap_or(redix.command(conn, ["SCARD", key]), 0)
     catch
       :exit, _ -> 0
     end
@@ -69,4 +71,8 @@ defmodule Towerops.Security.FourOhFourTracker do
   defp redis_key(ip_address) do
     "404_scan:#{ip_address}"
   end
+
+  # Lets the test suite swap in `Towerops.Redix.Fake` so this module
+  # doesn't reach out to a real Redis from CI.
+  defp redix_module, do: Application.get_env(:towerops, :redix_module, Redix)
 end

test/test_helper.exs

@@ -47,28 +47,13 @@ ExUnit.start()
 # aren't installed (e.g. CI test gate runs in a lean image without
 # gdal-bin). Locally we install GDAL via `brew install gdal` per
 # CLAUDE.md, so the test runs there.
-gdal_excludes =
+extra_excludes =
   if System.find_executable("gdal_translate") && System.find_executable("gdaldem") do
     []
   else
     [:requires_gdal]
   end
 
-# Skip Redis-backed tests when Redis isn't reachable on localhost:6379
-# (e.g. the Forgejo CI test gate has no Redis service). Local dev runs
-# Redis via direnv/services so the tests run there.
-redis_excludes =
-  case :gen_tcp.connect(~c"127.0.0.1", 6379, [:binary, active: false], 250) do
-    {:ok, sock} ->
-      :gen_tcp.close(sock)
-      []
-
-    {:error, _} ->
-      [:requires_redis]
-  end
-
-extra_excludes = gdal_excludes ++ redis_excludes
-
 # Exclude tests by default
 # Run specific tests with: mix test --only <tag>
 ExUnit.configure(

test/towerops/redis_health_check_test.exs

@@ -5,8 +5,6 @@ defmodule Towerops.RedisHealthCheckTest do
 
   alias Towerops.RedisHealthCheck
 
-  @moduletag :requires_redis
-
   describe "wait_for_redis/2" do
     test "returns :ok when Redis is available" do
       redis_config = Application.get_env(:towerops, :redis, host: "localhost", port: 6379)

test/towerops/security/four_oh_four_tracker_test.exs

@@ -1,56 +1,18 @@
 defmodule Towerops.Security.FourOhFourTrackerTest do
   use Towerops.DataCase
 
+  alias Towerops.Redix.Fake, as: RedixFake
   alias Towerops.Security.FourOhFourTracker
 
   @moduletag :four_oh_four_tracker
-  @moduletag :requires_redis
 
   setup do
-    # Ensure a real Redix connection exists for these tests.
-    # The application may have started a stub Agent instead of real Redix.
-    ensure_redix_connection()
-    :ok
-  end
-
-  defp ensure_redix_connection do
-    case Process.whereis(Towerops.Redix) do
-      nil ->
-        start_real_redix()
-
-      pid ->
-        # Check if it's a real Redix process or a stub
-        try do
-          case Redix.command(pid, ["PING"]) do
-            {:ok, "PONG"} -> :ok
-            _ -> restart_as_real_redix()
-          end
-        catch
-          :exit, _ -> restart_as_real_redix()
-        end
-    end
-  end
-
-  defp restart_as_real_redix do
+    # In test mode `Towerops.Redix` is the in-process fake — wipe it so
+    # state doesn't leak between tests.
     if pid = Process.whereis(Towerops.Redix) do
-      try do
-        Process.unregister(Towerops.Redix)
-      rescue
-        ArgumentError -> :ok
-      end
-
-      try do
-        GenServer.stop(pid, :normal, 100)
-      catch
-        :exit, _ -> :ok
-      end
+      RedixFake.reset(pid)
     end
 
-    start_real_redix()
-  end
-
-  defp start_real_redix do
-    {:ok, _pid} = Redix.start_link(host: "localhost", port: 6379, name: Towerops.Redix)
     :ok
   end