Release 0.4.1

changelog.md

@@ -1,4 +1,8 @@
-# 2022-04-20 - v0.4.0
+# 2022-06-13 - v0.4.1
+
+* Fix crash on failed SSHEx / Postgrex connection failure.
+
+# 2022-06-09 - v0.4.0
 
 * Add minimal clustering logic.
 * Fix crash on unavailable HAProxy socket.

lib/ha_handler/drbd/watcher.ex

@@ -28,8 +28,8 @@ defmodule HAHandler.DRBD.Watcher do
     hostname = Keyword.get(opts, :hostname)
     password = Keyword.get(opts, :password)
 
-    {:ok, pid} = connect(hostname, password)
-
+    case connect(hostname, password) do
+      {:ok, pid} ->
         state = %{
           backend: pid,
           hostname: hostname,
@@ -37,6 +37,14 @@ defmodule HAHandler.DRBD.Watcher do
         }
 
         {:ok, state}
+      {:error, err} ->
+
+        # Wait for 10 seconds so that the supervisor does not loop on dead node
+        # (and reach the max_restart threshold / stop trying).
+        Process.sleep(10_000)
+
+        {:error, err}
+    end
   end
 
   @impl true

lib/ha_handler/pgsql/watcher.ex

@@ -12,14 +12,19 @@ defmodule HAHandler.PGSQL.Watcher do
     # successful.
     # TODO: set dbconnections backoff and connect hooks
     # See https://github.com/elixir-ecto/db_connection/blob/master/lib/db_connection.ex#L343
-    {:ok, pid} = Postgrex.start_link(opts)
-
+    case Postgrex.start_link(opts) do
+      {:ok, pid} ->
         state = %{
           backend: pid,
           hostname: Keyword.get(opts, :hostname)
         }
 
         {:ok, state}
+      {:error, err} ->
+        # Will be catched by the supervisor if anything happen. It should not
+        # be triggered even if a PGSQL node down, since Postgrex has its own
+        # surpervision tree.
+        {:error, err}
   end
 
   @impl true

mix.exs

@@ -4,7 +4,7 @@ defmodule HAHandler.MixProject do
   def project do
     [
       app: :ha_handler,
-      version: "0.4.0",
+      version: "0.4.1",
       elixir: "~> 1.12",
       start_permanent: Mix.env() == :prod,
       deps: deps(),