fix: Improve errors logged out and erpc error handling (#1226)

--------- Co-authored-by: Stas <[email protected]>
supabase · Nov 18, 2024 · 31401d2 · 31401d2
1 parent b2803d6
commit 31401d2
Show file tree

Hide file tree

Showing 6 changed files with 38 additions and 41 deletions.
diff --git a/README.md b/README.md
@@ -228,7 +228,8 @@ This is the list of operational codes that can help you understand your deployme
 | ErrorExecutingTransaction          | Error executing a database transaction in tenant database                                                                           |
 | SynInitializationError             | Our framework to syncronize processes has failed to properly startup a connection to the database                                   |
 | JanitorFailedToDeleteOldMessages   | Scheduled task for realtime.message cleanup was unable to run                                                                       |
-| UnknownError                       | An unknown error occurred                                                                                                           |
+| UnknownErrorOnController           | An error we are not handling correctly was triggered on a controller                                                                |
+| UnknownErrorOnChannel              | An error we are not handling correctly was triggered on a channel                                                                   |
 
 ## License
 

diff --git a/lib/realtime/database.ex b/lib/realtime/database.ex
@@ -237,14 +237,19 @@ defmodule Realtime.Database do
   def transaction(db_conn, func, opts) when node() == node(db_conn),
     do: transaction_catched(db_conn, func, opts)
 
-  def transaction(db_conn, func, opts),
-    do: Rpc.enhanced_call(node(db_conn), __MODULE__, :transaction, [db_conn, func, opts])
+  def transaction(db_conn, func, opts) do
+    metadata =
+      Keyword.take(Logger.metadata(), [:project_id, :tenant_id])
+      |> Keyword.put(:target, node(db_conn))
 
-  defp transaction_catched(db_conn, func, opts) do
+    Rpc.enhanced_call(node(db_conn), __MODULE__, :transaction, [db_conn, func, opts, metadata])
+  end
+
+  defp transaction_catched(db_conn, func, opts, metadata \\ []) do
     Postgrex.transaction(db_conn, func, opts)
   rescue
     e ->
-      log_error("ErrorExecutingTransaction", e)
+      log_error("ErrorExecutingTransaction", e, metadata)
       {:error, e}
   end
 

diff --git a/lib/realtime/rpc.ex b/lib/realtime/rpc.ex
@@ -36,38 +36,29 @@ defmodule Realtime.Rpc do
   def enhanced_call(node, mod, func, args \\ [], opts \\ []) do
     timeout = Keyword.get(opts, :timeout, Application.get_env(:realtime, :rpc_timeout))
 
-    try do
-      :timer.tc(fn -> :erpc.call(node, mod, func, args, timeout) end)
-    catch
-      kind, reason ->
-        log_error("ErrorOnRpcCall", %{target: node, mod: mod, func: func, error: {kind, reason}})
-        {:error, "RPC call error"}
-    else
-      {_, {:EXIT, reason}} ->
-        log_error("ErrorOnRpcCall", %{target: node, mod: mod, func: func, error: {:EXIT, reason}},
-          mod: mod,
-          func: func,
-          target: node
-        )
+    with {latency, {status, _} = response} <-
+           :timer.tc(fn -> :erpc.call(node, mod, func, args, timeout) end) do
+      Telemetry.execute(
+        [:realtime, :rpc],
+        %{latency: latency, success?: status == :ok},
+        %{mod: mod, func: func, target_node: node, origin_node: node()}
+      )
 
-        {:error, "RPC call error"}
-
-      {latency, response} ->
-        Telemetry.execute(
-          [:realtime, :rpc],
-          %{latency: latency},
-          %{
-            mod: mod,
-            func: func,
-            target_node: node,
-            origin_node: node()
-          }
-        )
-
-        case response do
-          {status, _} when status in [:ok, :error] -> response
-          _ -> {:error, response}
-        end
+      case response do
+        {status, _} when status in [:ok, :error] -> response
+        _ -> {:error, response}
+      end
     end
+  catch
+    kind, reason ->
+      log_error(
+        "ErrorOnRpcCall",
+        %{target: node, mod: mod, func: func, error: {kind, reason}},
+        mod: mod,
+        func: func,
+        target: node
+      )
+
+      {:error, "RPC call error"}
   end
 end
diff --git a/lib/realtime_web/channels/realtime_channel.ex b/lib/realtime_web/channels/realtime_channel.ex
@@ -141,10 +141,10 @@ defmodule RealtimeWeb.RealtimeChannel do
         )
 
       {:error, error} ->
-        Logging.log_error_message(:error, "UnknownError", error)
+        Logging.log_error_message(:error, "UnknownErrorOnChannel", error)
 
       error ->
-        Logging.log_error_message(:error, "UnknownError", error)
+        Logging.log_error_message(:error, "UnknownErrorOnChannel", error)
     end
   end
 
@@ -419,7 +419,7 @@ defmodule RealtimeWeb.RealtimeChannel do
         {:error, :too_many_joins}
 
       error ->
-        Logging.log_error_message(:error, "UnknownError", error)
+        Logging.log_error_message(:error, "UnknownErrorOnCounter", error)
         {:error, error}
     end
   end

diff --git a/lib/realtime_web/controllers/fallback_controller.ex b/lib/realtime_web/controllers/fallback_controller.ex
@@ -68,7 +68,7 @@ defmodule RealtimeWeb.FallbackController do
   end
 
   def call(conn, response) do
-    log_error("UnknownError", response)
+    log_error("UnknownErrorOnController", response)
 
     conn
     |> put_status(:unprocessable_entity)

diff --git a/mix.exs b/mix.exs
@@ -4,7 +4,7 @@ defmodule Realtime.MixProject do
   def project do
     [
       app: :realtime,
-      version: "2.33.48",
+      version: "2.33.49",
       elixir: "~> 1.16.0",
       elixirc_paths: elixirc_paths(Mix.env()),
       start_permanent: Mix.env() == :prod,