Gun: Publisher job behavior improvement

Gun's connection pool also returns an error if duplicate workers are launched simultaneously. Snooze on this error as well, and lower the snooze to 3 seconds with the optimism that the connection will still be open by then and the delivery can be completed quickly.

The original setting of 30 seconds is pretty high and means there's an unnatural lag between deliveries of activities destined to the same server that were created at nearly the same time. This configuration should be more efficient.

2 files changed, 8 insertions, 1 deletions

diff --git a/changelog.d/oban_gun_snooze.change b/changelog.d/oban_gun_snooze.change
new file mode 100644
index 000000000..c94525b2a
--- /dev/null
+++ b/changelog.d/oban_gun_snooze.change
@@ -0,0 +1 @@
+Publisher behavior improvement when snoozing Oban jobs due to Gun connection pool contention.
diff --git a/lib/pleroma/web/activity_pub/publisher.ex b/lib/pleroma/web/activity_pub/publisher.ex
index e040753dc..e63b8ff1f 100644
--- a/lib/pleroma/web/activity_pub/publisher.ex
+++ b/lib/pleroma/web/activity_pub/publisher.ex
@@ -143,9 +143,13 @@ defmodule Pleroma.Web.ActivityPub.Publisher do
           _ -> {:error, e}
         end
 
+      {:error, {:already_started, _}} ->
+        Logger.debug("Publisher snoozing worker job due worker :already_started race condition")
+        connection_pool_snooze()
+
       {:error, :pool_full} ->
         Logger.debug("Publisher snoozing worker job due to full connection pool")
-        {:snooze, 30}
+        connection_pool_snooze()
 
       e ->
         unless params[:unreachable_since], do: Instances.set_unreachable(inbox)
@@ -155,6 +159,8 @@ defmodule Pleroma.Web.ActivityPub.Publisher do
     end
   end
 
+  defp connection_pool_snooze, do: {:snooze, 3}
+
   defp signature_host(%URI{port: port, scheme: scheme, host: host}) do
     if port == URI.default_port(scheme) do
       host