Trigger stop on db write failure (digital-asset#6233)

* Stop running trigger under DB write failure changelog_begin changelog_end * Another fix relating to initialization retries changelog_begin changelog_end * Handle the error case directly
kgeonw · Jun 4, 2020 · 71ff65d · 71ff65d
1 parent 884d4d8
commit 71ff65d
Showing 1 changed file with 16 additions and 9 deletions.
diff --git a/triggers/service/src/main/scala/com/digitalasset/daml/lf/engine/trigger/ServiceMain.scala b/triggers/service/src/main/scala/com/digitalasset/daml/lf/engine/trigger/ServiceMain.scala
@@ -393,26 +393,33 @@ object Server {
             server.logTriggerStatus(runningTrigger, "running")
             server.addRunningTrigger(runningTrigger) match {
               case Left(err) =>
-                val msg = "Failed to add running trigger to database.\n" + err
-                ctx.self ! TriggerInitializationFailure(runningTrigger, msg)
+                // The trigger has just advised it's in the running
+                // state but updating the running trigger table has
+                // failed. This error condition is exogenous to the
+                // runner. We therefore need to tell it explicitly to
+                // stop.
+                server.logTriggerStatus(
+                  runningTrigger,
+                  "stopped: initialization failure (db write failure)")
+                runningTrigger.runner ! TriggerRunner.Stop
                 Behaviors.same
               case Right(()) => Behaviors.same
             }
           case TriggerInitializationFailure(runningTrigger, cause) =>
-            // The trigger has failed to start.
+            // The trigger has failed to start. No need to update the
+            // running triggers tables since this trigger never made
+            // it there.
             server.logTriggerStatus(runningTrigger, "stopped: initialization failure")
-            // No need to update the running triggers tables since
-            // this trigger never made it there.
+            // Don't send any messages to the runner here (it's under
+            // the management of a supervision strategy).
             Behaviors.same
           case TriggerRuntimeFailure(runningTrigger, cause) =>
             // The trigger has failed. Remove it from the running
             // triggers tables.
             server.logTriggerStatus(runningTrigger, "stopped: runtime failure")
             server.removeRunningTrigger(runningTrigger)
-            // Don't send any messages to the runner. Its supervision
-            // strategy will automatically restart the trigger up to
-            // some number of times beyond which it will remain
-            // stopped.
+            // Don't send any messages to the runner here (it's under
+            // the management of a supervision strategy).
             Behaviors.same
           case GetServerBinding(replyTo) =>
             replyTo ! binding