Prefer to use after_transition

Before_transition does not play nice when used with
OptimisticLocking and this can result in triggering
multiple events

Closes https://gitlab.com/gitlab-org/gitlab-ce/issues/55306

app/models/ci/build.rb

@@ -220,6 +220,10 @@ module Ci
         next unless build.project
 
         build.deployment&.drop
+      end
+
+      after_transition any => [:failed] do |build|
+        next unless build.project
 
         if build.retry_failure?
           begin

app/models/clusters/applications/ingress.rb

@@ -23,7 +23,7 @@ module Clusters
       FETCH_IP_ADDRESS_DELAY = 30.seconds
 
       state_machine :status do
-        before_transition any => [:installed] do |application|
+        after_transition any => [:installed] do |application|
           application.run_after_commit do
             ClusterWaitForIngressIpAddressWorker.perform_in(
               FETCH_IP_ADDRESS_DELAY, application.name, application.id)

app/models/clusters/applications/knative.rb

@@ -20,7 +20,7 @@ module Clusters
       self.reactive_cache_key = ->(knative) { [knative.class.model_name.singular, knative.id] }
 
       state_machine :status do
-        before_transition any => [:installed] do |application|
+        after_transition any => [:installed] do |application|
           application.run_after_commit do
             ClusterWaitForIngressIpAddressWorker.perform_in(
               FETCH_IP_ADDRESS_DELAY, application.name, application.id)

app/models/merge_request.rb

@@ -105,7 +105,9 @@ class MergeRequest < ActiveRecord::Base
 
     before_transition any => :opened do |merge_request|
       merge_request.merge_jid = nil
+    end
 
+    after_transition any => :opened do |merge_request|
       merge_request.run_after_commit do
         UpdateHeadPipelineForMergeRequestWorker.perform_async(merge_request.id)
       end