poll instead of hardcoded sleep. 1 testcase can fail or timeout

CHANGELOG.md

@@ -3,6 +3,7 @@
 # 3.5.1
 
 - Fix local run for integration tests
+- Wait for creation of serviceaccount 'default' in newly created namespace to avoid race conditions with pod creation failures
 
 
 # 3.5.0

test/helpers/test_provisioner.rb

@@ -52,11 +52,25 @@ def prepare_pv(name, storage_class_name: nil)
 
     private
 
+    def wait_for_default_service_account(kubeclient, namespace)
+      30.times do
+        begin
+          sa = kubeclient.get_service_account('default', namespace)
+          return if sa
+        rescue Kubeclient::ResourceNotFoundError
+          # If the service account is not found, sleep for a second and then retry
+          sleep(1)
+        end
+      end
+      raise "Default service account in #{namespace} not ready after 30 seconds"
+    end
+
     def create_namespace(namespace)
       ns = Kubeclient::Resource.new(kind: 'Namespace')
       ns.metadata = { name: namespace }
       kubeclient.create_namespace(ns)
-      sleep(5) # wait for the serviceaccount 'default' to be created; https://github.com/kubernetes/kubernetes/issues/66689
+      # wait for the serviceaccount 'default' to be created; https://github.com/kubernetes/kubernetes/issues/66689
+      wait_for_default_service_account(kubeclient, namespace)
     end
   end
 end

test/integration/krane_deploy_test.rb

@@ -1350,7 +1350,7 @@ def test_resource_watcher_reports_failed_after_timeout
       f["missing_volumes.yml"]["Deployment"].first["spec"]["progressDeadlineSeconds"] = 30
       f["cannot_run.yml"]["Deployment"].first["spec"]["replicas"] = 1
     end
-    assert_deploy_failure(result)
+    assert_deploy_failure_or_timeout(result)
 
     bad_probe_timeout = "Deployment/bad-probe: TIMED OUT (progress deadline: 5s)"
 

test/test_helper.rb

@@ -116,6 +116,14 @@ def assert_deploy_failure(result, cause = nil)
     alias_method :assert_restart_failure, :assert_deploy_failure
     alias_method :assert_task_run_failure, :assert_deploy_failure
 
+    def assert_deploy_failure_or_timeout(result)
+      assert_equal(false, result, "Deploy succeeded when it was expected to fail.#{logs_message_if_captured}")
+      logging_assertion do |logs|
+        assert(logs.include?("Result: FAILURE") || logs.include?("Result: TIMED OUT"),
+          "'Result: FAILURE' or 'Result: TIMED OUT' not found in the following logs:\n#{logs}")
+      end
+    end
+
     def assert_deploy_success(result)
       assert_equal(true, result, "Deploy failed when it was expected to succeed.#{logs_message_if_captured}")
       logging_assertion do |logs|