Merge branch '232786-respect-stop-query-failures' into 'master'

Skip subsequent topology Prometheus queries if timeout occur

See merge request gitlab-org/gitlab!38293

changelogs/unreleased/232786-respect-stop-query-failures.yml

+---
+title: Skip subsequent topology Prometheus queries if timeout occur
+merge_request: 38293
+author:
+type: performance

lib/gitlab/usage_data/topology.rb

@@ -17,6 +17,9 @@ module Gitlab
         'registry' => 'registry'
       }.freeze
 
+      # If these errors occur, all subsequent queries are likely to fail for the same error
+      TIMEOUT_ERRORS = [Errno::ETIMEDOUT, Net::OpenTimeout, Net::ReadTimeout].freeze
+
       CollectionFailure = Struct.new(:query, :error) do
         def to_h
           { query => error }
@@ -158,6 +161,11 @@ module Gitlab
       end
 
       def query_safely(query, query_name, fallback:)
+        if timeout_error_exists?
+          @failures << CollectionFailure.new(query_name, 'timeout_cancellation')
+          return fallback
+        end
+
         result = yield query
 
         return result if result.present?
@@ -169,6 +177,14 @@ module Gitlab
         fallback
       end
 
+      def timeout_error_exists?
+        timeout_error_names = TIMEOUT_ERRORS.map(&:to_s).to_set
+
+        @failures.any? do |failure|
+          timeout_error_names.include?(failure.error)
+        end
+      end
+
       def topology_node_services(instance, all_process_counts, all_process_memory, all_server_types)
         # returns all node service data grouped by service name as the key
         instance_service_data =

spec/lib/gitlab/usage_data/topology_spec.rb

@@ -402,7 +402,8 @@ RSpec.describe Gitlab::UsageData::Topology do
       end
 
       context 'and an error is raised when querying Prometheus' do
-        it 'returns empty result with failures' do
+        context 'without timeout failures' do
+          it 'returns empty result and executes subsequent queries as usual' do
             expect_prometheus_api_to receive(:query)
               .at_least(:once)
               .and_raise(Gitlab::PrometheusClient::ConnectionError)
@@ -426,6 +427,38 @@ RSpec.describe Gitlab::UsageData::Topology do
             })
           end
         end
+
+        context 'with timeout failures' do
+          where(:exception) do
+            described_class::TIMEOUT_ERRORS
+          end
+
+          with_them do
+            it 'returns empty result and cancelled subsequent queries' do
+              expect_prometheus_api_to receive(:query)
+                .and_raise(exception)
+
+              expect(subject[:topology]).to eq({
+                duration_s: 0,
+                failures: [
+                  { 'app_requests' => exception.to_s },
+                  { 'node_memory' => 'timeout_cancellation' },
+                  { 'node_memory_utilization' => 'timeout_cancellation' },
+                  { 'node_cpus' => 'timeout_cancellation' },
+                  { 'node_cpu_utilization' => 'timeout_cancellation' },
+                  { 'node_uname_info' => 'timeout_cancellation' },
+                  { 'service_rss' => 'timeout_cancellation' },
+                  { 'service_uss' => 'timeout_cancellation' },
+                  { 'service_pss' => 'timeout_cancellation' },
+                  { 'service_process_count' => 'timeout_cancellation' },
+                  { 'service_workers' => 'timeout_cancellation' }
+                ],
+                nodes: []
+              })
+            end
+          end
+        end
+      end
     end
 
     context 'when embedded Prometheus server is disabled' do