turtlebot3: stop the debounce profile from hiding the goal-status faults

.github/workflows/ci.yml

@@ -94,6 +94,47 @@ jobs:
         working-directory: demos/turtlebot3_integration
         run: docker compose --profile ci down
 
+  # The debounce profile is a second configuration of the same demo, so it needs
+  # its own run: the default profile never loads the per-source threshold file,
+  # and a key that stops matching it fails silently by falling back to the
+  # global thresholds.
+  build-and-test-turtlebot-debounce:
+    runs-on: ubuntu-24.04
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+
+      # Static check, so it also covers the cpu and nvidia services that CI
+      # never starts. A service missing from the override runs the default
+      # profile while the caller believes it is debouncing.
+      - name: Check the debounce override covers every demo service
+        run: ./tests/check_debounce_overlay.sh
+
+      - name: Build and start turtlebot3 demo in debounce mode
+        working-directory: demos/turtlebot3_integration
+        run: |
+          docker compose --profile ci \
+            -f docker-compose.yml -f docker-compose.debounce.yml \
+            up -d --build turtlebot3-demo-ci
+
+      - name: Run debounce smoke tests
+        run: ./tests/smoke_test_debounce.sh
+
+      - name: Show container logs on failure
+        if: failure()
+        working-directory: demos/turtlebot3_integration
+        run: |
+          docker compose --profile ci \
+            -f docker-compose.yml -f docker-compose.debounce.yml \
+            logs turtlebot3-demo-ci --tail=200
+
+      - name: Teardown
+        if: always()
+        working-directory: demos/turtlebot3_integration
+        run: |
+          docker compose --profile ci \
+            -f docker-compose.yml -f docker-compose.debounce.yml down
+
   build-and-test-moveit:
     runs-on: ubuntu-24.04
     steps:

demos/turtlebot3_integration/config/entity_thresholds_debounce.yaml

@@ -0,0 +1,39 @@
+# Copyright 2026 bburda
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+# Per-source debounce for the debounce demo. Keys match a source_id by longest
+# prefix, so the entry below applies only to the goal-status reporter and the
+# global thresholds still cover the no-progress one.
+#
+# The prefix is anchored at the start of the source_id, which the detector builds
+# as its fully qualified node name plus a suffix. demo.launch.py puts the node in
+# the "bridge" namespace, so the name is /bridge/anomaly_detector and the leading
+# /bridge is part of the key. Renaming the node or its namespace in the launch
+# file breaks this match silently: the resolver falls back to the global
+# thresholds and the goal-status faults never leave PREFAILED.
+
+/bridge/anomaly_detector/goal_status:
+  # A goal aborting or being cancelled is a single status change, so exactly one
+  # FAILED arrives and exactly one PASSED on the next success. -1 confirms on
+  # that one event and 0 heals on it. Under the global -3 the counter stops at -1
+  # and both codes stay PREFAILED for good. They are still listed, because the
+  # default fault filter includes PREFAILED alongside CONFIRMED, but they never
+  # count as confirmed: confirmedDTC stays 0, a confirmed-only query does not
+  # return them, and nothing that keys off confirmation runs.
+  confirmation_threshold: -1
+  # These two repeat the global values rather than changing them. They are stated
+  # so this source keeps confirming and healing on a single event even if the
+  # global profile is retuned later; only confirmation_threshold differs today.
+  healing_enabled: true
+  healing_threshold: 0

demos/turtlebot3_integration/config/medkit_params.yaml

@@ -50,7 +50,7 @@ fault_manager:
     database_path: "/var/lib/ros2_medkit/faults.db"
 
     # Debounce configuration
-    confirmation_threshold: 0  # Immediate confirmation
+    confirmation_threshold: -1  # Immediate confirmation; 0 is rejected and falls back to this
     healing_enabled: false
     healing_threshold: 3
     auto_confirm_after_sec: 0.0

demos/turtlebot3_integration/config/medkit_params_debounce.yaml

@@ -1,8 +1,9 @@
 # ros2_medkit gateway configuration for TurtleBot3 demo
 # ARTICLE 3 - DEBOUNCE VERSION
-# Differences from default:
-#   confirmation_threshold: -3 (was 0) — requires 3 sustained FAILED events
-#   healing_enabled: true (was false) — auto-heal after PASSED events
+# Differences from the default profile in medkit_params.yaml:
+#   confirmation_threshold: -3, so three sustained FAILED events are needed
+#     before an operator sees the fault, instead of the default -1
+#   healing_enabled: true, so a fault clears itself once the condition is gone
 #
 # Node runs under /diagnostics namespace, so we need to match that here
 diagnostics:
@@ -38,11 +39,25 @@ fault_manager:
     database_path: "/var/lib/ros2_medkit/faults.db"
 
     # === DEBOUNCE CONFIGURATION (Article 3) ===
-    confirmation_threshold: -3   # Need 3 sustained FAILED events to confirm
-    healing_enabled: true        # Auto-heal when problem resolves
-    healing_threshold: 3         # Need 3 PASSED events to heal
+    # -3 filters the no-progress reporter, which repeats every 5 s while the
+    # robot is stuck: three reports, so roughly 15 s of it, before an operator
+    # sees anything. That is the contrast with the storm profile.
+    confirmation_threshold: -3
+    healing_enabled: true
+    # A confirmed fault sits at the confirmation threshold, because the counter is
+    # clamped to the band, so healing costs healing_threshold minus -3 PASSED
+    # events: three at 0, six at 3. The detector answers a recovery with a burst
+    # of three, which is what makes 0 the value that heals and 3 the value that
+    # would not.
+    healing_threshold: 0
     auto_confirm_after_sec: 0.0  # Disabled
 
+    # The goal-status faults fire once per status change, so the -3 above would
+    # hide them permanently. They report under their own source and take their
+    # own thresholds from here.
+    entity_thresholds:
+      config_file: "/root/demo_ws/src/turtlebot3_medkit_demo/config/entity_thresholds.yaml"
+
     # Snapshot configuration (freeze frames)
     snapshots:
       enabled: true

demos/turtlebot3_integration/docker-compose.debounce.yml

@@ -7,6 +7,13 @@
 #   DEBOUNCE (with filtering):
 #     docker compose --profile cpu -f docker-compose.yml -f docker-compose.debounce.yml up -d
 #
+#   DEBOUNCE in CI (headless, driven by tests/smoke_test_debounce.sh):
+#     docker compose --profile ci -f docker-compose.yml -f docker-compose.debounce.yml up -d
+#
+# Every service that can run the demo needs an entry here. A service left out
+# starts with the default profile while the caller believes it is debouncing,
+# and every assertion about thresholds then measures the wrong configuration.
+#
 # The override mounts the debounce config over the default one inside the container.
 # colcon build --symlink-install means the installed config points to the source,
 # so mounting over the source path works.
@@ -15,7 +22,14 @@ services:
   turtlebot3-demo:
     volumes:
       - ./config/medkit_params_debounce.yaml:/root/demo_ws/src/turtlebot3_medkit_demo/config/medkit_params.yaml:ro
+      - ./config/entity_thresholds_debounce.yaml:/root/demo_ws/src/turtlebot3_medkit_demo/config/entity_thresholds.yaml:ro
 
   turtlebot3-demo-nvidia:
     volumes:
       - ./config/medkit_params_debounce.yaml:/root/demo_ws/src/turtlebot3_medkit_demo/config/medkit_params.yaml:ro
+      - ./config/entity_thresholds_debounce.yaml:/root/demo_ws/src/turtlebot3_medkit_demo/config/entity_thresholds.yaml:ro
+
+  turtlebot3-demo-ci:
+    volumes:
+      - ./config/medkit_params_debounce.yaml:/root/demo_ws/src/turtlebot3_medkit_demo/config/medkit_params.yaml:ro
+      - ./config/entity_thresholds_debounce.yaml:/root/demo_ws/src/turtlebot3_medkit_demo/config/entity_thresholds.yaml:ro