Tweak Chaos Test Parameters (#357)

To reduce false positive failures

Author: kraftp

.github/workflows/chaos-test.yml

@@ -9,11 +9,7 @@ on:
 jobs:
   integration:
     runs-on: ubuntu-latest
-    strategy:
-      fail-fast: false
-      matrix:
-        python-version: ['3.13']
-
+    timeout-minutes: 60
     steps:
     - uses: actions/checkout@v4
       with:
@@ -23,13 +19,13 @@ jobs:
     - name: Setup PDM
       uses: pdm-project/setup-pdm@v4
       with:
-        python-version: ${{ matrix.python-version }}
+        python-version: '3.13'
         architecture: 'x64'
 
     - name: Install Dependencies
       run: pdm install
       working-directory: ./
 
     - name: Run Chaos Tests
-      run: pdm run pytest --timeout 600 chaos-tests
+      run: pdm run pytest --timeout 1200 chaos-tests
       working-directory: ./

chaos-tests/conftest.py

@@ -62,13 +62,13 @@ def __init__(self) -> None:
     def start(self) -> None:
         def _chaos_thread() -> None:
             while not self.stop_event.is_set():
-                wait_time = random.uniform(5, 20)
+                wait_time = random.uniform(5, 40)
                 if not self.stop_event.wait(wait_time):
                     print(
                         f"🐒 ChaosMonkey strikes after {wait_time:.2f} seconds! Restarting Postgres..."
                     )
                     stop_docker_pg()
-                    down_time = random.uniform(0, 5)
+                    down_time = random.uniform(0, 2)
                     time.sleep(down_time)
                     start_docker_pg()
 

chaos-tests/test_workflows.py

@@ -42,7 +42,7 @@ def test_recv(dbos: DBOS) -> None:
     def recv_workflow() -> Any:
         return DBOS.recv(topic, timeout_seconds=10)
 
-    num_workflows = 1000
+    num_workflows = 5000
 
     for i in range(num_workflows):
         handle = DBOS.start_workflow(recv_workflow)