[Node] update update wf retry details

Author: GuillaumeDSM

packages/node/octobot_node/constants.py

@@ -13,6 +13,8 @@
 #
 #  You should have received a copy of the GNU General Public
 #  License along with OctoBot. If not, see <https://www.gnu.org/licenses/>.
+import os
+
 try:
     import octobot.constants as octobot_constants
     BASE_LOGS_FOLDER = octobot_constants.LOGS_FOLDER
@@ -22,6 +24,11 @@
 AUTOMATION_LOGS_FOLDER = f"{BASE_LOGS_FOLDER}/automations"
 PARENT_WORKFLOW_ID_LENGTH = 36 # length of a UUID4
 
+# default to 10 retry after 1, 2, 4, 8, 16, ... 1024 seconds (total of 2047 seconds)
+AUTOMATION_WORKFLOW_RETRY_INTERVAL_SECONDS = float(os.getenv("AUTOMATION_WORKFLOW_RETRY_INTERVAL_SECONDS", 1.0))
+AUTOMATION_WORKFLOW_MAX_ITERATION_RETRIES = int(os.getenv("AUTOMATION_WORKFLOW_MAX_ITERATION_RETRIES", 10))
+AUTOMATION_WORKFLOW_BACKOFF_RATE = float(os.getenv("AUTOMATION_WORKFLOW_BACKOFF_RATE", 2))
+
 TASKS_ENCRYPTION_ENV_VARS = [
     "TASKS_SERVER_RSA_PRIVATE_KEY",
     "TASKS_SERVER_ECDSA_PRIVATE_KEY",

packages/node/octobot_node/scheduler/workflows/automation_workflow.py

@@ -17,6 +17,7 @@
 import time
 import typing
 import dbos
+import os
 
 import octobot_commons.logging
 
@@ -27,6 +28,7 @@
 import octobot_node.models
 import octobot_node.scheduler.octobot_flow_client as octobot_flow_client
 import octobot_node.scheduler.task_context
+import octobot_node.constants as constants
 import octobot_node.scheduler.workflows.params as params
 import octobot_node.errors as errors
 
@@ -36,11 +38,6 @@
 from octobot_node.scheduler import SCHEDULER  # avoid circular import
 
 
-INTERVAL_SECONDS = 1.0
-MAX_ITERATION_RETRIES = 3
-BACKOFF_RATE = 2.0
-
-
 @SCHEDULER.INSTANCE.dbos_class()
 class AutomationWorkflow:
     # Always use dict as input to parse minimizable dataclasses and facilitate data format updates
@@ -108,9 +105,9 @@ async def execute_automation(inputs: dict) -> typing.Optional[str]:
     @SCHEDULER.INSTANCE.step(
         name="execute_iteration",
         retries_allowed=True,
-        interval_seconds = INTERVAL_SECONDS,
-        max_attempts=MAX_ITERATION_RETRIES,
-        backoff_rate=BACKOFF_RATE,
+        interval_seconds = constants.AUTOMATION_WORKFLOW_RETRY_INTERVAL_SECONDS,
+        max_attempts=constants.AUTOMATION_WORKFLOW_MAX_ITERATION_RETRIES,
+        backoff_rate=constants.AUTOMATION_WORKFLOW_BACKOFF_RATE,
     )
     async def execute_iteration(inputs: dict, actions_update: typing.Optional[dict]) -> dict:
         """

packages/node/tests/scheduler/workflows/test_automation_workflow.py

@@ -29,6 +29,7 @@
 
 import octobot_copy.entities
 import octobot_node.config
+import octobot_node.constants
 import octobot_node.enums
 import octobot_node.scheduler
 import octobot_node.scheduler.workflows
@@ -410,7 +411,7 @@ async def test_execute_automation(
                 assert parsed_output.error == expected_error_status
                 assert (
                     run_mock.await_count
-                    == octobot_node.scheduler.workflows.automation_workflow.MAX_ITERATION_RETRIES
+                    == octobot_node.constants.AUTOMATION_WORKFLOW_MAX_ITERATION_RETRIES
                 )
                 mock_logger.exception.assert_called_once()
                 mock_process.assert_not_called()
@@ -1033,12 +1034,12 @@ async def test_execute_automation_execute_iteration_retries_octobot_actions_job_
         temp_dbos_scheduler,
     ):
         """
-        DBOS execute_iteration is configured with max_attempts=MAX_ITERATION_RETRIES.
+        DBOS execute_iteration is configured with max_attempts=AUTOMATION_WORKFLOW_MAX_ITERATION_RETRIES.
         When OctoBotActionsJob.run() fails on early attempts then succeeds, the step should
         retry and eventually complete without failing the workflow. A processed action may still
         report an error_status; that value is copied to AutomationWorkflowOutput.error on completion.
         """
-        max_attempts = octobot_node.scheduler.workflows.automation_workflow.MAX_ITERATION_RETRIES
+        max_attempts = octobot_node.constants.AUTOMATION_WORKFLOW_MAX_ITERATION_RETRIES
         task = octobot_node.models.Task(
             name="retry_policy_test",
             content="{}",
@@ -1120,8 +1121,8 @@ async def test_execute_automation_execute_iteration_exhausts_retries_when_octobo
         import_automation_workflow,
         temp_dbos_scheduler,
     ):
-        """After MAX_ITERATION_RETRIES failed OctoBotActionsJob.run() calls, the step must stop retrying."""
-        max_attempts = octobot_node.scheduler.workflows.automation_workflow.MAX_ITERATION_RETRIES
+        """After AUTOMATION_WORKFLOW_MAX_ITERATION_RETRIES failed OctoBotActionsJob.run() calls, the step must stop retrying."""
+        max_attempts = octobot_node.constants.AUTOMATION_WORKFLOW_MAX_ITERATION_RETRIES
         task = octobot_node.models.Task(
             name="retry_exhausted_test",
             content="{}",