estel/dify
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

Feat: Retry on node execution errors (#11871)

Browse Source 
Co-authored-by: Novice Lee <novicelee@NoviPro.local>
This commit is contained in:
Novice
2024-12-20 14:14:06 +08:00
committed by GitHub
parent f6247fe67c
commit 7abc7fa573
23 changed files with 736 additions and 188 deletions
Download Patch File Download Diff File Expand all files Collapse all files

137
api/services/workflow_service.py
View File

@@ -15,6 +15,7 @@ from core.workflow.nodes.base.entities import BaseNodeData
from core.workflow.nodes.base.node import BaseNode
from core.workflow.nodes.enums import ErrorStrategy
from core.workflow.nodes.event import RunCompletedEvent
from core.workflow.nodes.event.event import SingleStepRetryEvent
from core.workflow.nodes.node_mapping import LATEST_VERSION, NODE_TYPE_CLASSES_MAPPING
from core.workflow.workflow_entry import WorkflowEntry
from events.app_event import app_draft_workflow_was_synced, app_published_workflow_was_updated
@@ -220,56 +221,99 @@ class WorkflowService:
# run draft workflow node
start_at = time.perf_counter()
retries = 0
max_retries = 0
should_retry = True
retry_events = []
try:
node_instance, generator = WorkflowEntry.single_step_run(
workflow=draft_workflow,
node_id=node_id,
user_inputs=user_inputs,
user_id=account.id,
)
node_instance = cast(BaseNode[BaseNodeData], node_instance)
node_run_result: NodeRunResult | None = None
for event in generator:
if isinstance(event, RunCompletedEvent):
node_run_result = event.run_result
while retries <= max_retries and should_retry:
retry_start_at = time.perf_counter()
node_instance, generator = WorkflowEntry.single_step_run(
workflow=draft_workflow,
node_id=node_id,
user_inputs=user_inputs,
user_id=account.id,
)
node_instance = cast(BaseNode[BaseNodeData], node_instance)
max_retries = (
node_instance.node_data.retry_config.max_retries if node_instance.node_data.retry_config else 0
)
retry_interval = node_instance.node_data.retry_config.retry_interval_seconds
node_run_result: NodeRunResult | None = None
for event in generator:
if isinstance(event, RunCompletedEvent):
node_run_result = event.run_result
# sign output files
node_run_result.outputs = WorkflowEntry.handle_special_values(node_run_result.outputs)
break
# sign output files
node_run_result.outputs = WorkflowEntry.handle_special_values(node_run_result.outputs)
break
if not node_run_result:
raise ValueError("Node run failed with no run result")
# single step debug mode error handling return
if node_run_result.status == WorkflowNodeExecutionStatus.FAILED and node_instance.should_continue_on_error:
node_error_args = {
"status": WorkflowNodeExecutionStatus.EXCEPTION,
"error": node_run_result.error,
"inputs": node_run_result.inputs,
"metadata": {"error_strategy": node_instance.node_data.error_strategy},
}
if node_instance.node_data.error_strategy is ErrorStrategy.DEFAULT_VALUE:
node_run_result = NodeRunResult(
**node_error_args,
outputs={
**node_instance.node_data.default_value_dict,
"error_message": node_run_result.error,
"error_type": node_run_result.error_type,
},
)
else:
node_run_result = NodeRunResult(
**node_error_args,
outputs={
"error_message": node_run_result.error,
"error_type": node_run_result.error_type,
},
)
run_succeeded = node_run_result.status in (
WorkflowNodeExecutionStatus.SUCCEEDED,
WorkflowNodeExecutionStatus.EXCEPTION,
)
error = node_run_result.error if not run_succeeded else None
if not node_run_result:
raise ValueError("Node run failed with no run result")
# single step debug mode error handling return
if node_run_result.status == WorkflowNodeExecutionStatus.FAILED:
if (
retries == max_retries
and node_instance.node_type == NodeType.HTTP_REQUEST
and node_run_result.outputs
and not node_instance.should_continue_on_error
):
node_run_result.status = WorkflowNodeExecutionStatus.SUCCEEDED
should_retry = False
else:
if node_instance.should_retry:
node_run_result.status = WorkflowNodeExecutionStatus.RETRY
retries += 1
node_run_result.retry_index = retries
retry_events.append(
SingleStepRetryEvent(
inputs=WorkflowEntry.handle_special_values(node_run_result.inputs)
if node_run_result.inputs
else None,
error=node_run_result.error,
outputs=WorkflowEntry.handle_special_values(node_run_result.outputs)
if node_run_result.outputs
else None,
retry_index=node_run_result.retry_index,
elapsed_time=time.perf_counter() - retry_start_at,
execution_metadata=WorkflowEntry.handle_special_values(node_run_result.metadata)
if node_run_result.metadata
else None,
)
)
time.sleep(retry_interval)
else:
should_retry = False
if node_instance.should_continue_on_error:
node_error_args = {
"status": WorkflowNodeExecutionStatus.EXCEPTION,
"error": node_run_result.error,
"inputs": node_run_result.inputs,
"metadata": {"error_strategy": node_instance.node_data.error_strategy},
}
if node_instance.node_data.error_strategy is ErrorStrategy.DEFAULT_VALUE:
node_run_result = NodeRunResult(
**node_error_args,
outputs={
**node_instance.node_data.default_value_dict,
"error_message": node_run_result.error,
"error_type": node_run_result.error_type,
},
)
else:
node_run_result = NodeRunResult(
**node_error_args,
outputs={
"error_message": node_run_result.error,
"error_type": node_run_result.error_type,
},
)
run_succeeded = node_run_result.status in (
WorkflowNodeExecutionStatus.SUCCEEDED,
WorkflowNodeExecutionStatus.EXCEPTION,
)
error = node_run_result.error if not run_succeeded else None
except WorkflowNodeRunFailedError as e:
node_instance = e.node_instance
run_succeeded = False
@@ -318,6 +362,7 @@ class WorkflowService:
db.session.add(workflow_node_execution)
db.session.commit()
workflow_node_execution.retry_events = retry_events
return workflow_node_execution