deb-mistral/mistral/workflow/reverse_workflow.py
Istvan Imre 3f5092d7f4 Publish/output in case of task/workflow failure
Currently it is not possible to provide any reasonable output in case of
a task or workflow failure. Implementing this would greatly simplify error
handling in workflows. This blueprint is a proposal to introduce two new
attributes, publish-on-error for tasks and output-on-error for workflows
for this purpose.

Implements: blueprint mistral-publish-on-error
Change-Id: Ib3a64971effb02390206dc6f993e772a51f8f237
2016-10-10 10:00:48 +02:00

195 lines
6.2 KiB
Python

# Copyright 2014 - Mirantis, Inc.
#
# 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.
import networkx as nx
from networkx.algorithms import traversal
from mistral import exceptions as exc
from mistral.workflow import base
from mistral.workflow import commands
from mistral.workflow import data_flow
from mistral.workflow import lookup_utils
from mistral.workflow import states
class ReverseWorkflowController(base.WorkflowController):
"""'Reverse workflow controller.
This controller implements the workflow pattern which is based on
dependencies between tasks, i.e. each task in a workflow graph
may be dependent on other tasks. To run this type of workflow
user must specify a task name that serves a target node in the
graph that the algorithm should come to by resolving all
dependencies.
For example, if there's a workflow consisting of two tasks
'A' and 'B' where 'A' depends on 'B' and if we specify a target
task name 'A' then the controller first will run task 'B' and then,
when a dependency of 'A' is resolved, will run task 'A'.
"""
__workflow_type__ = "reverse"
def _find_next_commands(self, task_ex):
"""Finds all tasks with resolved dependencies.
This method finds all tasks with resolved dependencies and
returns them in the form of workflow commands.
"""
cmds = super(ReverseWorkflowController, self)._find_next_commands(
task_ex
)
# TODO(rakhmerov): Adapt reverse workflow to non-locking model.
# 1. Task search must use task_ex parameter.
# 2. When a task has more than one dependency it's possible to
# get into 'phantom read' phenomena and create multiple instances
# of the same task. So 'unique_key' in conjunction with 'wait_flag'
# must be used to prevent this.
task_specs = self._find_task_specs_with_satisfied_dependencies()
return cmds + [
commands.RunTask(
self.wf_ex,
self.wf_spec,
t_s,
self.get_task_inbound_context(t_s)
)
for t_s in task_specs
]
def _get_target_task_specification(self):
task_name = self.wf_ex.params.get('task_name')
task_spec = self.wf_spec.get_tasks().get(task_name)
if not task_spec:
raise exc.WorkflowException(
'Invalid task name [wf_spec=%s, task_name=%s]' %
(self.wf_spec, task_name)
)
return task_spec
def _get_upstream_task_executions(self, task_spec):
t_specs = [
self.wf_spec.get_tasks()[t_name]
for t_name in self.wf_spec.get_task_requires(task_spec)
or []
]
return list(
filter(
lambda t_e: t_e.state == states.SUCCESS,
lookup_utils.find_task_executions_by_specs(
self.wf_ex.id,
t_specs
)
)
)
def evaluate_workflow_final_context(self):
task_execs = lookup_utils.find_task_executions_by_spec(
self.wf_ex.id,
self._get_target_task_specification()
)
# NOTE: For reverse workflow there can't be multiple
# executions for one task.
assert len(task_execs) <= 1
if len(task_execs) == 1:
return data_flow.evaluate_task_outbound_context(task_execs[0])
else:
return {}
def get_logical_task_state(self, task_ex):
# TODO(rakhmerov): Implement.
return task_ex.state, task_ex.state_info, 0
def is_error_handled_for(self, task_ex):
return task_ex.state != states.ERROR
def all_errors_handled(self):
task_execs = lookup_utils.find_error_task_executions(self.wf_ex.id)
return len(task_execs) == 0
def _find_task_specs_with_satisfied_dependencies(self):
"""Given a target task name finds tasks with no dependencies.
:return: Task specifications with no dependencies.
"""
tasks_spec = self.wf_spec.get_tasks()
graph = self._build_graph(tasks_spec)
# Unwind tasks from the target task
# and filter out tasks with dependencies.
return [
t_s for t_s in
traversal.dfs_postorder_nodes(
graph.reverse(),
self._get_target_task_specification()
)
if self._is_satisfied_task(t_s)
]
def _is_satisfied_task(self, task_spec):
if lookup_utils.find_task_executions_by_spec(
self.wf_ex.id, task_spec):
return False
if not self.wf_spec.get_task_requires(task_spec):
return True
success_t_names = set()
for t_ex in self.wf_ex.task_executions:
if t_ex.state == states.SUCCESS:
success_t_names.add(t_ex.name)
return not (
set(self.wf_spec.get_task_requires(task_spec)) - success_t_names
)
def _build_graph(self, tasks_spec):
graph = nx.DiGraph()
# Add graph nodes.
for t in tasks_spec:
graph.add_node(t)
# Add graph edges.
for t_spec in tasks_spec:
for dep_t_spec in self._get_dependency_tasks(tasks_spec, t_spec):
graph.add_edge(dep_t_spec, t_spec)
return graph
def _get_dependency_tasks(self, tasks_spec, task_spec):
dep_task_names = self.wf_spec.get_task_requires(task_spec)
if len(dep_task_names) == 0:
return []
dep_t_specs = set()
for t_spec in tasks_spec:
for t_name in dep_task_names:
if t_name == t_spec.get_name():
dep_t_specs.add(t_spec)
return dep_t_specs