* This patch adds the first version of the REST endpoint that
can generate a report for a workflow execution. Without any
query parameters, GET method of the endpoint returns a tree-like
structure that includes information about all execution objects
associated with the specified workflow execution. The root object
is the workflow execution itself, its children are task executions,
each task execution has either action executions or workflow
executions and so on. So the structure describes the entire set
of execution objects regardless of how deep there are from the
root object. This kind of data itself can be used to better
understand and visualise the process related with a parent workflow,
see what paths were taken while the workflow was running.
If additional parameters are provided in a query string, the
endpoint can give only a subset of the entire tree. Currently,
the filters are:
1. "errors_only" (boolean) to retain only all execution objects
with the error state, meaning that only all error paths are
present in the report. It is useful when we need to do a root
cause analysis of the workflow failure. False by default.
2. "max_depth" (integer) to limit how deep the algorithm can go
into nested workflows. If set to 0, only the root workflow
execution will be in the report. If set to 1, then the report
will have only the root workflow execution and its direct
children. And so on. If negative (by default) then no limit
is set.
Additionally, the report contains statistics about task executions
examined while the report was being generated, like the number of
tasks in the error state, number of tasks that successfully
finished and so on.
* Added all main tests for the endpoint. Note that despite the fact
that this test verifies a REST API endpoint, unlike the other API
tests it runs a Mistral engine to run workflows. This is done to
simplify the test implementation so that we don't have to mock
everything with huge data structures like we do in other API tests.
Possible changes that may be made based on the feedback:
* Statistics can contain not only number of tasks in certain states.
We can also add things like number of actions, depth of the tree,
number of nested workflows, average task/action/workflow execution
time etc.
* Additional query parameters to configure a generated report. For
example, "statistics_only" just to get a general information about
the workflow execution tree, not the tree itself. Another example
is "running_only" to retain only not finished workflow paths.
Implements blueprint: mistral-error-analysis
Change-Id: Id3e17821e04b7a1b84dfea5126d223d90ad8e3c2
ae2c5fdbbb