From the code below (the '//' lines are the proposal fix) there are 2 calls for the DB (one for list, secound for delete - that call for get), but there is a little different between them.
The second call is with ‘_secure_query’ so when it comes to delete the execution we are getting an exception that this ‘execution.id’ is not valid/available in the DB.
This happened since every exaction has its own project.id while the task (ExecutionExpirationPolicy) running without any special project.id.
And since we may have few different projects we need to set for each execution his project ID for each deletion (like you are doing in periodic.py when the MistralPeriodicTasks going over the triggers – lines 36-40) .
From expiration_policy.py:
==========================
with db_api.transaction():
# TODO(gpaz): In the future should use generic method with
# filters params and not specific method that filter by time.
for execution in db_api.get_expired_executions(exp_time):
try:
// Here we need to set the ctx with project ID so delete execution will not fail in ‘_secure_query’
LOG.debug('DELETE execution id : %s from date : %s '
'according to expiration policy',
execution.id,
execution.updated_at)
db_api.delete_workflow_execution(execution.id)
except Exception as e:
msg = "Failed to delete [execution_id=%s]\n %s" \
% (execution.id, traceback.format_exc(e))
LOG.warning(msg)
finally:
//Here we need to return the ctx to the origin values
Fix proposed to branch: master
Review: https:/