Task Run API tests and Library

[gshif]

• Define and create tests for Task Run API
• Define and create wrapper function for Task Run API
• If any, helper functions.
• Reference latest swagger doc

[gshif]

Create wrapper functions for get all task runs and get task run endpoints.

https://github.com/influxdata/influxdb/blob/master/http/swagger.yml#L4121-L4191

• [ ] Create get_all_task_runs function (wraps /tasks/{task id}/runs endpoint):
  • function will take following arguments: task_id, token, after, limit, after_time and before_time.
  • function will return the list of task runs, where each run is a dictionary
• [ ] Create get_task_run function(wraps /tasks/{task id}/runs/{run id} endpoint):
  • function will take following arguments: task_id, token and task_run_id.
  • function will return a dictionary of key/values from a response object.

---

• [ ] Create helper function(s):
  • [ ] function to get count of all task runs for a specific task.

---

Challenges :

• Currently every task request is aligned to the hour/minute/sec, e.g if task was created at 2019-04-09T22:41:03Z and has every set to 15s, one would expect first run to happen at 2019-04-09T22:41:18Z, but it actually is set to 019-04-09T22:41:15Z

Tests

• Will be using create_org_and_bucket fixture.

• [x] All of the successful task runs within/between TIME1:TIME2 should be present. Verify count.

  • It will be hard to verify precise count of run tasks due to:
    • Current tasks implementation. With every option task run might start between task creation time + every. We should probably do the following:
      • Create a task.
      • Get a task start time by calling get a task endpoint since create task does not return the task creation time (https://github.com/influxdata/idpe/issues/3166)
      • The first task run should happen between task creation time and task creation time + every(it is possible that task run won't be reported within these times since task run is being written to kafka, to storage --> system bucket of the organization this task belongs to, so there could be some latency.)
      • Once we know the time of the first task run, we can add every * 4 (if we want to get 4 task runs) and then get task runs
  • Create a task that reads from a bucket and writes to a different(same?) bucket. (Will need to create a bucket to write to)
  • Task execution time should be less than task scheduled time, e.g. task is scheduled to run every 30 seconds and time execution time is less than 30 seconds.
  • Can write the same point from bucketA to a bucketB, but have a different timestamp for the point written to bucketB, i.e. bucketB would have multiple identical points (measurement, tag, field), but different timestamp. At the end of the test run a count(*) query to verify number of records.

[gshif]

• [x] Get all of the task runs with default query param limit = 20
  • create a task that runs every 15 sec (see previous test case)
  • wait for task to run more than 20 times.
  • get all of the task runs. Should only return the first 20

[gshif]

https://github.com/influxdata/idpe/pull/3329 (the above two tests)