1175115 - Spark API for telemetry needs proper test coverage

Status: RESOLVED FIXED

(Cloud Services Graveyard :: Metrics: Pipeline, defect, P2)

Description

[Roberto Agostino Vitillo (:rvitillo)]

The telemetry API can be found at [1]. As the API is slowly growing, we should write a proper test-suite for it.

[1] https://github.com/vitillo/python_moztelemetry/

Comment 1

[Sam Penrose]

Writing tests is my favorite way to learn a code base, so I grabbed this. Feel free to reassign if it makes sense.

Comment 2

[Roberto Agostino Vitillo (:rvitillo)]

I am increasing the priority as the data validation depends on it.

Comment 3

[Mark Reid [:mreid]]

Any progress on this?  I'm looking at adding support for derived streams, and it'd be nice to be able to test things before sending a possibly-broken PR :)

Comment 4

[Sam Penrose]

Sadly no.

(In reply to Mark Reid [:mreid] from comment #3)
> Any progress on this?  I'm looking at adding support for derived streams,
> and it'd be nice to be able to test things before sending a possibly-broken
> PR :)

Comment 5

[Mark Reid [:mreid]]

My approach to testing this was:
1. Spin up a single-instance Spark cluster from telemetry-dash.mozilla.org
2. Copy over the modified files using scp
3. Restart the Spark kernel
4. Run the "Telemetry Hello World" notebook and make sure the results look OK.

Comment 6

[Mauro Doglio [:mdoglio]]

:gfritzsche the python_moztelemetry library has a test suite already running in CI so

Comment 7

[Mauro Doglio [:mdoglio]]

:gfritzsche the python_moztelemetry library has a test suite already running in CI so you should be able to add as many tests as you want to it. I fixed the bug title to reflect that. Can we remove bug 1264984 from the block list?

Comment 8

[Mauro Doglio [:mdoglio]]

Added some refactoring work to the block list because it would make it much easier to write unittest.

Comment 9

[Mauro Doglio [:mdoglio]]

Now that bug 1282803 is fixed, the test coverage is 35% (was more or less 0% before). Another coverage bump up should come from solving bug 1288154.

Comment 10

[Mauro Doglio [:mdoglio]]

The overall coverage is now 54%. As shown here [1], the modules used by get_pings have a very high coverage (93%, 100%) which is what matters the most.

[1]https://coveralls.io/github/mozilla/python_moztelemetry?branch=master