summaryrefslogtreecommitdiffstats
path: root/toolkit/components/telemetry/docs/collection/custom-pings.rst
blob: 395a23aace958316480af92762f701894043c037 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
.. _submitting-customping:

=======================
Submitting custom pings
=======================

Custom pings can be submitted from JavaScript using:

.. code-block:: js

    TelemetryController.submitExternalPing(type, payload, options)

- ``type`` - a ``string`` that is the type of the ping, limited to ``/^[a-z0-9][a-z0-9-]+[a-z0-9]$/i``.
- ``payload`` - the actual payload data for the ping, has to be a JSON style object.
- ``options`` - optional, an object containing additional options:
   - ``addClientId``- whether to add the client id to the ping, defaults to ``false``
   - ``addEnvironment`` - whether to add the environment data to the ping, defaults to ``false``
   - ``overrideEnvironment`` - a JSON style object that overrides the environment data

``TelemetryController`` will assemble a ping with the passed payload and the specified options.
That ping will be archived locally for use with Shield and inspection in ``about:telemetry``.
If preferences allow the upload of Telemetry pings, the ping will be uploaded at the next opportunity (this is subject to throttling, retry-on-failure, etc.).

.. important::

    Every new or changed data collection in Firefox needs a `data collection review <https://wiki.mozilla.org/Firefox/Data_Collection>`__ from a Data Steward.

Submission constraints
----------------------

When submitting pings on shutdown, they should not be submitted after Telemetry shutdown.
Pings should be submitted at the latest within:

- the `observer notification <https://developer.mozilla.org/docs/Observer_Notifications#Application_shutdown>`_ ``"profile-before-change"``
- the :ref:`AsyncShutdown phase <AsyncShutdown_phases>` ``sendTelemetry``

There are other constraints that can lead to a ping submission getting dropped:

- invalid ping type strings.
- invalid payload types: E.g. strings instead of objects.
- oversized payloads: We currently only drop pings >1MB, but targeting sizes of <=10KB is recommended.

Tools
=====

Helpful tools for designing new pings include:

- `gzipServer <https://github.com/mozilla/gzipServer>`_ - a Python script that can run locally and receives and saves Telemetry pings. Making Firefox send to it allows inspecting outgoing pings easily.
- ``about:telemetry`` - allows inspecting submitted pings from the local archive, including all custom ones.

Designing custom pings
======================

In general, creating a new custom ping means you don't benefit automatically from the existing tooling. Further work is needed to make data show up in re:dash or other analysis tools.

In addition to the `data collection review <https://wiki.mozilla.org/Firefox/Data_Collection>`__, questions to guide a new ping design are:

- Submission interval & triggers:
   - What events trigger ping submission?
   - What interval is the ping submitted in?
   - Is there a throttling mechanism?
   - What is the desired latency? (submitting "at least daily" still leads to certain latency tails)
   - Are pings submitted on a clock schedule? Or based on "time since session start", "time since last ping" etc.? (I.e. will we get sharp spikes in submission volume?)
- Size and volume:
   - What’s the size of the submitted payload?
   - What's the full ping size including metadata in the pipeline?
   - What’s the target population?
   - What's the overall estimated volume?
- Dataset:
   - Is it opt-out?
   - Does it need to be opt-out?
   - Does it need to be in a separate ping? (why can’t the data live in probes?)
- Privacy:
   - Is there risk to leak PII?
   - How is that risk mitigated?
- Data contents:
   - Does the submitted data answer the posed product questions?
   - Does the shape of the data allow to answer the questions efficiently?
   - Is the data limited to what's needed to answer the questions?
   - Does the data use common formats? (i.e. can we re-use tooling or analysis know-how)