Crashlytics integration akin to Bugsnag

kitallis commented 2 months ago

Context

This is the current set of health metics data stored by Tramline:

# Table name: release_health_metrics
#  daily_users                :bigint
#  daily_users_with_errors    :bigint
#  errors_count               :bigint
#  new_errors_count           :bigint
#  sessions                   :bigint
#  sessions_in_last_day       :bigint
#  sessions_with_errors       :bigint
#  total_sessions_in_last_day :bigint

We will need to transform the data from Crashlytics into a similar form and store it in Tramline.

Figure out the query required to get this data. Make sure to filter by the correct version code and version name.

Fetch frequency can be lower than Bugsnag (every 30 minutes?).

Once the data is fetched and stored, the rest of the health rule application logic remains the same. So, all that is needed from this issue to fetch, transform and store the release health metrics data.

Crashlytics is special because there's no direct API to get this data. We expect Firebase to be connected to the user's account, and we would make calls to the BQ dataset instead. Link to sample BQ data exported to a spreadsheet.

Requirements

Write an API layer and integration layer for Crashlytics
Ensure the interfaces adhere to the common functions, see Bugsnag integration

Acceptance Criteria

UI change required only in connecting the integration, the rest should work seamlessly.
Hook up the APIs and integration layer
Unit tests & some API-mocked fixture-driven tests
Ensure things work as closely as Bugsnag. Test manually properly.
Find an estimate of the costs associated with each query. This will help us tweak polling rate and help communicate the unprecedented costs (if any) to our users.

Additional Notes

Here's a sample BQ query that can be used as a starting point for all the other queries to be written:

SELECT
  COUNT(DISTINCT event_id) AS number_of_crashes,
  FORMAT_TIMESTAMP("%F", event_timestamp) AS date_of_crashes
FROM
 `decoded-theme-355014.firebase_crashlytics.com_tramline_ueno_ANDROID_REALTIME`
GROUP BY
  date_of_crashes
ORDER BY
  date_of_crashes DESC
LIMIT 30;

gitstart-connector[bot] commented 1 month ago

Issue synced into GitStart. Please review, edit and hand it off to GitStart by following this link.

gitstart-connector[bot] commented 3 weeks ago

This issue has been handed off to GitStart: https://clients.gitstart.com/tramline/11653/tickets/TRAB-645

kitallis commented 2 weeks ago

Some additional definitions, referenced from https://bugsnagapiv2.docs.apiary.io/#reference/projects/releases/view-a-release

errors_introduced_count: the new errors that have been introduced in this release
errors_seen_count: the total errors that have been seen in this release
sessions_count_in_last_24h: total number sessions for the release in the last 24h
total_sessions_count: total number of sessions for the release
unhandled_sessions_count: total number of sessions for the release that have had an exception/error
accumulative_daily_users_with_unhandled: the total number of users in the release with errors
accumulative_daily_users_seen: the total number of users in the release

tramlinehq / tramline