Taking a look at one problematic client on Aurora leads to a broad examination of the types of hosts that are sending us this data and some seriously-speculative conclusions.
This notebook transforms pings from the Pulse testpilot test to a parquet dataset. Docs at https://github.com/mozilla/pulse/blob/master/docs/metrics.md
This notebook transforms pings from the SnoozeTabs testpilot test to a parquet dataset. Docs at https://github.com/bwinton/SnoozeTabs/blob/master/docs/metrics.md
This job basically just takes core pings and puts them in parquet format.
This notebook maps Fennec saved_session pings to some useful information about clients. This is a 1:1 mapping.
This job takes the Fennec saved session pings and maps them to just client, submissionDate, activeAddons, and persona.
This job takes the Fennec saved session pings and transforms them, where there could be multiple events per ping.
The longitudinal, main_summary, and cross_sectional datasets can yield misleading Linux user counts over time