We should extend the v2/v4 comparison in this script: https://gist.github.com/georgf/1b0831a6b81b6c9fe240 ... with totalTime & browser defaults.
Any other fields?
Component: Telemetry → Metrics: Pipeline
Product: Toolkit → Cloud Services
Version: Trunk → unspecified
Updated the script: * added comparisons of v2 totalTime to v4 totalTime, sessionLength, subsessionLength * if on a "newish" profile on 42+, don't match up v2 or v4 dates if they cover roughly the same historic dates, use all the data instead * show the cutoff date used for date matching (and when it wasn't used) This gives fine totalTime comparisons on my first run on a new profile, but looks rather broken on my profile with long history. Caveats: * sessionLength was introduced relatively recently (Nightly >= 2015-07-30) * when not on a newish profile on 42+, we have to try date matching, so certain discrepancies are expected
So we can approach this from two angles: * trying to reproduce big mismatches on a fresh profile on 42+ * for bigger mismatches in our clients, manually dig into the historic data to check this (potentially time consuming)
Mismatch data from Katie: https://docs.google.com/spreadsheets/d/1XCi_yYTb6RVO89lfnAAf1MTqBV7F2YXPXJYY-6fmDU8/edit#gid=0
(In reply to Georg Fritzsche [:gfritzsche] from comment #4) > Mismatch data from Katie: > https://docs.google.com/spreadsheets/d/1XCi_yYTb6RVO89lfnAAf1MTqBV7F2YXPXJYY- > 6fmDU8/edit#gid=0 The sesssion chaining problems there went away after waiting for 60sec after startup (i.e. after letting Telemetry load the session data from disk). The mismatches persisted.
Updated: * added the current browser default * consolidated the output in one tab * v2 history is now shown in table form too for manual inspection
Updated, mostly added the historical browser default validation.
This part seems to work fine, lets collect any suspicious data and file new bugs for investigations.
Status: NEW → RESOLVED
Last Resolved: 3 years ago
Resolution: --- → WORKSFORME
You need to log in before you can comment on or make changes to this bug.