Now that the remediation has made it to release, how's it look? ### "late" client ids seem to have the same partial improvement as on other channels As we saw on Nightly and Beta, [the number of "late" client_ids diminished by the Remediation is... not absolute](https://sql.telemetry.mozilla.org/queries/79801/source#198169). My eyeballs suggest that the decrease is significant (statistically if in no other way), but it clearly hasn't had the effect I desired. ### New Profiles measured by clients sending us pings for the first time appear inflated by around 20% The larger-in-magnitude effect appears to be [how many client_id creations the Remediation prevents](https://sql.telemetry.mozilla.org/queries/79801/source#198167). That's on the order of 20%, and does indeed result in a decrease in new profile counts [as measured by "main" pings with `profile_subsession_counter = 1`](https://sql.telemetry.mozilla.org/queries/79803/source#198173). As measured by "new-profile" pings, [new profile ping counts and client counts remain unchanged](https://sql.telemetry.mozilla.org/queries/79804/source#198175). This appears to be visible on the New Profiles dashboard as a progressive decline in new profiles starting April 20: https://mozilla.cloud.looker.com/dashboards-next/53?Date=after%202021%2F01%2F01&Channel=&Activity%20Segment=&OS=&Attributed%20(Yes%20%2F%20No)=&Country%20Name= ### So how do we fix the rest of the "late" client_ids, hm? I'm really not sure. I haven't yet looked at the added instrumentation which reached release on 88 to see if there's anything obvious hidden in there, but I'm not hopeful it holds the answer. It appears as though the answer is "Sometimes we're able to save the session state file but not the Prefs, for no obvious reason", which doesn't lead itself for particularly strong ideas for fixes. I think the next step for me is to go hunting for ideas in the Release data. Maybe there's something that only becomes obvious at release volumes. If nothing comes up, the task will be to as exhaustively as possible document the behaviour of these in [the literature](https://docs.telemetry.mozilla.org) and continue wishing for a speedy migration of Firefox Desktop collections to Glean where this problem appears to be minimized.
Bug 1700188 Comment 3 Edit History
Note: The actual edited comment in the bug view page will always show the original commenter’s name and original timestamp.
Now that the remediation has made it to release, how's it look? ### "late" client ids seem to have the same partial improvement as on other channels As we saw on Nightly and Beta, [the number of "late" client_ids diminished by the Remediation is... not absolute](https://sql.telemetry.mozilla.org/queries/79801/source#198169). My eyeballs suggest that the decrease is significant (statistically if in no other way), but it clearly hasn't had the effect I desired. ### New Profiles measured by clients sending us pings for the first time appear inflated by around 20% The larger-in-magnitude effect appears to be [how many client_id creations the Remediation prevents](https://sql.telemetry.mozilla.org/queries/79801/source#198167). That's on the order of 20%, and does indeed result in a decrease in new profile counts [as measured by "main" pings with `profile_subsession_counter = 1`](https://sql.telemetry.mozilla.org/queries/79803/source#198173). As measured by "new-profile" pings, [new profile ping counts and client counts remain unchanged](https://sql.telemetry.mozilla.org/queries/79804/source#198175). This appears to be visible on the New Profiles dashboard as a progressive decline in new profiles starting April 20: https://mozilla.cloud.looker.com/dashboards-next/53?Date=after%202021%2F01%2F01&Channel=&Activity%20Segment=&OS=&Attributed%20(Yes%20%2F%20No)=&Country%20Name= ### So how do we fix the rest of the "late" client_ids, hm? I'm really not sure. I haven't yet looked at the added instrumentation which reached release on 88 to see if there's anything obvious hidden in there, but I'm not hopeful it holds the answer. It appears as though the answer is "Sometimes we're able to save the session state file but not the Prefs, for no obvious reason", which doesn't lend itself to particularly strong ideas for fixes. I think the next step for me is to go hunting for ideas in the Release data. Maybe there's something that only becomes obvious at release volumes. If nothing comes up, the task will be to as exhaustively as possible document the behaviour of these in [the literature](https://docs.telemetry.mozilla.org) and continue wishing for a speedy migration of Firefox Desktop collections to Glean where this problem appears to be minimized.