Decide what to do with WEBEXT_* legacy telemetry histograms
Categories
(WebExtensions :: General, task)
Tracking
(Not tracked)
People
(Reporter: florian, Assigned: florian)
References
(Blocks 1 open bug)
Details
(Whiteboard: [addons-jira])
Attachments
(2 files)
Legacy telemetry histograms are currently being migrated to glean (bug 1950710), with a mirroring to legacy telemetry histograms using GIFFT.
Some WEBEXT_* histograms in toolkit/components/extensions/ExtensionTelemetry.sys.mjs already exist in Glean and the Glean metrics have been introduced more than a year ago in bug 1820158:
WEBEXT_BACKGROUND_PAGE_LOAD_MS
WEBEXT_BROWSERACTION_POPUP_OPEN_MS
WEBEXT_BROWSERACTION_POPUP_PRELOAD_RESULT_COUNT
WEBEXT_CONTENT_SCRIPT_INJECTION_MS
WEBEXT_EVENTPAGE_RUNNING_TIME_MS
WEBEXT_EVENTPAGE_IDLE_RESULT_COUNT
WEBEXT_EXTENSION_STARTUP_MS
WEBEXT_PAGEACTION_POPUP_OPEN_MS
WEBEXT_STORAGE_LOCAL_IDB_GET_MS
WEBEXT_STORAGE_LOCAL_IDB_SET_MS
Is this data needed in both Glean and legacy telemetry, or can we now remove the legacy telemetry histograms?
What should we do with the other histograms? I haven't looked at their implementation yet to see if they also record the same data in Glean, but here's the list of them:
WEBEXT_BACKGROUND_PAGE_LOAD_MS_BY_ADDONID
WEBEXT_BROWSERACTION_POPUP_OPEN_MS_BY_ADDONID
WEBEXT_BROWSERACTION_POPUP_PRELOAD_RESULT_COUNT_BY_ADDONID
WEBEXT_EVENTPAGE_RUNNING_TIME_MS_BY_ADDONID
WEBEXT_EVENTPAGE_IDLE_RESULT_COUNT_BY_ADDONID
WEBEXT_CONTENT_SCRIPT_INJECTION_MS_BY_ADDONID
WEBEXT_EXTENSION_STARTUP_MS_BY_ADDONID
WEBEXT_PAGEACTION_POPUP_OPEN_MS_BY_ADDONID
WEBEXT_STORAGE_LOCAL_IDB_GET_MS_BY_ADDONID
WEBEXT_STORAGE_LOCAL_IDB_SET_MS_BY_ADDONID
WEBEXT_STORAGE_LOCAL_IDB_MIGRATE_RESULT_COUNT
Assignee | ||
Comment 1•6 days ago
|
||
(In reply to Florian Quèze [:florian] from comment #0)
What should we do with the other histograms? I haven't looked at their implementation yet to see if they also record the same data in Glean, but here's the list of them:
WEBEXT_BACKGROUND_PAGE_LOAD_MS_BY_ADDONID
WEBEXT_BROWSERACTION_POPUP_OPEN_MS_BY_ADDONID
WEBEXT_BROWSERACTION_POPUP_PRELOAD_RESULT_COUNT_BY_ADDONID
WEBEXT_EVENTPAGE_RUNNING_TIME_MS_BY_ADDONID
WEBEXT_EVENTPAGE_IDLE_RESULT_COUNT_BY_ADDONID
WEBEXT_CONTENT_SCRIPT_INJECTION_MS_BY_ADDONID
WEBEXT_EXTENSION_STARTUP_MS_BY_ADDONID
WEBEXT_PAGEACTION_POPUP_OPEN_MS_BY_ADDONID
WEBEXT_STORAGE_LOCAL_IDB_GET_MS_BY_ADDONID
WEBEXT_STORAGE_LOCAL_IDB_SET_MS_BY_ADDONID
Actually, these are implemented in ExtensionTelemetry.sys.mjs too (search for "_BY_ADDONID").
WEBEXT_STORAGE_LOCAL_IDB_MIGRATE_RESULT_COUNT
The code records a Glean event in addition to the legacy telemetry histogram: https://searchfox.org/mozilla-central/rev/917fb41190af26c3536ecc883212fc1e6710aa51/toolkit/components/extensions/ExtensionStorageIDB.sys.mjs#116,136
Comment 2•6 days ago
|
||
Hi Florian,
yes the only ones still missing from Glean telemetry should be the subset of metrics collected through ExtensionTelemetry.sys.mjs that are keyed by addon id, at the time we filed a jira spike (WEBEXT-1541) to track touching base with a Glean team (to determine if there were already plans to add a labeled timing distribution, which we didn't have yet at the time) and determine if we wanted to collect of them in Glean or only a subset of them.
I just looked back to that jira spike and I see that in the meantime we have introduced supprot for a labeled_timing_distribution metric in Glean (and it looks we also support a labeled counter metric too) and so we should be able to port these to Glean too now.
Is this data needed in both Glean and legacy telemetry, or can we now remove the legacy telemetry histograms?
Based on the notes we collected last December while going through all our telemetry probes, there isn't any dashboard using these histograms data that we depend on (most of them have been used during investigations, the ones keyed by addon id were initially added specifically for a derived dataset, which it is not used nor maintained anymore: https://github.com/mozilla/addons_daily/) and so it should be ok to only look to the Glean data in the future.
I'll touch base with William asap to double-check that "not mirroring these histograms into legacy telemetry" does sound reasonable also from William's perspective.
Assignee | ||
Comment 3•5 days ago
|
||
Updated•5 days ago
|
Assignee | ||
Comment 4•5 days ago
|
||
Assignee | ||
Comment 5•5 days ago
|
||
(In reply to Luca Greco [:rpl] [:luca] [:lgreco] from comment #2)
Thanks!
I'll touch base with William asap to double-check that "not mirroring these histograms into legacy telemetry" does sound reasonable also from William's perspective.
Ok, I put up patches that migrate everything to Glean using legacy telemetry mirroring, except for the 2 histograms where a Glean equivalent is not supported yet. We can easily follow-up later to remove the legacy telemetry histograms.
Updated•5 days ago
|
Updated•5 days ago
|
Description
•