Increased error rate for metrics.search_count and browser.search.in_content metrics due to invalid label

[badboy]

We see a spike in errors for the metrics.search_count metric (Our monitoring dashboard, see the "Invalid Label" graph on the top right).

Channel: release.
The relevant code:

fenix/app/src/main/java/org/mozilla/fenix/components/metrics/MetricsUtils.kt

Line 48 in 01be732

Metrics.searchCount["$identifier.$source"].add()

Labels are restricted to lowercase alphanumerics, underscores, dashes and dots (see also our docs).
Maybe the identifier or source contains other characters? or the identifier is not always lowercased?

┆Issue is synchronized with this Jira Task

[cpeterso]

Looks like we're still seeing "Invalid Label" errors for the metrics.search_count and browser.search.in_content metrics, about 400 and 300 errors per day, respectively:

https://mozilla.cloud.looker.com/explore/fenix/metrics?qid=y2mXoHm29rnZSzSbpnfiPi&origin_space=746&toggle=vis

[Mugurell]

The source can only take some values and it is always lowercased:

fenix/app/src/main/java/org/mozilla/fenix/components/metrics/MetricsUtils.kt

Lines 30 to 46 in f356df7

	enum class Source {
	ACTION, SHORTCUT, SUGGESTION, TOPSITE, WIDGET, NONE
	}
	/**
	* Records the appropriate metric for performed searches.
	* @engine the engine used for searching.
	* @isDefault whether te engine is the default engine or not.
	* @searchAccessPoint the source of the search. Can be one of the values of [Source].
	*/
	fun recordSearchMetrics(
	engine: SearchEngine,
	isDefault: Boolean,
	searchAccessPoint: Source
	) {
	val identifier = if (engine.type == SearchEngine.Type.CUSTOM) "custom" else engine.id
	val source = searchAccessPoint.name.lowercase()

 

The identifier of bundled search engines is not lowercased before recording the telemetry and seems like we have some search engines which do use uppercase letters, eg: wikipedia-zh-CN.

Seems like lowercasing the search engine ids in Fenix/Focus before recording the telemetry is the best approach which would also ensure no newer search engines could cause this issue even if they are set to use uppercase letters.

[Mugurell]

Looking in Nightly it seems like the occurences went down a bit but might be too early to tell.
https://mozilla.cloud.looker.com/dashboards/382?Channel=mozdata.org%5E_mozilla%5E_fenix.metrics&Submission+Date=14+day+ago+for+14+day

[SoftVision-LorandJanos]

For the past 14 days:
On Nightly metrics.search_count was between 0 and 4.

On Nightly browser.search.in_content was between 0 and 7.

On Beta metrics.search_count was between 0 and 2.

On Beta browser.search.in_content was between 2 and 28.

Closing the ticket as verified.