Kibana stats: less aggressive usage collection if failed

[afharo]

Metricbeat for Kibana collects usage stats once a day.

But, if the collection request fails (we've seen timeouts lately for requests taking longer than 10s default timeout), it will request the usage again on the next poll interval (10s later), overloading the server even more.

beats/metricbeat/module/kibana/stats/stats.go

Lines 166 to 173 in 03748b3

	content, err = m.statsHTTP.FetchContent()
	if err != nil {
	return err
	}
	if shouldCollectUsage {
	m.usageLastCollectedOn = now
	}

Can we set up a backoff mechanism that will try to skip the collection of the usage for a few rounds (maybe for another day?), prioritising metric-stats over usage-collection?

[elasticmachine]

Pinging @elastic/integrations-services (Team:Services)

[elasticmachine]

Pinging @elastic/stack-monitoring (Stack monitoring)

[afharo]

@andresrc thank you for amending the labels 😅
I've created a tentative PR in here: #20772

But I'm not a go developer so maybe there's a better way to implement it 😉

[andresrc]

Thank you very much @afharo !