Check your metric names is the Important line in Telegraf 1.40

So Telegraf Releases put v1.40.0 on the feed last night. Same row on telegraf-updates. We flagged it Important as a breaking change. The section that matters is literally titled Important Changes. Not the new plugins list.

Recency wants fritzbox_smarthome and zerobus. Fine. Those are the names people screenshot. The partner-break lines are the ones that say please check your metric names and queries. Cisco telemetry MDT now respects configured aliases, which might change the metric name for microburst and RIB data if you set aliases. Spurious RIB metrics get cleaned up. Table row numbering changes. Tags stop leaking into parent metrics. And power_on_hours in the smart input used to report duration values in seconds. Now it reports hours. That is a dashboard break dressed as a unit fix.

I keep almost filing Telegraf under "agent bump, ignore." Kind of the wrong habit when you sell next to observability stacks. New plugins are what your AE forwards. Metric renames and unit flips are what break a Grafana panel that still assumes the old name or the old seconds scale.

To the best of my understanding the Breaking Changes list is the other half of the same story. Deprecated plugins go away: aerospike, sflow, amon. Deprecated options leave kafka_consumer, openstack, procstat, and tail. If a customer still pins those, let's say, the release notes already told them. The Important flag caught the check-your-metrics cohort. Recency still wants the shiny plugin names. Fine. Recency is not the unit flip.

We pull it into a uniform entry shape at /sources/telegraf-releases. Same JSON as everything else. The talking point is check your metric names, then power_on_hours is now hours. The new plugins are the decoy.