What can classic monitoring like Nagios do that Prometheus deliberately doesn't, and what should you use instead?

Event monitoring: triggering an action because a specific line shows up in a log file. Prometheus only handles numeric metrics, so log analysis belongs in a central logging system.

A Nagios setup can watch a log and alert when, say, OutOfMemoryError appears. Prometheus has no notion of individual events or log lines; it samples numbers.

The Prometheus developers' answer: don't build a workaround into your monitoring. Analyse logs with central logging and automatic log analysis (for example the Elastic stack, Splunk or Loki), which large platforms run anyway. Right tool for the job: metrics for trends and thresholds, logs for individual events.

Go deeper:

• Grafana Loki documentation — a log system designed to sit next to Prometheus, using the same labels.
• Elasticsearch — Wikipedia — the search engine behind the Elastic stack for central logging.