Monitoring

What a running cluster reports about itself, and how to read it.

What to watch sorts those readings by how fast a bad value causes damage, and says which of them the cluster repairs without you. That page is the place to start. The pages below describe where to take each reading.

MooseFS shows its state through two interfaces of its own. The GUI draws the cluster and marks values that need attention. The GUI is the only place with a picture of the whole installation on one screen. The CLI prints the same numbers as text, one section at a time. A script uses the CLI.

Two pages cover the readings themselves rather than the tools. Cluster health parameters explains what each number means, and what a value other than zero says about the data. Charts and statistics covers the history the cluster keeps on its own: how far back the history reaches and what each series counts.

For monitoring that keeps running with no browser tab open, Prometheus metrics covers the exporter and its output. Grafana covers turning those metrics into panels. Alerting covers the rules that send you notifications, so you don't have to keep checking the cluster. Logs covers what each component writes as it works, and which messages to recognize.

When a reading turns bad, go to Troubleshooting, which starts from the symptom rather than from the cause. The full list of values the exporter exposes is in Metrics. The messages the cluster raises on its own are in Alerts.