Monitor

Pigsty ships with a complete observability stack — it's all there once installed. Default endpoints in single-node mode:

Component	URL	Default account
Grafana	http://<IP>:3000	admin / pigsty
Prometheus	http://<IP>:9090	—
AlertManager	http://<IP>:9093	—
Loki (logs)	Embedded as a Grafana data source	—

---

Grafana dashboards

Log in and go to Home → Dashboards, organised by prefix:

• PGSQL — Cluster / Instance / Database / Query / Session / Replication
• PGCAT — table / index / extension details via the system catalog
• PGLOG — slow-query log analysis
• INFRA — Prometheus, Grafana, Loki themselves
• NODE — CPU / memory / disk / network
• REDIS / MINIO / ETCD — dashboards for each module

Dashboards to start with:

• PGSQL Overview — whole-cluster health at a glance
• PGSQL Cluster — replica lag, TPS, connection count
• PGSQL Instance — deep per-instance metrics
• PGSQL Query — slow queries via pg_stat_statements
• PGSQL Alert — currently active alerts

---

Metric sources

Source	Description
pg_exporter	600+ PostgreSQL metrics (custom queries included)
node_exporter	Host metrics
patroni	HA state
haproxy	Routing / health / connections
pgbouncer_exporter	Pool metrics

See pg_exporter module.

---

Alerts

Alert rules are preshipped in Prometheus, living at /etc/prometheus/rules/*.yml on each node.

Common default alerts:

• PostgreSQL primary unreachable
• Replication lag exceeds threshold
• Connection count nearing the limit
• Disk usage > 85%
• pgBouncer connection backlog

Wire them to webhooks / DingTalk / WeCom by editing the receivers section of alertmanager.yml.

---

Logs

• PostgreSQL logs are shipped by promtail into Loki
• Grafana → Explore → pick the Loki data source → query {cluster="pg-meta"}
• Structured / slow-query analysis dashboard: PGLOG Analysis

---

Going deeper

• INFRA module: /docs/infra/
• pg_exporter: /docs/pg_exporter/
• Monitoring concept: /docs/concept/monitor