การตรวจสอบและการบันทึกใน Neo4j

A production graph database must stay healthy. Monitoring tells you about slow queries, memory pressure, and failed transactions before users notice.

Neo4j writes several logs in the logs/ directory:

• neo4j.log — general output
• debug.log — detailed diagnostics
• query.log — executed queries

Turn on query logging in neo4j.conf to capture every query and its duration. This is invaluable for spotting slow patterns.

Use SHOW TRANSACTIONS to see what is executing right now, including how long each has been running.

If a query runs away, terminate it by its transaction id to free resources.

Neo4j exposes metrics like transaction rate, page cache hit ratio, and heap usage. These can be exported to monitoring systems.

• Page cache hit ratio should stay high
• Watch heap and GC pauses

Enable the Prometheus endpoint in config so external dashboards (like Grafana) can scrape Neo4j metrics.

Track how large your stores grow. SHOW DATABASES and store files reveal node, relationship, and property counts over time.

The page cache keeps graph data in memory. If your dataset fits in cache, reads are very fast. Monitor the hit ratio to size it correctly.

Set alerts for: low cache hit ratio, long GC pauses, sustained high transaction failures, and disk filling up. Catch issues early.

A healthy admin routine: review query.log weekly, watch metrics dashboards daily, and verify backups regularly.

You learned to monitor Neo4j:

• Read neo4j.log, debug.log, and query.log
• Enable query logging with thresholds
• Inspect and terminate transactions
• Export metrics to Prometheus/Grafana
• Tune and watch the page cache