What metrics matter when administering a database?

When administering a database, it's important to track key metrics that show its performance, stability, and health.

Here are the main groups:

#Server performance

• CPU usage. If it stays above 80% for a long time, look for heavy queries.
• Memory usage. A shortage leads to disk spilling and slowdowns.
• Disk I/O. High I/O is a sign the disk can't keep up.

#Database activity

• Connection count. How many clients are currently working with the DB.
• Active queries. Currently running queries and their execution time.
• Locks / Waits. Locks and waits between transactions.

#Transaction state

• Commit / Rollback ratio. The ratio of successful to cancelled transactions.
• Deadlocks. Mutual locks (there shouldn't be any).

#Data storage

• Disk space usage. How much space the database and logs occupy.
• Table size / Index size. Which tables are growing the fastest.

#Backups

• Backup status. When the last backup ran, and whether it succeeded.

Summary: by tracking these metrics, an administrator can see where the bottleneck is - in queries, hardware, or settings - and prevent a failure in time.