One database, one machine
Your shop keeps every order in one table on one database server. Each glowing dot is a write: a new order, a status change, a refund. The app servers in front hold no data, so adding more of them is easy.
The database is different. It has one disk, one CPU and a fixed amount of memory. Watch its two gauges: the fill level is disk space, and the bar underneath is write load, where 100% means the machine can't absorb writes any faster.
Business is good and traffic keeps climbing. Soon writes queue up and time out, and the disk runs out of room.
Each dot stands for thousands of writes, and time is squeezed: a real disk fills over months, not seconds.
The classic first fix:
$ psql shop -Atc "SELECT pg_size_pretty(pg_total_relation_size('orders'))"3212 GB$ df -h /var/lib/postgresqlFilesystem Size Used Avail Use% Mounted on/dev/nvme1n1 4.0T 3.3T 700G 83% /var/lib/postgresql
One table, one disk. Indexes, the write-ahead log and vacuum all need room on that same disk too.