Architecture for PostgreSQL DBAs
PostgreSQL architecture is how the server is built: a postmaster that forks one backend process per connection, shared memory including shared_buffers, background processes such as the checkpointer and WAL writer, and MVCC for concurrency. Autovacuum cleans dead row versions. Knowing this explains server behaviour and anchors the PostgreSQL DBA path.
Stage: Foundations · Part of the PostgreSQL DBA roadmap · Last reviewed 19 November 2026
What to learn in Architecture
- Name the main background processes and say what each one does
- Explain how MVCC lets readers and writers avoid blocking each other
- Describe how dead tuples form after UPDATE and DELETE
- Explain what VACUUM, autovacuum and transaction ID wraparound protection do
- Describe how the write-ahead log makes commits durable
Architecture interview questions
Why does an UPDATE in PostgreSQL create bloat?
An UPDATE writes a new row version and marks the old one dead instead of overwriting it, because readers with older snapshots may still need the old version. Dead tuples keep occupying space until VACUUM makes it reusable, so heavy updates can leave tables larger than their live data.
What is transaction ID wraparound and how is it prevented?
Transaction IDs are 32-bit and compared circularly, so after about two billion transactions very old rows would appear to be in the future. VACUUM freezes old tuples to keep them visible. Autovacuum runs anti-wraparound passes, and failing to keep up can force the database to stop accepting writes.
Where Architecture fits in the PostgreSQL DBA roadmap
Architecture belongs to the Foundations stage (beginner level), which takes about 6 hours. Relational basics and the Postgres way.
Other units in this stage
Back to the full PostgreSQL DBA roadmap