Connection Pooling: PgBouncer
Run PgBouncer in transaction-pooling mode, size pools correctly, and avoid the prepared-statement trap.
Connection Pooling: PgBouncer is a free SQL Academy lesson on CoddyKit — lesson 3 of 4. You can read the complete lesson below for free — then practise it hands-on in the browser with a built-in code editor and a 24/7 AI tutor. It is part of the SQL Academy learning path, one of 4 lessons in the course, and your progress syncs across the web and the CoddyKit app.
Why Pool Connections?
Every Postgres connection is a separate OS process — memory and CPU heavy. Apps that open and close connections per request quickly exhaust the server. Pooling reuses a small set of long-lived connections.
Per-Connection Cost
A typical Postgres backend uses 5–10 MB of RAM. 500 connections = ~5 GB RAM just for backends. PgBouncer can multiplex 1000s of clients onto 50 backend connections.
Three Pooling Modes
- Session — client gets a dedicated connection for the whole session
- Transaction — client gets a connection per transaction
- Statement — connection per statement (rarely used)
Transaction Mode (Recommended)
Each transaction gets a server connection. The same client may use different connections across transactions.
# pgbouncer.ini
[databases]
mydb = host=primary port=5432 dbname=mydb
[pgbouncer]
pool_mode = transaction
listen_port = 6432
max_client_conn = 1000
default_pool_size = 50Transaction Mode Caveats
Features that require session state break:
- Prepared statements (need same backend across calls)
- SET LOCAL ... (per-transaction is fine; SET ... persisting across is not)
- LISTEN / NOTIFY (need persistent connection)
- Cursors that outlive a transaction
Session Mode
Use when transaction mode breaks your app — but you get fewer concurrent clients.
Sizing Pools
Rule of thumb: default_pool_size ≈ vCPU count × 2 + spindles. Too high = context switching kills throughput; too low = waiting on pool.
PgBouncer in Front of HAProxy
Typical production stack:
App → HAProxy (read/write routing) → PgBouncer (pooling) → Postgres
Prepared Statements in Transaction Mode
Recent PgBouncer (1.21+) supports protocol-level prepared statements in transaction mode. Older versions: use simple queries or session mode.
Monitoring PgBouncer
Connect to the admin db (special database):
psql -p 6432 -U pgbouncer pgbouncer
-- Commands:
SHOW STATS;
SHOW POOLS;
SHOW CLIENTS;
SHOW SERVERS;Alternatives
- pgpool-II — pooling + load balancing + query rewriting
- Odyssey — Yandex's pooler, multi-threaded
- Driver-side pools — usually combined with PgBouncer for cluster-wide pooling
Application Pool + PgBouncer
Apps typically run their own connection pool (HikariCP, pgxpool) AND go through PgBouncer. Two-layer pooling — app pool keeps connections to the bouncer; bouncer multiplexes them onto Postgres.
Recap
PgBouncer is essential at any nontrivial scale.
- Transaction mode is the standard
- Watch out for prepared statements / SET / LISTEN
- Pool size ~ vCPU × 2
- Monitor with SHOW POOLS
Quick Check
Which PgBouncer mode multiplexes the most clients on the fewest server connections?
Frequently asked questions
Is the “Connection Pooling: PgBouncer” lesson free?
Yes — the full text of “Connection Pooling: PgBouncer” is free to read here on the web, and the SQL Academy course includes 4 lessons in total. To practise it interactively (a built-in code editor and a 24/7 AI tutor) and unlock the rest of the SQL Academy course, upgrade to CoddyKit PRO.
What will I learn in “Connection Pooling: PgBouncer”?
Run PgBouncer in transaction-pooling mode, size pools correctly, and avoid the prepared-statement trap. You practise SQL Academy with hands-on code you run directly in the browser, and a 24/7 AI tutor answers your questions as you work through the lesson.
Do I need any experience to start SQL Academy?
No prior experience is required. SQL Academy on CoddyKit is structured for beginners through advanced learners; this is — lesson 3 of 4, so you can start here or from the beginning and move at your own pace.
How long does the “Connection Pooling: PgBouncer” lesson take?
Most CoddyKit lessons take about 5–10 minutes. Each one is bite-sized and interactive, so you make steady progress and pick up exactly where you left off across the web and the app.
Can I write and run code in this SQL Academy lesson?
Yes. Every SQL Academy lesson includes a built-in code editor, so you write and run real code right in your browser and get instant AI feedback — no local setup required.
All lessons in this course
- pg_stat_statements: Top Queries
- pgBadger for Log Analysis
- Connection Pooling: PgBouncer
- Capacity Planning and Bloat Audits