Connection Pooling

pyrsedis uses an async connection pool internally. Connections are created on demand and reused across calls.

How it works

1. Each Redis() instance owns one pool
2. Pool creates connections lazily (up to pool_size)
3. Idle connections are reused in LIFO order (better cache warmth)
4. Connections idle longer than idle_timeout_ms are dropped
5. Connections are initialized with AUTH + SELECT on creation

Configuration

r = Redis(
    pool_size=8,             # max concurrent connections
    idle_timeout_ms=300_000, # drop idle connections after 5 minutes
)

Monitoring

r.pool_idle_count     # number of idle connections in the pool
r.pool_available      # idle + free capacity

Best practices

Match pool size to concurrency

If your app uses 4 threads hitting Redis, pool_size=4 is sufficient. Extra connections waste server memory (~10 KB each).

Don't create multiple Redis instances for the same server

Each Redis() creates its own pool. Sharing one instance across threads is safe and efficient.

Idle timeout

The default 5-minute idle timeout works well for most apps. Lower it if you're behind a firewall that kills idle TCP connections sooner.