Interface ConnectOptions
acquireTimeoutMs?: number;
line?: string;
maxConnections?: number;
minConnections?: number;
pollOnly?: boolean;
}
Index
Properties
OptionalacquireTimeoutMs
Acquire timeout in milliseconds. Default 30_000.
Optionalline
Migration line name. Default "main".
OptionalmaxConnections
Max sqlx pool connections per process. Default 5.
Size this for your fleet, not just one process.
Each worker process opens max_connections + 1 connections to Postgres
(the +1 is a dedicated LISTEN socket). At 10 pods with the default you
use 60 connections. Postgres ships with max_connections = 100, so
you'll run out fast if you scale without adjusting this.
Job handlers themselves do NOT hold a connection while running — a
connection is only acquired briefly for fetch, heartbeat ticks, and
complete/fail. So max_connections can be much smaller than
workerConcurrency (roughly concurrency ÷ 5 is a reasonable starting
point). When using PgBouncer, set this to your per-process PgBouncer
pool allocation, not your raw Postgres max_connections.
OptionalminConnections
Min idle pool connections. Default 0.
OptionalpollOnly
Disable the LISTEN/NOTIFY subscriber and use polling only.
Required when connecting through PgBouncer in transaction-pooling mode — LISTEN needs a persistent session connection, which transaction pooling does not provide. In poll-only mode the worker falls back to a configurable poll interval (default 1 s) instead of being woken immediately on new jobs. Session-mode PgBouncer and direct Postgres connections do not need this.
Connection options for
Queue.connect.