Cloudflare Workers

Workers have no long-lived process, so connection reuse is the whole story. Two good options:

PostgreSQL via Hyperdrive

Hyperdrive keeps a warm pool near the origin and speaks TLS to Databasezy.

1. Create the Hyperdrive config with the direct endpoint (Hyperdrive is the pooler).

   shell

   npx wrangler hyperdrive create zb-pg-prod \
     --connection-string="postgresql://app:<password>@pg-7f3k.us-east.databasezy.com:5432/app?sslmode=verify-full"

2. Bind it and use pg with the binding’s connection string.

   wrangler.toml

   [[hyperdrive]]
   binding = "DB"
   id = "<hyperdrive id>"

   src/index.ts

   import { Client } from "pg";
   
   export default {
     async fetch(_req: Request, env: { DB: Hyperdrive }) {
       const client = new Client({ connectionString: env.DB.connectionString });
       await client.connect();
       const { rows } = await client.query("select now()");
       await client.end();
       return Response.json(rows[0]);
     },
   };

3. Allow Cloudflare’s egress on the instance: Hyperdrive connects from Cloudflare’s published IP ranges (https://www.cloudflare.com/ips-v4), which you can add as CIDRs.

libSQL over HTTP

No sockets needed; the Worker calls the Hrana HTTP endpoint directly.

shell

npx wrangler secret put ZB_TOKEN

src/index.ts

import { createClient } from "@libsql/client/web";

export default {
  async fetch(_req: Request, env: { ZB_TOKEN: string }) {
    const db = createClient({ url: "https://libsql-7f3k.us-east.databasezy.com", authToken: env.ZB_TOKEN });
    const rs = await db.execute("select datetime('now') as now");
    return Response.json(rs.rows[0]);
  },
};

Cloudflare’s egress ranges go on the allow-list the same way.

Valkey and FerretDB

Use connect() from cloudflare:sockets with secureTransport: "on"; ioredis and the MongoDB driver do not run on Workers. For caching, Workers KV or a Node-runtime service in front of Valkey is usually simpler.