If you're connecting Activepieces (or n8n/Make) directly to PostgreSQL in production, you're likely bleeding connections. Most automation platforms keep HTTP requests open for 30-120s. If your workflow hits a database node 5 times with standard

session

pooling, you're holding 5 connections open simultaneously. With Supabase's default 60 connection limit, you hit the wall at ~12 concurrent executions. The fix: Switch to

transaction

pooling mode (PgBouncer) and set statement timeouts < 30s. In Activepieces, use the Postgres piece with explicit

LIMIT

clauses and avoid long-running

SELECT *

without pagination. Even better: cache lookup tables in memory using the Storage piece instead of querying the DB on every loop iteration. What's your current DB connection setup? Seeing a lot of

FATAL: sorry, too many clients already

errors in the wild? https://cdn.discordapp.com/attachments/1053778461275201576/1481858138440208486/1773372892166-post.png?ex=69b4d76f&is=69b385ef&hm=cdfcb16ce9c1faff4f30e521cc3820ba5cfcbb5c60605254ea58011693f035e4&