Hey there! We've found the following might help you get unblocked faster:

• 🧵 bash: line 1: psql: command not found

• 🧵 We are unable to connect to the database via SSH

• 🧵 Urgent: Unable to connect to production database via SSH (critical data)

If you find the answer from one of these, please let us know by solving the thread!

WHAT HAPPENED (The Real Issue)

When you edited workflow_entity.versionId manually in the Railway Postgres GUI, you did something n8n does not allow at the DB level:

versionId is a NOT NULL + foreign key–dependent field

In n8n:

• workflow_entity.versionId is required

• It references workflow_version.id

• It is used during boot to migrate, load workflows, and lock versions

• Setting it to NULL or empty string means…

n8n boot MIGRATIONS now crash mid-transaction

When the Worker/Primary starts, n8n loads all workflows:

SELECT ... FROM workflow_entity w
JOIN workflow_version v ON w.versionId = v.id

But since a row has versionId = null, Postgres throws:

ERROR:  null value in column "versionId" violates not-null constraint

Depending on where the failure occurred, n8n leaves a transaction open or keeps retrying.

WHY THE POSTGRES SERVICE NOW HANGS

Railway Postgres logs show the “Database Connection” spinner forever because:

n8n is repeatedly attempting migrations → locking system catalog

This causes:

• Table locks

• Long-running transactions

• Any new connection (psql, Railway GUI) = timeout

GOOD NEWS: You probably do NOT need to delete your database

Because…

Railway almost never becomes unrecoverable

99% of these cases are fixed by killing the bad query and correcting the corrupted row.

THE FIX (safe, no data loss)

You need to:

1. Stop the Worker + Primary

2. Kill stuck connections

3. Connect to Postgres

4. Restore the broken row’s versionId

STEP 1 — Stop services

In Railway:

Stop:

• Worker

• Primary (main n8n)

Leave only:

• Postgres

• Redis

This prevents n8n from instantly reconnecting and re-locking the DB.

STEP 2 — Kill all active pids in Postgres

Railway → Postgres → Shell → run:

SELECT pg_terminate_backend(pid)
FROM pg_stat_activity
WHERE datname = current_database()
AND pid <> pg_backend_pid();

If GUI is frozen, run via psql:

psql "postgresql://USER:PASSWORD@HOST:PORT/railway"

Then run the terminate query again.

STEP 3 — Fix the broken row

Find the corrupted workflow:

SELECT id, name, versionId FROM workflow_entity WHERE versionId IS NULL OR versionId = '';

You should get 1 row—the one you changed.

Now find the latest version:

SELECT id FROM workflow_version WHERE workflowId = '<workflow-id>' ORDER BY createdAt DESC LIMIT 1;

Then restore it:

UPDATE workflow_entity
SET "versionId" = '<the version id you found>'
WHERE id = '<workflow id>';

STEP 4 — Verify the schema

Check:

\d workflow_entity

Confirm from output:

versionId UUID NOT NULL

If it's NOT NULL, you MUST set a value again (n8n requires it).

STEP 5 — Restart n8n Primary (only)

If it boots normally, restart Worker after.

If this works → your DB is safe

If all connections still hang after terminating backends

You have 2 possibilities:

Option A — There's still an open transaction from n8n (most likely)

Run:

SELECT * FROM pg_stat_activity;

Look for long-running queries like:

idle in transaction

Kill them with:

SELECT pg_terminate_backend(<pid>);

Keep terminating until none remain.

Option B — Corrupt row causes boot-loop

If you cannot restore the versionId because you don't know it:

You can copy version from another workflow version:

UPDATE workflow_entity SET "versionId" = (
  SELECT id FROM workflow_version WHERE workflowId = workflow_entity.id LIMIT 1
)
WHERE workflow_entity.versionId IS NULL OR workflow_entity.versionId = '';

🧨 MUST I DELETE THE DATABASE? (Rare)

Only if ALL of these are true:

• You cannot connect

• Terminating backends fails

• Even psql times out when services are stopped

• Railway DB logs show corrupted Postgres cluster, not locked tables

From your description, this is only a lock, not corruption.

You should NOT need to delete your DB.

Hi, I would ignore the obvious bot response before mine. Here's what I think is going on as I'm having the exact same issue (not with N8n though). Internal networking seems to not work at all in my new project. Using public Host & Port / URL works everytime. So far i've tried with a Mysql DB, Redis and MinIO service..