n8n cannot use private networking for Redis or Postgres — forced to use TCP proxy

Referencing ray-chen's post from 4 months ago about n8n connection issues and private networking fixes.

Setup

- Template: n8n with workers (UI + Worker + Redis + Postgres)

- Region: asia-southeast1

- n8n Version: 2.1.4

- Running stable for: 2 months before incident

What Happened

On Dec 31 2025 at ~3pm IST, all services crashed simultaneously. Logs showed:

```

[Redis client] connect ETIMEDOUT

Unable to connect to Redis after trying to connect for 10s

Exiting process due to Redis connection error

```

Postgres logs:

```

database system was not properly shut down; automatic recovery in progress

```

Resource metrics showed low usage (~100-150MB memory) before crash — not OOM.

Resolution

Switching both Redis and Postgres from private networking to TCP proxy fixed the issue.

Configuration Details

I have all recommended private networking flags enabled:

- ENABLE_ALPINE_PRIVATE_NETWORKING="true"

- N8N_LISTEN_ADDRESS="::"

- QUEUE_BULL_REDIS_DUALSTACK="true"

Working config (TCP proxy):

DB_POSTGRESDB_HOST="${{Postgres.RAILWAY_TCP_PROXY_DOMAIN}}"

DB_POSTGRESDB_PORT="${{Postgres.RAILWAY_TCP_PROXY_PORT}}"

QUEUE_BULL_REDIS_HOST="${{Redis.RAILWAY_TCP_PROXY_DOMAIN}}"

QUEUE_BULL_REDIS_PORT="${{Redis.RAILWAY_TCP_PROXY_PORT}}"

Failing config (private network):

DB_POSTGRESDB_HOST="${{Postgres.RAILWAY_PRIVATE_DOMAIN}}"

DB_POSTGRESDB_PORT="5432"

QUEUE_BULL_REDIS_HOST="${{Redis.RAILWAY_PRIVATE_DOMAIN}}"

QUEUE_BULL_REDIS_PORT="6379"

Error: Could not establish database connection within the configured timeout of 20,000 ms

The Confusing Part

I deployed a NocoDB instance in the same Railway project with the same Redis and Postgres services — and private networking works perfectly:

NC_DB="pg://${{Postgres.RAILWAY_PRIVATE_DOMAIN}}:${{Postgres.RAILWAY_TCP_APPLICATION_PORT}}?u=${{Postgres.PGUSER}}&p=${{Postgres.PGPASSWORD}}&d=${{Postgres.PGDATABASE}}"

NC_REDIS_URL="${{Redis.REDIS_URL}}?family=6"

Connection test results:

• n8n → Redis (private): Fails

• n8n → Redis (TCP proxy): Works

• n8n → Postgres (private): Fails

• n8n → Postgres (TCP proxy): Works

• NocoDB → Redis (private): Works

• NocoDB → Postgres (private): Works

Known Issue Supposedly Fixed

GitHub Issue #13117 documented this exact problem — ioredis not resolving IPv6 for Railway's private network. It was reportedly fixed in n8n 1.79.0 with the QUEUE_BULL_REDIS_DUALSTACK flag.

I'm on n8n 2.1.4 — well past the fix. It still doesn't work.

The n8n template with internal Redis states:

> IMPORTANT: If your deployment fails, make sure you've opted out of IPv4 Private Networks in your Railway Account Settings → Feature Flags.

This feature flag does not exist in my Railway account settings.

Documentation vs Reality

Railway docs state:

- TCP Proxy is for "access from outside the private network"

- Private networking provides "faster communication and increased throughput"

- Using private network avoids "service-to-service egress costs"

Yet I'm forced to use TCP proxy for internal n8n Redis/Postgres communication — paying egress for what should be free internal traffic.

Questions

1. Why does NocoDB work on private networking but n8n doesn't, on the same Railway setup?

2. Is there a known issue with n8n's QUEUE_BULL_REDIS_DUALSTACK implementation?

3. The IPv4 Private Networks feature flag mentioned in templates doesn't exist — is this deprecated?

4. Is private networking in asia-southeast1 known to be less stable?

5. Was there an incident on Dec 31 not reflected on the status page?

Environment Info

n8nVersion: 2.1.4

platform: docker (self-hosted)

nodeJsVersion: 22.21.1

database: postgres

executionMode: scaling (single-main)

Happy to provide any additional logs or configuration details.