REDIS INVESTIGATION REPORT

Documentation for Sessions

Redis Investigation Report — Session 128.2

From: Browser/Cowork Agent To: Session 128.2 (Cursor Agent) Date: 2026-02-20 Status: COMPLETE — Root cause identified, fix known

Redis Service

Exists: YES
Status: Online
Source Image: redis:8.6.0
Region: US East (Virginia, USA), 1 Replica
Internal hostname: redis.railway.internal (IPv6 enabled, alias: redis)
Public Networking: nozomi.proxy.rlwy.net:14089 → :6379
Private Networking: redis.railway.internal:6379
Data present: YES — BullMQ keys from backend-vitto-api visible (bull:vitto-tasks:*), 2 pages of key-value pairs
Metrics: CPU ~0 vCPU, Memory near 0, Public Network Traffic 0 B (expected — internal only)

Redis Service Variables (Raw Editor)

REDIS_PASSWORD="PIf..."  (redacted)
REDIS_PUBLIC_URL="redis://default:${REDIS_PASSWORD}@${RAILWAY_TCP_PROXY_DOMAIN}:${RAILWAY_TCP_PROXY_PORT}"
REDIS_URL="redis://${REDISUSER}:${REDIS_PASSWORD}@${REDISHOST}:${REDISPORT}"
REDISHOST="${{RAILWAY_PRIVATE_DOMAIN}}"
REDISPASSWORD="${{REDIS_PASSWORD}}"
REDISPORT="6379"
REDISUSER="default"

Key finding: REDIS_URL on the Redis service itself resolves to redis://default:***@redis.railway.internal:6379 (uses private domain). There is NO REDIS_PRIVATE_URL in the 7 user-defined variables — it must be one of the 13 Railway-auto-added variables.

backend-api (REDIS_URL) — api.gograb.space

Variable exists: YES
Raw value: ${{Redis.REDIS_PRIVATE_URL}}/1
DB suffix: /1 ✅ (correct)
Problem: Uses REDIS_PRIVATE_URL which resolves to 127.0.0.1 → BROKEN
PORT: 3000
Last deploy: "fix(deploy): ensure correct railway.toml fo..." — 16 hours ago, ACTIVE

backend-ai-api (REDIS_URL) — ai-api.gograb.space

Variable exists: YES
Raw value: ${{Redis.REDIS_PRIVATE_URL}}/2
DB suffix: /2 ✅ (correct)
Problem: Uses REDIS_PRIVATE_URL which resolves to 127.0.0.1 → BROKEN
PORT: 3001
Last deploy: "fix(deploy): ensure correct railway.toml fo..." — 14 hours ago, ACTIVE
Note: "This deployment was initiated by Railway. Your service was automatically redeployed."

backend-docs-api (REDIS_URL) — docs-api.gograb.space

Variable exists: YES
Raw value: ${{Redis.REDIS_PRIVATE_URL}}/3
DB suffix: /3 ✅ (correct)
Problem: Uses REDIS_PRIVATE_URL which resolves to 127.0.0.1 → BROKEN
PORT: 3002

backend-vitto-api (VITTO_REDIS_URL) — vitto-api.gograb.space

Variable exists: YES
Raw value: ${{Redis.REDIS_URL}}/0 ✅ (ALREADY FIXED)
DB suffix: /0 ✅ (correct)
Status: WORKING — Fixed earlier in this session by changing from ${{Redis.REDIS_PRIVATE_URL}}/0 to ${{Redis.REDIS_URL}}/0
PORT: 3003

Deploy Logs (backend-vitto-api — post-fix)

After fixing VITTO_REDIS_URL to use ${{Redis.REDIS_URL}}/0, the deploy logs confirmed:

LOG [RedisService] Redis connected
LOG [RedisService] Redis connected via VITTO_REDIS_URL
LOG [TaskQueue] BullMQ connected to Redis at redis.railway.internal:6379
LOG [SmartRouter] Health check complete: 2/2 providers healthy
LOG [NestApplication] Nest application successfully started
🚀 VITTO Super-Agent API READY on port 3003 (Railway)

Redis is fully working when using ${{Redis.REDIS_URL}} instead of ${{Redis.REDIS_PRIVATE_URL}}.

Root Cause Analysis

The root cause is identical across all 4 backends: All services were configured with ${{Redis.REDIS_PRIVATE_URL}} which is a Railway-auto-generated variable that resolves to 127.0.0.1:6379 instead of the actual private network address redis.railway.internal:6379.

The working reference is ${{Redis.REDIS_URL}} which resolves to redis://default:***@redis.railway.internal:6379 — this uses the actual private domain name and works correctly within Railway's internal network.

Why REDIS_PRIVATE_URL resolves to 127.0.0.1: This appears to be a Railway platform behavior where REDIS_PRIVATE_URL is meant for use within the Redis container itself (where localhost makes sense), NOT for other services connecting to Redis.

Recommended Fix

Change REDIS_PRIVATE_URL → REDIS_URL in 3 services:

Service	Current Value	Fix To
backend-api	`${{Redis.REDIS_PRIVATE_URL}}/1`	`${{Redis.REDIS_URL}}/1`
backend-ai-api	`${{Redis.REDIS_PRIVATE_URL}}/2`	`${{Redis.REDIS_URL}}/2`
backend-docs-api	`${{Redis.REDIS_PRIVATE_URL}}/3`	`${{Redis.REDIS_URL}}/3`
backend-vitto-api	`${{Redis.REDIS_URL}}/0`	✅ Already fixed

This is the exact same fix that was already successfully applied to backend-vitto-api earlier in this session. The deploy passed healthcheck and Redis connected at redis.railway.internal:6379.

Each variable change in Railway will trigger an automatic redeploy. All 3 fixes can be applied simultaneously — they are independent.

Verification Plan

After applying fixes:

Each service will auto-redeploy
Check deploy logs for Redis connected messages
Verify no more Redis connection failed warnings in logs
Confirm Redis Metrics show increased connections (currently only vitto connects)