9 · Gotchas & limits
Everything that will surprise you, in one place. Each item links back to the
step that explains it.
Redis
Gotcha
Impact
Step
No persistence, volume, or replica
restart ⇒ all keys gone
4
No maxmemory/eviction policy
memory full ⇒ writes fail (noeviction default)
4
KEYS * for listing
blocks Redis on big datasets
4
Re-PUT without ttlSeconds
silently clears the TTL
4
No auth/ACL/TLS on redis://
anyone on the network has full access
1
REDIS_URL logged at startup
credential leak if the URL has a password
4
No command timeout
requests queue during outages instead of failing
5
Backend
Gotcha
Impact
Step
String(value) on all values
objects become "[object Object]"
4
ttlSeconds: 0 = no TTL; "abc" → 503
truthiness ≠ validation
4
Malformed JSON → 500 not 400
middleware drops parser status
3
/ready flag set only on ready/end
stale 200 during reconnects
5
CORS *, errors leak internals
trusted networks only
1
depends_on isn't readiness
backend starts before Redis accepts
2
Frontend
Gotcha
Impact
Step
proxy_connect_timeout 2s in nginx
dead backend ⇒ fast /api failure (the banner)
7
Stale BACKEND_URL
needs container restart, not rebuild
7
UI "shared state" demo
only means same-Redis views — no cross-region sync
8
Architecture
Gotcha
Impact
Step
No cross-region replication
data is per-cluster
8
No global LB/DNS
failover is manual/DNS
8
Release workflow tags lightweight tags pushed with --follow-tags
tag may not reach remote
deploy ref
If this went to production
AuthN/Z, restricted CORS/ingress, secrets + log redaction, input validation,
bounded requests and Redis timeouts, durable/HA storage (persistence,
replica, or managed Redis), SCAN not KEYS, eviction policy decision,
observability, and a real cross-region consistency strategy. All future work —
none of it is here today.
Done. Reference: API guide ·
Configuration ·
Deployment ·
Maintaining docs
Answered: 0 / 0
Correct: 0 / 0
Reset