Programming · by Nicolás Díaz · 2026-10-12

Redis 8 + Postgres as your only datastore — when does this combo actually break?

We've been running a 'boring stack' of Postgres for source of truth + Redis for caching/queues/pubsub for 2 years now at decent scale (few million requests/day) and honestly it just works. Keep seeing posts about people reaching for Kafka or dedicated search engines way earlier than they need to. At what point did you actually outgrow this combo and what pushed you there?

0 sparks · 9 comments

Comments

Avery Johnson · 0 likes

we hit the wall at full text search, postgres FTS is fine until you need real relevance tuning then elasticsearch becomes necessary

Bianca Santos · 0 likes

redis pubsub message loss during failover is what finally pushed us to kafka for anything that cant tolerate dropped messages

Brave Sparrow · 0 likes

yeah redis streams help a bit here but kafka's durability guarantees are just different tier

Aria Thompson · 0 likes

pg_partman + proper indexing got us further than i expected honestly, postgres is criminally underrated for scale

Tomás Gómez · 0 likes

the boring stack wins until you need true horizontal write scaling, then youre looking at citus or just sharding manually which is its own pain

Huy Le · 0 likes

we're at like 50m req/day on this exact combo still, people way overestimate when they need 'real' infra

Luciana Vargas · 0 likes

this is the real answer, most companies never need to leave postgres+redis honestly

Felix Weber · 0 likes

jsonb columns in postgres replaced what would have been a mongo use case for us, another reason to stay boring

Amelia Hughes · 0 likes

biggest gotcha for us was redis memory fragmentation over time, had to restart instances periodically until we tuned maxmemory-policy properly