It starts harmlessly enough:
โฝ Postgres for core data ๐ช
โฝ Redis for caching โก
โฝ Elasticsearch because โsearch is hardโ ๐ค
โฝ DynamoDBโฆ no one remembers why, but itโs in prod now ๐ฌ
โฝ ClickHouse for analytics because someone read a blog post about it once ๐
โฝ SQLite for that one internal CLI tool, obviously ๐ก
โฝ Oh, and MongoDB is still running that one service from 2019 โณ
Fast-forward 12 months and your stack looks like a ๐ฑ๐ฎ๐๐ฎ๐ฏ๐ฎ๐๐ฒ ๐ฏ๐๐ณ๐ณ๐ฒ๐โฆ
Except now you need ๐ฟ๐ฒ๐ฎ๐น ๐ด๐๐ฟ๐ to monitor it, a second team to maintain it,
and a ๐๐ต๐ฒ๐ฟ๐ฎ๐ฝ๐ ๐๐ต๐ถ๐๐๐น๐ฒ๐ฟ ๐บ๐ฎ๐ฟ๐บ๐ผ๐ for on-call rotation.
๐๐ฟ๐ฒ ๐๐ต๐ฒ๐๐ฒ ๐ฐ๐ต๐ผ๐ถ๐ฐ๐ฒ๐ ๐๐๐ฟ๐ฎ๐๐ฒ๐ด๐ถ๐ฐโฆ ๐ผ๐ฟ ๐ท๐๐๐ ๐๐๐บ๐ฝ๐๐ผ๐บ๐ ๐ผ๐ณ ๐ฑ๐ฒ๐๐ฒ๐น๐ผ๐ฝ๐ฒ๐ฟ ๐ผ๐ฝ๐๐ถ๐บ๐ถ๐๐บ?
(aka โsure, Iโll just spin up one more DB – what could go wrong?โ)
โ ๐ฆ๐ผ๐บ๐ฒ ๐๐ถ๐ด๐ป๐ ๐๐ผ๐โ๐ฟ๐ฒ ๐๐๐ฟ๐๐ถ๐๐ถ๐ป๐ด, ๐ป๐ผ๐ ๐บ๐ฎ๐ป๐ฎ๐ด๐ถ๐ป๐ด:
โฝ Backups? “I think they’re happeningโฆ”
โฝ Nobody touches the legacy MongoDB because it might be possessed ๐ป
โฝ Every new feature adds a new datastore – and a new panic attack
โฝ Your monitoring dashboard looks like a spaghetti monster
โ ๐ฆ๐ผ ๐๐ต๐ฎ๐โ๐ ๐๐ผ๐๐ฟ ๐ฟ๐ฒ๐ฎ๐น๐ถ๐๐?
Are you keeping things tight with 1-2 core databases?
Or just leaning into the chaos with no limits?
