One shared cache for every replica
Why Redis?
With several app replicas, the on-disk cache becomes inconsistent: each replica holds its own copy of the page. A shared Redis fixes that.
- A single source of truth for ISR
- Instant revalidation across all replicas
- tag markers live in Redis
- no invalidations lost on deploy
- Memory kept in check with TTLs
Pitfalls
- Never touch Redis during
next build. - Put a timeout on every command.
- Flush the cache when the build changes.
A cache you can't invalidate correctly is just a source of bugs.
In short: worth it, but only with timeouts, fallbacks and clear logs.
Published: