U6: Valkey tuning

[jaturapornchai]

Summary

Tune Valkey for cache-heavy usage per U6 plan: 1GB maxmemory, allkeys-lru, no RDB/AOF persistence, tcp-backlog 2048, tcp-keepalive 60.

Test plan

• docker compose config parses successfully
• Verify valkey container starts Up after deploy

[Copilot]

Pull request overview

This PR updates the Valkey (Redis-compatible) service configuration in docker-compose.yml to run as an aggressive, cache-first instance tuned for heavy cache workloads per the U6 plan.

Changes:

• Increase Valkey memory limit to 1GB and switch eviction policy to allkeys-lru.
• Disable AOF and RDB snapshot scheduling, and tune TCP settings (backlog/keepalive).

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

--save "" and --appendonly no stop new RDB/AOF writes, but the service still mounts a persistent volume at /data. If a dump.rdb already exists in that volume, Valkey can still load it on startup, which conflicts with the “no persistence” goal. Consider removing the named volume (or switching to tmpfs: /data) and ensuring /data is empty on deploy to make the cache truly ephemeral.

[Copilot]

--tcp-backlog 2048 may not take effect if the kernel net.core.somaxconn is lower (Valkey will cap the effective backlog). If you rely on 2048 for connection spikes, consider also setting an appropriate sysctl for the container/host so the backlog setting is actually honored.