Azure Managed Redis filling with seasonal event Q&A cache entries risks out-of-memory errors for the city FAQ bot. Which eviction approach fits?
Select an answer to reveal the explanation.
Short Explanation
When the festival FAQ drawer overflows, allkeys-lru is the polite ‘toss the least-touched pamphlets first’ rule so Redis stays up instead of face-planting under memory pressure.
Full Explanation
When Azure Managed Redis fills with seasonal event Q&A cache entries, memory pressure can take the city FAQ bot offline unless an eviction policy reclaims space under load. Configuring a policy such as allkeys-lru lets less-recently-used seasonal entries leave memory first so the managed cache remains available for hot civic FAQ traffic. Disabling eviction and allowing Redis to crash when the seasonal corpus fills memory sacrifices availability for a temporary festival workload and fails municipal uptime goals. Evicting only configuration keys while retaining every seasonal Q&A entry forever does not free the working set that is actually exhausting memory, so OOM risk remains. Switching the entire FAQ store to local disk files on each Container Apps replica removes the shared cache coherence Redis provides and multiplies cold-path work across replicas without solving citywide FAQ latency. Exam caveat: eviction policy choice is a deliberate availability–freshness tradeoff for civic caches, not a substitute for sizing or TTL design. Operationally, set an appropriate maxmemory policy, watch used_memory and evicted_keys during peak event weeks, and confirm the FAQ bot still serves after synthetic memory pressure tests.