Operations and ecosystem

What to observe

Track Redis connection and reconnect state, command latency and timeouts, pipeline batch size, and codec decode failures. Loaded maps add hit/miss rates, loader latency, write-behind queue depth, retries, dead letters, and shutdown draining. Queue saturation usually means the writer cannot keep up with Redis writes, not merely that the configured capacity is small.

From tests to production failures

LettuceClientsTest covers connection reuse and multi-thread, virtual-thread, and coroutine access. RedisFutureSupportTest checks ordering and failure propagation. Loaded-map tests cover writer failures, dead letters, and caller-scope ownership. Benchmark numbers describe one machine and payload set; they are not an application SLA.

Choose the next layer

• Stay here for direct Redis commands and coroutine adapters.
• Use bluetape4k-cache-lettuce for memoization and cache abstractions.
• Use bluetape4k-hibernate-cache-lettuce for Hibernate second-level caching.
• For database-backed loading and writing, define transaction boundaries with bluetape4k-jdbc, bluetape4k-hibernate, or Exposed repositories.
• Continue with bluetape4k-workshop and Exposed Workshop for complete examples.

Do not label cache-aside PUT management as true write-through. A write-through or write-behind example must include the database writer and its consistency boundary.

Source and tests

• Benchmark.md
• LettuceSuspendedLoadedMapTest.kt
• FastForyCompatibilityTest.kt