💨 Aeon::Refresher – 4 + 1 Architecture

Breathes new life into stale caches.

---

1️⃣ Logical View

Purpose:
Proactively refresh cached query results before they expire.
Acts as Aeon’s self-healing temporal daemon.

Responsibilities - Detect near-expiry keys (TTL < threshold). - Recompute occurrences via Timekeeper. - Write fresh values back to Redis. - Emit Argus heartbeat metrics.

---

2️⃣ Process View

1. Heartbeat job runs every N minutes.
2. Refresher scans aeon:cache:* keys.
3. If TTL < 30 min → load window metadata.
4. Calls Aeon::Timekeeper.within(span: window) to rebuild.
5. Stores new payload with extended TTL.
6. Logs aeon.refresher.refreshed.

---

3️⃣ Development View

File: app/services/aeon/refresher.rb

class Aeon::Refresher
  THRESHOLD = 30.minutes

  def self.run
    keys = Redis.current.keys("aeon:cache:*")
    keys.each do |key|
      ttl = Redis.current.ttl(key)
      next unless ttl.between?(0, THRESHOLD.to_i)
      period = decode_period_from_key(key)
      refreshed = Aeon::Timekeeper.within(span: period)
      Redis.current.set(key, JSON.dump(refreshed), ex: Aeon::CacheManager::TTL_DEFAULT)
      Argus.emit("aeon.refresher.refreshed", meta: { key:, ttl: })
    end
  end

  def self.decode_period_from_key(key)
    # optional encoded metadata for reconstruction
  end
end

---

4️⃣ Physical View

Attribute	Description
Runtime	Sidekiq / Cron job (Aeon::Heartbeat)
Store	Redis and Postgres
Dependencies	Timekeeper, CacheManager

---

➕ 1 Scenario View

Scenario A: Cache Prewarming

At 23:45, Refresher detects next day’s window TTL → rebuilds early.

Scenario B: Burst Invalidations

Several timestamps updated → Refresher regenerates windows in parallel.

Observability:
Metrics aeon_refresher_runs_total, aeon_refresher_duration_seconds.

---