dmn diagram for Redis Cache Invalidation Flow

Tech architecture infographic titled "Redis Cache Invalidation" using archetype DATA PIPELINE (extract → transform → load). Show a technically accurate cache invalidation flow as a vector-clean whiteboard-style engineering diagram for a senior engineer: Browser client, API service, Primary Database, Redis Cache, Message Queue, Invalidation Worker, and optional Read Replica / Metrics box, all as labeled boxes connected with directional arrows. Include box details for each component: icon, canonical English name, and one-line English role description. Required boxes: Browser — "Web Browser" — "Sends read and write requests over HTTPS"; API — "REST API Service" — "Handles reads, writes, cache lookups, and invalidation publish"; DB — "Primary Database" — "Source of truth for durable records"; Cache — "Redis Cache" — "Stores hot key-value data with TTL"; Queue — "Message Queue" — "Buffers invalidation events for asynchronous fan-out". Add Invalidation Worker — "Invalidation Worker" — "Consumes events and deletes or updates Redis keys". Optional support box: "Metrics / Logs" — "Tracks cache hit rate, stale reads, and invalidation latency". Structure the flow in ETL/ELT style lanes: Extract = write event captured from API after DB commit; Transform = invalidation event normalized into key patterns or entity IDs; Load = Redis DEL, UNLINK, EXPIRE, or write-through refresh. Show arrows with short English labels such as: Browser → API "HTTPS REST request"; API → Redis "GET cache:key"; Redis → API "cache hit / miss"; API → DB "SELECT row"; DB → API "result set"; API → Browser "200 OK JSON"; Browser → API "HTTPS PUT / POST"; API → DB "BEGIN; UPDATE row; COMMIT"; API → Queue "publish invalidation event"; Queue → Invalidation Worker "entity ID, cache keys"; Invalidation Worker → Redis "DEL key" or "UNLINK key"; Invalidation Worker → Redis "EXPIRE key TTL"; API → Redis "SET key JSON EX 300" for cache repopulation. Make the lifecycle technically accurate: on read, API checks Redis first, falls back to DB on miss, then repopulates cache and returns 200 OK JSON; on write, API commits to the primary DB before publishing invalidation; invalidation is asynchronous via queue and worker; note common strategies visually: delete-on-write, write-through refresh, TTL backstop, and key-versioning tag. Include small callout notes in English such as "Cache-aside pattern", "Eventual consistency window", "Avoid invalidating before DB COMMIT", and "Use idempotent consumers". Add a numbered legend 1-7 in English: 1. Client sends HTTPS read or write request to REST API. 2. API checks Redis for the requested key. 3. On cache miss, API reads from Primary Database and returns 200 OK JSON. 4. API stores fresh value in Redis with TTL for subsequent reads. 5. On data change, API commits UPDATE or INSERT to Primary Database first. 6. API publishes invalidation event to Message Queue with entity ID or cache key pattern. 7. Invalidation Worker consumes event and deletes, expires, or refreshes Redis entries to prevent stale reads. Visual style: hand-drawn whiteboard, cool blue and cyan palette, sketchy marker outlines, subtle grid or whiteboard texture, engineering notebook mood, high readability, balanced spacing, editorial developer-blog illustration, isometric or flat tech-diagram style, vector-clean infographic layout. Emphasize arrows, staged ETL lanes, compact annotations, and senior-engineer precision. Do not imply audited security or vendor-specific reference architecture. No real cloud-vendor logos; if needed, use only generic cloud icons. All text MUST be written in English (array). Every heading, label, caption, legend and metric name in the image must be in English — not English. Spell each English word correctly using English characters and diacritics. Numbers stay as digits, no real cloud-vendor logos (AWS / GCP / Azure) — use generic cloud icons, no watermarks No real cloud-vendor logos (AWS, GCP, Azure) beyond generic cloud icons. Common protocol names (HTTPS, TCP, JWT, OAuth, REST, GraphQL) stay in canonical English form. No security-claim overstatements (do not present diagrams as audited reference architectures).