Your AEM go live checklist advanced: perspective, decisions, and practical tradeoffs.
It is a late night in the war room. Pizza boxes on the table. Slack popping every minute. The AEM author is quiet, the publish farm looks calm, and the CDN is primed. Ads are booked and the press note is queued. Then someone tests the product detail page and sees a cache miss storm. Traffic hits origin, CPUs climb, and a single forgotten filter in the Dispatcher lets a query sling wide open. That is how most launch stories start. Not because teams do not care, but because **AEM go live** has lots of moving parts and small misses add up fast.
Analysis: the AEM launch pieces that matter
Before a launch, decide the shape of the stack and how it will behave under load. Start with **topology**. Most teams are best served with **TarMK** on author and publish, **cold standby** for author, and at least two publish nodes. Go **MongoMK** only for very large editorial teams or very large write volume. Keep **data store** external for assets, file or S3 compatible, to avoid ballooning the repository.
Lock down **Dispatcher** rules. Be strict with filters and cache rules. Only cache what is safe. Block POST and selectors you do not own. Add **flush agents** to clear content with selectors and extensions your site actually uses. If you add a CDN on top, define who wins on **TTL** and **invalidation** so you do not chase ghosts between two caches.
Check **Oak indexes**. If your queries depend on a property or a path constraint, ship the matching indexes already built. A surprise reindex during peak is a rough way to meet your SRE. Plan **compaction** windows so they never collide with a campaign.
Sort **security** early. Terminate **TLS** where you can control ciphers. Force **HTTPS** from the edge. Wire **SAML** or **LDAP** for author if you need single sign on. Review ACLs for editors so **Publish** does not inherit anything it should not.
Get **assets** under control. Define renditions, check ImageMagick or GraphicsMagick calls, and cap concurrency. Heavy asset workflows can swamp a node. Offload long workflows to a dedicated worker if you can.
Finally, keep **search and SEO** clean. Ship the right 301 map, canonical tags, XML sitemaps, and a sane **robots.txt**. Exclude author from bots. Do not cache the sitemap forever at the edge.
Risks you can see from a mile away
- Reindex surprise: an index flag flips and your publish nodes crawl during prime time.
- Cache stampede: CDN miss pushes every request to Dispatcher, which then misses due to a rule gap.
- DNS delay: long TTL on the apex domain means your fix will take hours to land.
- SSL chain: a missing intermediate certificate breaks old mobile browsers and kills conversion.
- Workflow pileup: asset ingestion triggers hundreds of transcodes and drains CPU.
- Session leaks: custom code forgets to close sessions and file handles.
- Content freeze drift: editors keep publishing and flushing while you are rolling out.
- Logs without eyes: everything is logged, nothing is watched, alerts never fire.
AEM go live decision checklist
- Topology: TarMK or Mongo. How many publish nodes. Cold standby for author set and tested.
- Data store: File or S3 style. Backup and restore time measured for real.
- Dispatcher: Filters locked. Cache rules mapped. Vanity URLs and rewrites placed in the right layer.
- CDN: Origin health checks, TTL policy, and invalidation path pattern agreed.
- Indexes: Custom indexes shipped and warm. No reindex flag in production during launch week.
- Compaction: Schedule outside business peaks. Manual run tested.
- Security: HTTPS only, HSTS ready, SAML or LDAP tested on author. Admin users audited.
- Workflows: Max parallel jobs set. Long jobs offloaded. Failed job alerting wired.
- Monitoring: JVM, GC, thread pools, queue depth, replication, and Dispatcher cache ratio on a screen.
- Content: Freeze window communicated. Rollback path defined. Editors trained on emergency unpublish.
- SEO: 301 map verified, sitemap reachable, robots rules correct, tracking tags firing through Launch or GTM.
- Release: Blue green or rolling plan clear. Health endpoint agreed. Smoke tests scripted.
Action items before you press go
- Run a full **production rehearsal** with the CDN on and the same TLS cert. Push traffic with a load test that reflects real pages and cache mix.
- Warm **critical pages** through Dispatcher and CDN. Check cache keys for cookies and headers that should not vary.
- Export and import a **known good backup**. Time the restore. Keep it handy during launch.
- Ship a **readiness dashboard** that shows publish health, replication queues, error rate, and cache hit ratio.
- Freeze **oak run** and package installs during the window. No surprise tools.
- Turn on **alerts** for 5xx burst, slow response, queue growth, low disk, and young gen thrash.
- Confirm **DNS** cutover plan and TTL. Have a rollback host ready.
- Validate **security headers**: HSTS, X Content Type Options, X Frame Options, and CSP if you can support it.
- Test **author logins** through SSO with a real editor group, not only admins.
- Do a last mile **mobile test** on slow networks with cache cold to check page weight and time to first paint.
AEM can be a joy on launch day when the basics are steady. Keep the stack boring, the rules strict, and the playbook short. Your future self will thank you when the traffic hits and the graphs stay flat.