Cumulative Static Chunk Pooling
One of the most frustrating errors in production single-page and server-rendered Next.js applications is:
ChunkLoadError: Loading chunk 842 failed.NextDeploy eliminates this error permanently through Cumulative Static Chunk Pooling.
Why ChunkLoadError Happens
When you run next build, Webpack or Turbopack creates content-hashed assets:
- Release 1:
/_next/static/chunks/app/profile-a1b2c3.js - Release 2:
/_next/static/chunks/app/profile-d4e5f6.js
If User A loaded your app during Release 1, their browser has the old manifest in memory. When they navigate to /profile five minutes after Release 2 is deployed:
- Browser requests
/_next/static/chunks/app/profile-a1b2c3.js. - Traditional deployments wipe Release 1 container/directory.
- Server returns
404 Not Found. - Browser throws fatal
ChunkLoadError, crashing user state.
NextDeploy's Pool Architecture
Instead of isolating static assets inside each individual release directory, NextDeploy routes all static requests to a shared cumulative directory:
/var/www/apps/my-app/
├── releases/
│ ├── 2026-11-20_v1/
│ └── 2026-11-20_v2/
└── shared_static/
├── chunks/
│ ├── app/profile-a1b2c3.js (retained from v1)
│ └── app/profile-d4e5f6.js (added in v2)
└── css/
├── 8f4a.css
└── c91e.css1. Cumulative Asset Merge
During each deployment, nextdeploy ship rsyncs all content-hashed static files into /var/www/apps/my-app/shared_static/. Existing files are never overwritten or deleted during the deploy.
2. Immediate NVMe Serving via Caddy
Caddy maps /_next/static/* directly to shared_static:
handle /_next/static/* {
root * /var/www/apps/my-app/shared_static
file_server
header Cache-Control "public, max-age=31536000, immutable"
}Because these files are immutable and content-hashed, Caddy serves them directly from disk with maximum caching headers.
3. Safe Garbage Collection
To prevent unlimited disk growth, nextdeployd implements an LRU (Least Recently Used) cleanup policy that runs as a background task. Chunks not accessed for 14 days and older than 10 deployments are safely pruned without impacting active sessions.