AI & Agent Dev Bug Sandbox Back to Radar

Default "Use Cache" Handler Pins Buffer Pool Slabs Causing Memory To Exceed CacheMaxMemorySize By Up To 22X On Node 26

The default `use cache` handler stores `streamToBuffer` results that may be small slices of a shared Buffer pool slab. Long-lived cache entries therefore retain whole pool slabs (64 KiB on Node 26) while the LRU only counts the slice's bytes, causing actual memory usage to exceed `cacheMaxMemorySize` by up to 22x.

highConfidence 95%Next.js Affected V16.4.0

Origin Analysis

In `dist/server/lib/cache-handlers/default.js`, `set()` calls `streamToBuffer(entry.value)` and stores the returned Buffer. When total bytes are below `Buffer.poolSize >>> 1`, `Buffer.concat` returns a slice of a shared pool slab; retaining this slice keeps the entire slab alive, so the LRU's `value.byteLength` accounting vastly underestimates real retained memory.

1. Run Node.js 26.7.0. 2. Build and start Next.js with `--expose-gc` using the reproduction repo. 3. Generate 10,000 cache entries of ~1 KB via `/item/<n>` routes. 4. Force GC and read `/api/memory`; compare ArrayBuffers after GC: 221 MB with 64 KiB pool vs 42 MB with 8 KiB pool vs 10 MB after applying unpooled copy.

Fixing Code Block

0x7F 45 4C 46 02 01 01 000x90 12 AB 04 C1 7E 33 08
0xA4 1C 90 3E 77 0B D2 580x11 F6 6A 2C 80 19 DE 44
0x00 00 00 00 03 00 3E 000x01 00 B7 4D 55 90 0C 61
0x6C 69 62 63 2E 73 6F 2E0x36 00 8E 17 42 AA 05 F3
0xDE C0 DE 01 8B 45 FC 890xC7 5D 20 6F 91 03 E4 78
0x13 37 00 4F 9A 2D E1 700x08 B4 5C 22 71 D0 4A 16
0x48 8D 3D 22 01 00 00 E80x91 FF 0A 63 2E 57 B8 04
0x5A 91 C3 0E 44 70 1B 880x02 6F 39 C4 15 80 ED 27
0xFF 15 40 20 60 00 48 890xC3 90 7A 01 4C 28 B1 56
0x0F 1F 44 00 00 55 48 890xE5 C3 18 6D 92 0B 33 A7

Edge Case Audit

The unconditional copy adds an extra allocation and copy for every entry; for entries larger than half the Buffer pool size, `streamToBuffer` already returns a non-pooled exact-size buffer, so this adds avoidable peak memory and CPU overhead. In high-throughput cache writes, prefer conditionally copying only when `pooled.byteLength < pooled.buffer.byteLength` (or checking `pooled.buffer.byteLength !== pooled.byteLength`) to limit overhead. Rollback: keep the original `const value = await streamToBuffer(entry.value);` line and remove the copy; affected deployments can temporarily set `Buffer.poolSize = 8192` to reduce the gap. Also note the fix must remain in the transpiled `dist` output used by production builds; ensure source changes are rebuilt or patched in both source and dist.