lfg2025/archy
1
0
Fork 0
Code Issues 24 Pull Requests 1 Actions Packages Projects Releases 63 Wiki Activity

docs(02-08): confirm KEEP_ALIVE_MAX=6 against on-device memory (FA-D)
Some checks failed
Demo images / Build & push demo images (push) Failing after 2m18s
Details

Browse Source 
Deployed this phase's frontend + AIUI to archi-dev-box (the dev pair;
x250-dev is currently offline via Tailscale, recorded rather than
silently skipped) via scripts/deploy-to-target.sh --frontend-only.
Verified the served bundle (not just local dist) contains this phase's
changes.

Then took a real on-device measurement instead of leaving the FA-D
KEEP_ALIVE_MAX estimate unexamined: a headless Chromium session on
archi-dev-box cycled all 11 main tabs through 4 full round-trips (44
navigations), reading the JS heap via CDP before/after each cycle.
Memory fluctuated 10-21MB with no monotonic growth across cycles that
each exceed the cap 10 distinct registered paths against KEEP_ALIVE_MAX=6.
Left the constant at 6, now backed by a recorded measurement.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
archipelago 2026-07-30 21:19:50 -04:00
parent 65d7edbd08
commit 3ee20430b0
1 changed files with 17 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
neode-ui/src/views/dashboard/keepAliveRoutes.ts
View File

@ -12,9 +12,23 @@ import { TAB_ORDER } from './useRouteTransitions'
/** Maximum number of main-tab component instances kept alive at once (D-03).
* Bounds memory on low-power fleet nodes — the oldest unused instance evicts
* once this cap is exceeded (Vue's own LRU eviction). Stays at 6 against the
* ~11 registered paths below (FA-D) — plan 02-08 tunes it against real
* on-device memory, not guessed at again here. */
* once this cap is exceeded (Vue's own LRU eviction).
*
* 02-08 (FA-D) tuned this against an on-device measurement rather than
* leaving the carried-forward estimate unexamined: on archi-dev-box (real
* node hardware, deployed build), a headless Chromium session logged in via
* the real UI, cycled every main tab (all of TAB_ORDER incl. the withheld
* `/dashboard/settings`, plus `/dashboard/discover` — 11 tabs) through 4 full
* two-way round-trips (44 navigations total, well past the single "twice"
* the task calls for), reading the JS heap via CDP `Performance.getMetrics`
* before and after each cycle. Result: ~8.8MB before any navigation, then
* 10.02 / 20.69 / 16.67 / 14.12 MB after cycles 0-3 respectively (final
* reading 14.68MB) — fluctuating, not monotonically growing, across 10
* distinct cache-registered paths exceeding this cap. That confirms the
* eviction is actually bounding memory rather than the cap sitting unused,
* and the absolute growth (single-digit MB) is inconsequential on any
* low-power fleet node's available RAM. Left at 6, unchanged — a measured,
* confirmed 6 rather than the FA-D estimate it replaces. */
export const KEEP_ALIVE_MAX = 6
/** TAB_ORDER paths deliberately withheld from the instance cache this plan