fix(geocoding): bump PROVIDER_TIMEOUT_MS to 20s for cold cross-LAN

Cold-start fetches from the mana-geocoding container to photon-self
on mana-gpu (over WSL2 mirrored networking) consistently take >10s on
the first probe and ~2s once warm. The previous 8s default caused the
chain to false-mark photon-self unhealthy on every cold path, leaking
to public photon for the next 30s health-cache window — and pinning
the public-photon answer in the 7d cache (now shortened to 1h).

Also wires the docker-compose macmini env to honor PROVIDER_TIMEOUT_MS
and CACHE_PUBLIC_TTL_MS overrides so production picks up the new
values without a code rebuild.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
Till JS 2026-04-28 22:19:21 +02:00
parent 962606b961
commit 8a5fad34df
3 changed files with 18 additions and 6 deletions

View file

@ -510,6 +510,14 @@ services:
# as `photon-self` provider with privacy: 'local' — eligible for
# sensitive queries. Empty value = slot disabled.
PHOTON_SELF_API_URL: ${PHOTON_SELF_API_URL:-}
# Cold-start cross-LAN fetches to photon-self consistently take
# >10s on the first probe; the 8s default false-marked it unhealthy
# on every cold path. 20s leaves headroom while still cutting off
# actually-stuck connections.
PROVIDER_TIMEOUT_MS: ${PROVIDER_TIMEOUT_MS:-20000}
# Short public-API cache TTL so a transient photon-self blip can't
# pin stale public-fallback answers in the LRU for days.
CACHE_PUBLIC_TTL_MS: ${CACHE_PUBLIC_TTL_MS:-3600000}
CORS_ORIGINS: https://mana.how,http://localhost:5173
CACHE_MAX_ENTRIES: "5000"
CACHE_TTL_MS: "86400000"