bug description:
When the redis/valkey backend restarts (crash, OOM, pod reschedule, planned maintenance), the standalone server never reconnects.
Cap creates a single Bun RedisClient in standalone/src/db.js and relies on its autoReconnect. That reconnect logic doesn't appear to be stable in Bun:
Since cap has no reconnect/recreate handling of its own, the broken client is kept forever.
affects:
to reproduce:
- Run standalone (
tiago2/cap:latest) against redis or valkey, e.g. via the bundled docker-compose.yml or in kubernetes.
- Create a site key and verify
POST /<siteKey>/challenge returns 200.
- Restart the redis server (
docker restart cap-valkey / delete the redis pod). It is back and accepting connections within ~2 seconds.
- Keep requesting
POST /<siteKey>/challenge.
Result: after one ERR_REDIS_CONNECTION_TIMEOUT, every subsequent request fails with ERR_REDIS_CONNECTION_CLOSED — observed for 5+ minutes (151 consecutive failures) with no recovery, while a freshly created RedisClient from inside the same container connects and gets PONG immediately. Restarting cap fixes it instantly.
expected behavior:
After redis is reachable again, cap should reconnect and continue serving requests without a process restart.
screenshots:
API response after the redis restart (stays like this forever):
{"success":false,"error":"ERR_REDIS_CONNECTION_CLOSED","detail": "troubleshooting":"http://trycap.dev/guide/standalone/..."}}Log output — one connection timeout at the moment redis goes away, then an endless stream of ERR_REDIS_CONNECTION_CLOSED for every request:
[ERR_REDIS_CONNECTION_TIMEOUT 0e1999447f4a] {"timestamp":"2026-08-13T13:20:07.092Z","error":{"name":"RedisError","message":"Connection timeout reached after 10000ms","code":"ERR_REDIS_CONNECTION_TIMEOUT"},"env":{"bun":"1.3.14","platform":"linux",...}}
[ERR_REDIS_CONNECTION_CLOSED c383e2cc0033] {"timestamp":"2026-08-13T13:20:09.116Z","error":{"name":"RedisError","message":"Connection has failed","stack":"RedisError: Connection has failed\n at hget (unknown)\n at <anonymous> (/usr/src/app/src/cap.js:209:38)\n at <anonymous> (/usr/src/app/src/ratelimit.js:70:30)\n at handle (file:///usr/src/app/node_modules/elysia/dist/compose.mjs:10:27)\n at map (file:///usr/src/app/node_modules/elysia/dist/adapter/bun/compose.mjs:9:67)\n at processTicksAndRejections (native:7:39)","code":"ERR_REDIS_CONNECTION_CLOSED"},"env":{"bun":"1.3.14","platform":"linux",...}}
[ERR_REDIS_CONNECTION_CLOSED 5765415ffbb8] {"timestamp":"2026-08-13T13:20:11.147Z","error":{"name":"RedisError","message":"Connection has failed", ...}}
[cap] RSW keypair refresh failed: Connection has failed
[ERR_REDIS_CONNECTION_CLOSED 82cdaf1aa496] ... (repeats for every request until cap is restarted)
versions and environment:
cap standalone 3.1.9 (tiago2/cap:latest), Bun 1.3.14, redis:7-alpine
bug description:
When the redis/valkey backend restarts (crash, OOM, pod reschedule, planned maintenance), the standalone server never reconnects.
Cap creates a single Bun
RedisClientinstandalone/src/db.jsand relies on itsautoReconnect. That reconnect logic doesn't appear to be stable in Bun:Since cap has no reconnect/recreate handling of its own, the broken client is kept forever.
affects:
to reproduce:
tiago2/cap:latest) against redis or valkey, e.g. via the bundleddocker-compose.ymlor in kubernetes.POST /<siteKey>/challengereturns 200.docker restart cap-valkey/ delete the redis pod). It is back and accepting connections within ~2 seconds.POST /<siteKey>/challenge.Result: after one
ERR_REDIS_CONNECTION_TIMEOUT, every subsequent request fails withERR_REDIS_CONNECTION_CLOSED— observed for 5+ minutes (151 consecutive failures) with no recovery, while a freshly createdRedisClientfrom inside the same container connects and getsPONGimmediately. Restarting cap fixes it instantly.expected behavior:
After redis is reachable again, cap should reconnect and continue serving requests without a process restart.
screenshots:
API response after the redis restart (stays like this forever):
{"success":false,"error":"ERR_REDIS_CONNECTION_CLOSED","detail": "troubleshooting":"http://trycap.dev/guide/standalone/..."}}Log output — one connection timeout at the moment redis goes away, then an endless stream of
ERR_REDIS_CONNECTION_CLOSEDfor every request:versions and environment:
cap standalone 3.1.9 (
tiago2/cap:latest), Bun 1.3.14, redis:7-alpine