new-usemame/Calibre-Web-NextGen v4.0.86

v4.0.86 — container actually reports unhealthy when ingest/metadata services die

on GitHub

What's new

docker pull ghcr.io/new-usemame/calibre-web-nextgen:v4.0.86

(or :latest)

Container healthcheck no longer reports "healthy" when ingest or metadata write-back is silently broken

Previously, the container could report healthy to Docker / Kubernetes / Compose while:

• cwa-ingest-service had crashed and books dropped in /cwa-book-ingest/ weren't being picked up, OR
• metadata-change-detector had crashed and metadata edits in the UI weren't propagating back to your book files, OR
• the web worker was wedged by a heavy in-process job (large library duplicate scan, sustained ingest under load) — /health couldn't respond, but Docker's outer probe timeout sometimes left curl processes lingering instead of cleanly marking the probe failed.

/health now probes the s6 longruns that user features depend on, and the Docker HEALTHCHECK curl is bounded so a wedged web worker surfaces as a failed probe within two seconds. Reverse-proxy keepalive sockets also can't accumulate against the gevent process anymore.

Body shape now includes a services map:

{
  "status": "ok",
  "uptime": 12345.6,
  "version": "Calibre-Web-NextGen/v4.0.86",
  "services": {
    "cwa-ingest-service": "up",
    "metadata-change-detector": "up"
  }
}

Any service reporting "down" flips the response to HTTP 503 with "status": "degraded", so autoheal / k8s liveness / Compose can restart the container.

Credits

• Reporters: @droM4X (fork #193) and @FRaccie (~200k-book production trace for the wedged-worker symptom).
• Backport: @I-Would-Like-To-Report-A-Bug-Please (CWA #1335) — bounded healthcheck curl + gevent connection close.