UnraidUnraid
Unraid OS Bug Reports

[7.4.0] Docker update check shows "not available" for all lscr.io images (auth token dropped on 302 redirect to ghcr.io)

A
arcanineinUnraid OS Bug Reports
8 days ago

Expected Behavior

The Docker update check shows "not available" for every container using an lscr.io image (LinuxServer.io). Containers from Docker Hub and ghcr.io check correctly. Cause: with Unraid's user agent, lscr.io answers the authenticated manifest request with a 302 redirect to ghcr.io. The Authorization header is not forwarded to the new host, so ghcr.io returns 401 with no Docker-Content-Digest header, and getRemoteVersionV2() returns null.

Actual Behavior

The Update column shows "not available" for every lscr.io image. The status file records "remote": null, "status": "undef" for each.

docker pull and docker manifest inspect work for the same images; network, DNS and gateway are fine.

Manual test of Unraid's request sequence:

- curl with Unraid's user agent (docker/Unraid os/Docker-Client) and an Authorization header: lscr.io returns 302 to https://ghcr.io/v2/linuxserver/radarr/manifests/latest, which returns 401 (token dropped on the cross-host redirect).

- Same request with --location-trusted (credentials sent on redirect): 302, then 200 with a docker-content-digest header.

Steps to Reproduce

1. Run any container from an lscr.io/linuxserver/* image (radarr, sonarr, prowlarr, calibre-web, duplicati, etc.).

2. Docker tab -> Check for Updates.

3. Check /var/lib/docker/unraid-update-status.json for the image.

Additional Context

Environment:

- Unraid 7.4.0

- About 25 Docker containers; the 9 using lscr.io images are affected: radarr, sonarr, prowlarr, calibre, calibre-web, lazylibrarian, duplicati, nextcloud, duckdns

- Containers using Docker Hub or ghcr.io images (Jellyfin, Cleanuparr, open-webui, shelfmark, tdarr, etc.) check correctly

What I ruled out:

- Network, DNS and gateway (docker pull and docker manifest inspect work for the same images)

- Docker Hub rate limiting (ratelimit-remaining showed the full allowance)

- Stale status data (I cleared /var/lib/docker/unraid-update-status.json and re-ran dockerupdate check; the result is still "remote": null, "status": "undef" until the workaround is applied)

Verification of the workaround:

- Before: "remote": null, "status": "undef"

- After adding CURLOPT_UNRESTRICTED_AUTH: remote digest populated (sha256:51a9ac2a...) and status "false" (update available) for radarr, which matches the registry

Related report of the same symptom:

https://forums.unraid.net/topic/186325-updates-not-available-for-lscrio-containers/

(that thread was on 7.1.4, so this has existed across several releases)

Diagnostics (anonymized) attached.

Workarounds

Workaround that fixes it: add

curl_setopt($ch, CURLOPT_UNRESTRICTED_AUTH, true);

after the CURLOPT_FOLLOWLOCATION line in getCurlHandle() in

/usr/local/emhttp/plugins/dynamix.docker.manager/include/Helpers.php

After that, remote digests are populated and statuses show correctly. The patch is lost on reboot because /usr/local/emhttp is in RAM, so I reapply it with a User Scripts job at array start.

Suggested fix: keep the token on the redirect, or request a new token for the redirect target host.

Older forum report of the same symptom:

https://forums.unraid.net/topic/186325-updates-not-available-for-lscrio-containers/

Screenshots / Videos / Diagnostics

Diagnostics attached (anonymized).

Closed
Loading content, please wait...
Loading content, please wait...
Loading content, please wait...
Loading content, please wait...
Loading content, please wait...
Loading content, please wait...
Loading content, please wait...
Loading content, please wait...
Loading content, please wait...
Loading content, please wait...
Loading content, please wait...
Loading content, please wait...