Arr apps1.x

Radarr and Sonarr credentials and automatic stuck-queue handling. Config key: arr.instances (NZBDAV_CONFIG__ARR__INSTANCES; see headless).

Setting	What it does
Name	Optional display name; Overview Arr Health falls back to the host URL.
Enabled	When off, the instance is excluded from queue management, Arr-linked repairs and Arr Health. Default on.
Host / API Key	One pair per Radarr or Sonarr instance. Test Conn appears once both are filled.
Show Arr Health on Overview arr.health-enabled	Polls enabled instances for the Overview widget (NZBDAV_CONFIG__ARR__HEALTH_ENABLED). Default on.
Automatic replacement searches / Per window (minutes)	Replacement-search limit per movie or episode (1–10 searches, 1–1440 minutes). Default 3 in 30 minutes.
Automatic queue management	Per status message: Do Nothing, Remove, Remove and Blocklist, or Remove, Blocklist, and Search. Default Do Nothing.

Test Conn validates host reachability, API key auth, and a successful API response (GET /api). It works with already-saved (masked) API keys — you do not need to re-enter the key after reload. Failures show a reason (authentication, HTTP status, or network error).

Only Usenet queue items that Radarr or Sonarr reports as completed or awaiting import are acted upon. Downloads that are still queued or downloading are never removed by these rules. Typical mappings:

• Remove, Blocklist, and Search — no eligible files, samples, no audio tracks
• Remove and Blocklist — not an upgrade / custom format
• Remove — already imported
• Do Nothing — ID mismatches and similar manual-import cases

Any action other than Do Nothing tells the Arr to delete the queue record with removeFromClient=true. The Arr then removes the download from infinidysk History even when its own Remove Completed checkbox is off. That is independent of mounted files, which stay.

Docker hostnames and proxies

Radarr, Sonarr, and Prowlarr hosts whose hostname has no dot — Docker Compose service names such as http://sonarr:8989 — are always contacted directly, even when the container has HTTP_PROXY / HTTPS_PROXY set. Names like that only resolve on the local container network, so sending them through a proxy produced connection failures that disappeared when the host was changed to an IP address. Dotted hostnames and IP literals keep the normal proxy policy, including NO_PROXY. This applies only to infinidysk’s own Arr and Prowlarr management calls; imported indexer traffic still follows the per-indexer proxy settings.

Connections to Arr and Prowlarr hosts are recycled after two minutes so a stale DNS answer or a recreated container cannot pin a dead connection.

Request budgets are unchanged: Arr health polls allow 10 seconds per call, other Radarr/Sonarr requests 30 seconds, and Prowlarr requests 15 seconds. When one expires, logs, Test Conn, and the Overview Arr Health panel report it as a timeout with the routing used, for example:

Arr health poll failed for "http://sonarr:8989". Reason: "Sonarr queue
request to http://sonarr:8989 timed out after 10 seconds; routing: direct
(single-label hostname)."

“Operation canceled” is no longer used for expired waits; it now indicates only a genuine shutdown or caller cancellation.

Releases stuck in Awaiting import

Sometimes Sonarr completes a download but refuses to import it because the season numbering in the release name disagrees with Sonarr’s metadata. A representative queue reason is:

Episode 11x01 was unexpected considering the <release-name> folder name

Sonarr determines episode identity and import eligibility. infinidysk cannot remap the episode or force Sonarr to import it, but it can automatically resolve the stuck queue record.

In Settings → Arr Apps → Automatic queue management, choose an action for Episode(s) was/were unexpected considering the folder name:

• Remove removes the stuck queue record.
• Remove and Blocklist also prevents Sonarr from accepting that release again.
• Remove, Blocklist, and Search also asks Sonarr to find a replacement.

All queue rules default to Do Nothing. With that default, the item remains in Awaiting Import until an operator resolves it. The Overview Arr Health panel reports unusually long waits but does not mutate the queue.

Use Remove, Blocklist, and Search when an automatic replacement is wanted. A numbering disagreement can affect every indexed release for the episode; the per-media replacement-search limit eventually withholds additional searches while continuing to remove and blocklist rejected releases.

Replacement-search safety limit since 1.2.4

Remove, Blocklist, and Search is bounded to 3 automatic replacement searches in 30 minutes per Radarr movie or Sonarr episode by default. Once a media item reaches that limit, infinidysk still removes and blocklists its rejected release but does not start another automatic search. Adjust both values in Automatic queue management when a library needs a stricter or more permissive policy. The same per-media cap also bounds replacement searches triggered by health-check repairs, which additionally keep their own per-library-file rate limit.

The queue-monitor warning includes the matching Arr import message, so support packs show the actual rejection (for example, an archive, an ineligible file, or an import path problem) that led to the action.

Disabling an instance opts it out of stuck-queue actions, Arr-linked repairs, and health polling. Turn Show Arr Health on Overview off when you still want queue rules without Overview polling.

Blocklisted re-grab protection since 1.4.0

When a queue rule applies Remove and Blocklist or Remove, Blocklist, and Search, infinidysk remembers a bounded sample of article IDs from the removed download. A later NZB containing any remembered article is failed in the download queue before import or Usenet article requests, even when another indexer supplied it under a different release name. Plain Remove does not mark an upload as rejected.

This memory is process-local and bounded. Evidence capture is also limited to two seconds per monitoring pass so queue actions are not delayed by local evidence collection. It is lost on restart, entries can be evicted, and provider configuration changes can clear it. Protection also requires the completed download’s local history and stored NZB to still exist. A replacement that starts importing before Arr’s delete request returns can pass the check; later re-grabs are protected after Arr confirms the blocklist action succeeded.

Arr Health on the Overview dashboard since 1.2.0

When at least one Radarr or Sonarr instance is configured and enabled, Overview shows a compact Arr Health section: instance reachability, imports in the selected dashboard window, median/P95 handoff latency (infinidysk download completed → Arr DownloadFolderImported), queue depth, and items waiting unusually long for import.

Awaiting detail since 1.2.5

Queue is every item Arr currently tracks. Awaiting is the completed or importPending/importing subset, so the two counts intentionally differ. The widget reports the true Awaiting total while showing only its longest waits; each displayed row includes the Arr status reason when one is available. Related season-pack rows with the same download are grouped in the detail list.

The feature is completely dormant otherwise:

• No Overview widget
• No background polling or Arr HTTP
• No metrics writes or Arr Health log lines

Turn off Show Arr Health on Overview (arr.health-enabled) to keep instances for queue rules/repairs without health polling.

Handoff events are stored in the local metrics database (metrics.sqlite) for 90 days, so the Overview All window for this widget is effectively about 90 days. API keys are never sent to the widget.

Back up /config before upgrading: the metrics store gains an additive ArrImportEvents table that auto-applies on startup.

Headless NZBDAV_CONFIG__ARR__INSTANCES JSON that includes the new Name / Enabled properties cannot roll back to an older image (structured ENV config is not forward compatible — see headless).

Connect Arr