Puca (Radio and TV Archiver)

Puca is a Raspberry Pi (10.7.160.28) that continuously records Irish-language radio and TV for the lab's speech corpus.

Configured by ansible/puca/setup-puca.yml. Unlike every other host, it isn't managed via the standard ProxyJump pattern — it's administered from a separate control node (setanta), which reaches it directly by SSH.

Radio recording

Five stations are recorded continuously, each as its own systemd service instance: RTÉ Raidió na Gaeltachta, RTÉ Radio 1, RTÉ 2FM, Raidió na Life, and Newstalk. The recording engine (rte_recorder.py, from the lab's private Radio-Archivist repo) runs as the ansible-control user, writing .m4a/.mp3 files to local storage.

TV recording

Three Saorview channels — TG4, RTÉ One, and Cúla4 — are recorded 24/7 via a USB DVB-T tuner and tvheadend, transcoded audio-only (video dropped, audio and subtitles kept) to save space.

Storage

Recordings are written to a local external drive first (a live DVB-T capture can't tolerate an NFS stall), then synced hourly to the lab's NAS. A daily prune job keeps roughly 300 GB of local retention, only deleting files already confirmed present on the NAS.

Clock sync

The Pi has no RTC and the lab network blocks NTP, so the clock is set at boot (and re-synced daily) via htpdate against a few public HTTPS endpoints, ordered to run before tvheadend starts — starting with a stale clock causes recordings to be mis-dated against the EPG.

Monitoring

Rather than a database or per-file probing, Puca infers whether a recording succeeded from its file size (a reliable proxy for duration on near-constant-bitrate audio) compared against the published schedule, classifying each as OK / partial / short / empty. Host metrics and full system logs ship to a separate, publicly-reachable Grafana/Loki/Prometheus endpoint (corais.abair.ie) rather than the fleet's observability VM.