host_probes retention: prune rows >30d (table grows ~2.5k/day unbounded, 32.5k rows, v2.sqlite bloat driver)

Sub-tickets

Questions

Activity

• wi-cli-whey created · 2026-06-05
• wi-cli-whey assigned · 2026-06-05
  hub-llmmsgsrv-cc
• wi-cli-whey note · 2026-06-05
  Surfaced by nw-whey during DB triage. host_probes = 32,502 rows since 2026-05-23, ~2,500/day, no retention -> primary v2.sqlite growth driver (14,702 rows already >7d). Proposed: hub-side periodic prune deleting host_probes WHERE ts < now-30d (30d keeps #606-style outage analysis viable while bounding growth). hub owns hub.mjs+schema; pick mechanism (in-hub sweeper vs systemd timer) + window. Low pri.
• wi-cli-whey canceled · 2026-06-05
  Deprioritized per Elazar: host_probes is only 3.9MB (~10% of a 39.7MB DB), trivial; probes have active diagnostic value for ongoing lezama #606 single-path work. Pruning >30d would destroy the May23 baseline for marginal savings. Revisit only if table gets large (>100MB) or #606 closes and probes lose value.