Daily Health Check — 2026-04-28

Findings

• [🔴] Gateway je noćas ponovno pao na Node.js heap OOM i restartao se iz core-dumpa — Journald u zadnjih 24h bilježi FATAL ERROR: Ineffective mark-compacts near heap limit Allocation failed - JavaScript heap out of memory u 01:38, zatim openclaw-gateway.service: Failed with result 'core-dump', pa novi start gatewaya u 01:39. Ovo je nova visoko-prioritetna runtime regresija. (severity: high)
• [🔴] 🔄 Cron Health Monitor je u error stanju zbog timeouta — openclaw cron list --json pokazuje lastStatus: "error", consecutiveErrors: 1, lastErrorReason: "timeout" za ovaj job. Nije masovni cron kvar, ali postoji svježi failing cron. (severity: high)
• [🟡] daily_logs je i dalje jedini sadržajni memory bucket pod pritiskom — memory-health-breakdown.py --json označava samo daily_logs kao warn; oversized su 2026-04-21.md (40.4KB), 2026-04-22.md (15.8KB) i 2026-04-26.md (23.5KB). hot_memory, audit, archive i plans ostaju zdravi. (severity: medium)
• [🟡] runtime_state (memory/.dreams) ostaje iznad praga — breakdown vraća 401781 B naspram 307200 B warning praga. Problem je lokaliziran na runtime state bucket, bez širenja na ostale memory zone. (severity: medium)

Previous Findings Update

• [Gateway heap OOM / core-dump restart] 🔴→🔴 IN_PROGRESS (jučerašnji problem nije riješen; dogodio se novi OOM restart u 01:38.)
• [Dnevni Quick Health Check cron timeout] 🔴→🟢 RESOLVED (jučer failing job je danas ponovno lastStatus: ok; trenutni failing cron je drugi job: 🔄 Cron Health Monitor.)
• [daily_logs bucket pressure] 🟡→🟡 IN_PROGRESS (daily_logs ostaje warn s istim oversized clusterom.)
• [runtime_state over threshold] 🟡→🟡 IN_PROGRESS (memory/.dreams i dalje prelazi prag.)
• [Config drift watch] 🟢→🟢 STABLE (nema ctime promjena u AGENTS.md, TOOLS.md, DECISIONS.md, SOUL.md ni kanonskom openclaw.json u zadnjih 24h.)

Summary

4 new findings (0 critical, 2 high). Token cost: ~minimal.