This thread has everything. OOM killers, SD card wear, RPKI, and now cron scheduling. I just want to say that if your monitoring granularity is 1-minute pings, your 99.94% vs 99.97% discrepancy is probably four failed checks during a single restart. Not everything needs a postmortem. Sometimes a box just sneezes.
Member
Jajaja I use acme.sh in my router, is work very good. The flash is only 4MB but I put in tmpfs and download when need. The certificate is for my DDNS domain, free from DuckDNS. Maybe is not same like big server but the principle is same. Small software for small machine.
Expected. From Falkenstein you're ~7ms to Hetzner's own anycast resolvers if you peer direct, but Cloudflare FRA is across DE-CIX with congested ports some hours. For the 12MB savings I'd still take it on 128MB, but on 256MB the hot cache wins. This is why I said mitigation, not fix.
Heresy. Postmortems are how we learn. Though in my day we called them "staring at the blinkenlights until 3am." Same outcome, worse lighting.
