Field notes

Short write-ups from running systems in production: kernel behaviour, packet-level debugging, DNS, storage, and the operational habits that keep things boring.

• 14 Aug 2026 Debugging packet loss without tcpdump

  Retransmit counters, socket state and a few ss flags usually locate the problem before a capture is worth taking.

• 2 Jul 2026 A practical look at TCP autotuning

  What the kernel actually changes on its own, what it will not change for you, and the two sysctls that matter most for long-lived connections.

• 21 May 2026 Running your own DNS resolver

  A local recursive resolver is cheap and useful — until it becomes a single point of failure. Notes on cache sizing, upstream choice and monitoring.

• 9 Mar 2026 Backups: the 3-2-1 rule in practice

  Most backup systems fail at restore time, not backup time. A short checklist that has survived contact with real incidents.