Field archive

Notes

Commands, failure boundaries, and verification steps for small Linux systems. Examples stay generic; conclusions stay close to observable state.

13 notes

Networking4 min read

A complete port map names the transport, address family, bind address, and process.

Networking5 min read

Use local forwarding for a loopback-only service, then verify a fresh connection instead of trusting an old tunnel.

Networking5 min read

A sysctl is only a default. Prove the active socket and compare distributions, not one peak result.

Networking5 min read

Build the policy around established traffic, explicit transports, and a timed rollback.

Networking5 min read

Plan around the TTL, query each resolver layer, and keep an overlap window for rollback.

Linux4 min read

systemd can restart a crashed process, but only when exit semantics and rate limits are understood.

Linux5 min read

A retention drop-in is incomplete until a new record can be read from the persistent journal.

Linux5 min read

Keep a second session, capture state, arm a rollback, and separate configuration validity from service health.

Linux5 min read

Check synchronization state, not daemon liveness, and let restored virtual machines step once.

Linux5 min read

Track bytes and inodes separately, respect read-only remounts, and find growth early.

Operations5 min read

Use SQLite's backup API, verify an independent database, and rehearse restore outside production.

Operations5 min read

Confirm file ownership, validate the candidate, reload narrowly, and compare the served certificate.

Operations5 min read

Expand the calendar expression, read the journal receipt, and verify the effect on disk.

All notes