Page 16 - Linux

When a Linux server resolves domain names sometimes and not other times, the cause is rarely the DNS server itself but the local resolution chain in front of it: resolv.conf, systemd-resolved, nsswitch.conf, or a stubborn cache. This guide walks step by step through the correct order of checks, from the configuration file to the actual application runtime, and shows why a successful dig call alone proves nothing.

A single network port is a structural single point of failure on production servers, no matter how reliable the rest of the hardware is. Linux bonding combines several physical interfaces into one logical bond, delivering, depending on the chosen mode, failover protection, aggregated bandwidth, or both at once.

The official Magento system requirements list only names version numbers, yet a great deal of configuration work on the Linux server sits between those numbers and a working shop. This article translates PHP extensions, memory_limit, OpenSearch, Redis and Varnish into concrete commands and configuration files, and closes with a pre-flight checklist that catches setup mistakes before the first deployment.

A standalone Bash CLI tool distributed without a package manager needs its own way to stay current on users' machines. A self-update mechanism downloads new versions, verifies them, replaces the running script atomically, and rolls the change back on failure before users even notice anything went wrong.

As soon as a Bash script reads configuration from more than one source, the load order determines its behavior in production. Scripts that do not deliberately prioritize system environment, .env file and command-line overrides tend to work fine on one machine and silently pick up the wrong value on a colleague's laptop or in a CI runner.

A backup that has never actually been restored is an unproven assumption, not a reliable safety net. Automated restore tests regularly check whether a backup really works, whether the data is complete, and how long an actual restore takes. This article shows how to establish restore tests for Linux servers as a fixed, monitored part of operations.

When top only shows that the CPU is busy but not with what, classic monitoring stops being useful. Kernel tracing with ftrace and perf makes visible which function inside the kernel runs for how long, which syscall is blocking, and exactly where the time in a seemingly simple request disappears to.

An unclean shutdown or a power outage is enough to leave a filesystem with inconsistent metadata. fsck detects and repairs such errors, but only if you follow the workflow, unmounting, checking read-only, and repairing only afterward, in the right order.

The Linux kernel maintains not one but several clocks at the same time, and picking the wrong one for the wrong job leads to bugs that only surface during an NTP correction or a manual time jump. This guide explains the difference between CLOCK_REALTIME and CLOCK_MONOTONIC, why timeouts and retry logic should avoid the wall clock, and how scripts stay robust against time jumps.

An unencrypted backup is a second copy of the most sensitive data in a company, often protected even less than the original. Backup encryption is therefore not a nice-to-have, but a basic requirement as soon as a backup leaves your own server. This article shows how GPG, OpenSSL, restic, and Borg are used on Linux for encrypted backups, and what to watch for in key management.