Page 4 - Linux

Alongside the system clock the kernel keeps in memory, almost every physical and virtual machine has a battery backed hardware clock that keeps running even while the machine is powered off. Anyone who does not know the difference between the system clock and the hardware clock will sooner or later be puzzled by wrong boot timestamps, mistimed certificate checks, or a dual boot system that shows a different time after every Windows startup. This article explains how hwclock works, how to set the hardware clock correctly, and how to keep it reliably synchronized with the system clock.

High system load is not automatically a sign of a slow disk. This guide shows how iostat with interval sampling separates real trends from short bursts, how iotop identifies the specific process behind heavy I/O usage, and how metrics like %util and await reliably reveal whether you are looking at a disk problem or an application problem.

A script using mapfile or sed -i runs fine on the Linux server and breaks on a colleague's MacBook. The reason is the outdated Bash 3.2 that ships with macOS and the different coreutils Apple provides for licensing reasons.

Planning server storage around rigid partitions runs into limits the moment you need to grow. LVM decouples logical volumes from the physical disk layout and lets you resize, snapshot, and add new disks while the application keeps running, with no restart required.

When a PHP-FPM worker seems to hang for no reason, or an application throws mysterious file errors, application logic often stops providing useful clues. strace exposes exactly what a process is telling the kernel, syscall by syscall, and often reveals within seconds what a process is actually stuck on.

Grouping a log file with a hundred thousand lines by status code, calculating totals per customer, or building a formatted report from raw CSV: awk does all of that without Python, without a database, right inside the pipe. Fields, associative arrays and BEGIN/END blocks turn it into a full featured reporting tool.

Anyone connecting via SSH to the same server ten or twenty times in a row inside a deployment script pays the full price of a TCP handshake, key exchange and authentication every single time. ControlMaster shares a single established SSH connection over a local Unix socket with every further call, making Bash automation noticeably faster.

The EMFILE error, reported as "Too many open files", is rarely a coincidence and almost never a pure capacity issue. Usually an application opens connections, files, or sockets and never closes them again. This guide shows how to use lsof, the /proc/pid/fd directory, and targeted monitoring to find out which code path is leaking file descriptors and how to configure the limits correctly without papering over the actual leak.

Bash scripts that run without locking in cron jobs or deployment pipelines risk race conditions when executed more than once. PID files are common but not atomic. flock offers kernel backed locks that are released automatically even after a crash, making it the right tool for reliable locking in shell scripts.

An OnCalendar expression looks like a simple date at first glance but can combine weekdays, ranges, lists and repeats in a single, compact line. Anyone who does not really understand the syntax copies examples from the internet and then wonders why the timer never fires or fires at the wrong time. This article explains the OnCalendar syntax systematically, from simple shorthands to complex expressions, and shows how to reliably test every expression before deployment with systemd-analyze calendar.