roadmap
- 01
- 02
- 03
- 04
- 05
- 06
- 07
- 08
Write Bash scripts that hold up in production and drive a Unix command line with confidence. Covers the shell language, pipes and redirection, error handling, processes, and everyday system tools.
241 of 241 questions in your plan: 0 known, 0 in progress, 0 due, 0 skipped, 241 new. 0 cut, outside your plan.
Shell questions belong to the operations side of hiring: Linux administration, SRE, DevOps and platform loops, plus backend teams that carry their own on-call. They take two forms: a short script to write or repair, in a screen or a take-home, and a troubleshooting conversation where the interviewer reports a symptom and waits for your next command.
The script questions turn on what the shell does to a line before it runs. for f in $(ls) splits a filename containing a space in two. A counter incremented inside cmd | while read reads zero after the loop, because bash ran the loop in a subshell. cmd >log 2>&1 and cmd 2>&1 >log send errors to different places. The troubleshooting questions are about the machine: why free shows almost no free memory on a healthy server, or why a job that works from your prompt fails under cron, with its minimal environment and short PATH.
Higher in the level range, the question becomes what happens when someone else runs the script. Which failures does set -e still let through, say in a function called inside an if? What must be cleaned up when the script is interrupted, and which signals can no trap catch? What does a file named -rf do to rm *? And when should a job leave Bash for a language with real data structures?
Before any scripting, the plan settles which shell is running. The interactive shell comes first, then POSIX sh and its relatives: on Debian and Ubuntu /bin/sh is dash, so a #!/bin/sh script that uses [[ or arrays fails there and passes where sh is bash.
The scripting module is the longest stretch, and everything after it leans on it. Bash as a language opens on expansion, because quoting, splitting and globbing decide what every later construct receives. Redirection and pipelines sit next to process control, which is where a lost assignment gets its explanation: a pipeline stage is a child process, and a child cannot change its parent's variables. Scripting practices close the module, once the bugs strict mode and a linter catch have been met.
Then the plan turns from the language to the system a script acts on. Mode bits and umask decide whether a script runs at all and who can read what it writes. Process monitoring, down to where memory has gone, comes before timers and cron, for the script that runs with nobody watching. The network module looks past one host: connectivity and DNS checks and socket state for a service nobody can reach, then rsync and SSH for work on other machines. Package management closes the plan, apt for Debian and dnf with its predecessor yum for Red Hat, because a provisioning script must know which family it is on.
If you live in the terminal but rarely write more than a few lines of script, the scripting module is the gap. Read the expansion topic as a page before drilling anything, because by frequency, where study order starts, can put a `pipefail` question ahead of the splitting rules that explain it. Administrators and SREs usually have the opposite profile: monitoring and networking describe their working day, and the follow-ups tend to land on scripting practices and process control. Rehearse the system half aloud at a real machine; watching `vmstat 1` while something writes to disk teaches more than its man page. Days out, narrow importance to its must know end, then treat each scripting question as a prediction: say which arguments the command receives before you reveal the answer, as a whiteboard round would ask. Disks and filesystems, and how the machine boots and supervises long-running services, get no module here; those belong to the [Linux roadmap](/roadmaps/linux). If the shell is one round in a wider operations loop, the roadmap for [DevOps and SRE engineers](/roadmaps/devops) places it beside containers, delivery pipelines and cloud infrastructure.
If you live in the terminal but rarely write more than a few lines of script, the scripting module is the gap. Read the expansion topic as a page before drilling anything, because by frequency, where study order starts, can put a pipefail question ahead of the splitting rules that explain it.
Administrators and SREs usually have the opposite profile: monitoring and networking describe their working day, and the follow-ups tend to land on scripting practices and process control. Rehearse the system half aloud at a real machine; watching vmstat 1 while something writes to disk teaches more than its man page.
Days out, narrow importance to its must know end, then treat each scripting question as a prediction: say which arguments the command receives before you reveal the answer, as a whiteboard round would ask.
Disks and filesystems, and how the machine boots and supervises long-running services, get no module here; those belong to the Linux roadmap. If the shell is one round in a wider operations loop, the roadmap for DevOps and SRE engineers places it beside containers, delivery pipelines and cloud infrastructure.