Every cohort of learners eventually reaches the same awkward moment. The pipelines work, the containers start, and then someone asks why the build server ran out of disk on a Friday, or why a scheduled job never fired. The answer lives one layer down, in systems operations, usually called SysOps. We collected the questions students ask most often and answered them in a conversational way, borrowing facts from the Ops Atlas entry at xOps School on SysOps wherever a claim needs a source.
What exactly is SysOps?
It is the care of individual servers: which services run, who may do what, what the logs say, which jobs fire on a schedule, which packages are installed, how disks fill and how clocks stay right, plus the first aid of working out why a machine is slow. The Atlas describes Linux as the main target and gives Windows Server a lighter touch. It also admits the label is slippery. Nobody formally defines it, and it sits alongside system administration, so employers use both words for much the same work.
Is it still relevant now that everything runs in containers and the cloud?
Yes, with a caveat about what “relevant” means. Containers run on hosts, and those hosts have kernels, clocks, disks, limits and updates. On rented virtual machines the provider handles the data centre while the guest system, with its patches and accounts, is the customer’s. The Atlas calls the idea that the cloud needs no SysOps false for exactly this reason. Where the work does change is in how it is applied: through configuration tools and images more than through hand edits.
Which certification should I take first?
For hands-on host skills, the Atlas names two long-standing credentials: LFCS from the Linux Foundation and RHCSA from Red Hat. Each one puts you in front of a live machine and asks you to perform tasks. A third name you may have heard, the cloud provider’s SysOps Administrator associate, no longer exists under that title. It became CloudOps Engineer, and the last sitting of the previous exam code was on 29 September 2025. Pick based on the jobs you are targeting, and let a practical exam reward the labs you have already done.
How much Linux is enough to start?
Enough to be dangerous in a lab and cautious on production. In practice that means managing services with systemd, reading logs, handling permissions, installing and removing packages, editing the remote-login configuration safely and understanding mounts. Being able to say why you keep a second terminal open while reloading that configuration is a good sign.
What is the one thing beginners get wrong?
Restarting first. When something misbehaves, the urge to restart is strong, and it wipes the evidence: memory contents, open connections, the very queue you wanted to examine. The Atlas lists restart-first as an anti-pattern. A short sequence of looks beats a quick restart: capture load, memory, disk and network readings, read the last log lines and write down the time. Then act.
The second most common error is trusting one number. A processor graph at forty percent can hide short bursts, and a high load figure can mean waiting on storage. Learn what each reading includes before you let it decide your next step.
What should I read when a machine feels slow?
Start with a fixed first-minute routine. The Atlas mentions a well-known list of ten commands from 2015 and an approach that checks utilisation, saturation and errors for each resource, and notes that newer kernels report pressure-stall information that says directly how long tasks spent waiting. Mastering the whole set in the first week is unnecessary. You need the habit of running the same checks in the same order, and of saving the output so that a colleague can see what you saw.
Do I need to learn cron, or just timers?
Both, but spend your energy on timers. Cron is everywhere in older documentation, and it has quirks: a minimal shell, and a percent sign that becomes a newline unless escaped. Timers can catch up after downtime, add a random delay to avoid synchronised load, and trigger another unit when they fail. The Atlas recommends moving cron jobs to timers as part of tightening a baseline.
What is toil, and is the fifty percent rule real?
Toil means repetitive manual operational work that scales with the size of the estate. You may have read that a certain large company caps it at half of an engineer’s time. The Atlas stresses that this is that company’s own goal, not a universal rule. Measure your team’s share, watch the trend and choose a target that suits your context.
How do I practise without a production system?
Build a home lab with two or three virtual machines. Break them on purpose. Fill a disk and watch what fails. Set a wrong value early in the remote-login file and see why it wins. Revert a firewall change with a timer you armed beforehand. Restore a file set from a backup onto a clean machine and time it. The Atlas’s maturity scale rewards exactly this kind of evidence, with its highest level resting on drill, console and restore records.
Where does this fit in a DevOps learning path?
Think of it as the floor under the pipeline. Our article on sharpening your reasoning when DevOps problems appear argues that good engineers ask what changed before they ask who is to blame, which is the heart of SysOps triage. For the bigger picture of how skills, roadmap and strategy connect, the longer read about a practical DevOps roadmap for organisations changing how they run IT is a sensible companion.
What would a good first month look like?
Week one: list the hosts you can reach, remove shared logins and write down who owns what. Week two: install a tool that records performance history and write a ten-line triage script. Week three: set sensible limits on the journal and move one scheduled task onto a timer. Week four: choose a small data set, back it up, restore it elsewhere and write down how long it took. The Atlas groups similar tasks into its first-month list, and so a learner who completes them has covered the same ground as a working team.
A last word of encouragement
None of this is glamorous. Nobody hands out trophies for a disk that never filled up or a job that quietly ran every night. But when the pipeline you build sits on a host you understand, the Friday surprises become rarer and shorter, and you get to spend your evenings away from the pager.

