Lin Wei had been “learning containers” for four months. He had watched tutorials, followed along with two long workshops and kept a terminal open most evenings. Yet on the day his team asked him to package a small service and ship it, he froze at the same place he had frozen in week one: writing a clean image build and understanding why the second build took far longer than it should. The problem was not effort. It was the shape of his effort. This how-to follows what he changed, and shows how you can practise deliberately with a focused kata on a single DevOps skill.
Step 1: Admit that hours were not the issue
Lin had logged perhaps eighty hours. Most of it was comfortable: re-running commands he already knew, re-reading the same tutorial and tweaking things until they worked. That builds a feel for the commands. It does not fix the one thing that goes wrong every time. A kata is the opposite: a tiny exercise you repeat on purpose, aimed at a gap you have named.
So the first step is a short admission written on paper. For Lin it read: “I cannot structure an image build so rebuilds are fast.” That single line replaced a vague wish to “get better at containers”.
Step 2: Shrink the skill to a ten-minute exercise
A kata must fit in one sitting without fatigue. Lin took a small sample app and wrote the goal: produce a working image, change one source line, rebuild, and finish the second build in a fraction of the first. No networking, no orchestration, no registry. Just that loop.
If your own weakness is something else, such as reading a failing pipeline log or writing a short infrastructure definition, the same principle holds. Cut it down until a single attempt takes ten minutes or less, and until success or failure is obvious.
Step 3: Decide in advance what “good” looks like
Before the first attempt, Lin wrote his pass mark: image builds, app starts, and a rebuild after a one-line change takes less than a quarter of the first build time. Having the standard on paper stopped him from moving the goalposts mid-session. Feedback only works when you know what you are comparing against.
Step 4: Run it, timed, then stop
He set a ten-minute timer, made the first attempt and recorded the numbers: the first build and the rebuild took almost the same time, which told him layers were being invalidated at the wrong step. When the timer rang he stopped, even though he wanted to keep fiddling. Stopping on time preserves the habit and keeps sessions repeatable.
Step 5: Get honest feedback quickly
Numbers helped, but Lin also wanted a human view. He asked a teammate one narrow question: “Looking at this build file, what is the first thing you would reorder?” The reply, to copy dependency manifests before the rest of the source, was something he would not have guessed after weeks of random tinkering.
Good feedback is quick, specific and willing to say “no”. A tool’s output, a checklist and a colleague who can be blunt all qualify. When he later wanted to understand the requirements of a bigger service, he read a clear overview of the questions worth answering before a custom web application is built and used it as a checklist for his next kata, not as light reading.
Step 6: Change exactly one thing
Lin applied his teammate’s suggestion, and nothing else, then reran the drill. Rebuild time dropped sharply. In the next round he changed only the base image choice and observed the effect. Because each round altered a single variable, he could tell what mattered. When people change four things after a failure, success teaches them nothing, and failure teaches even less.
Step 7: Raise the difficulty when it gets boring
After a few days the exercise became easy, which was the cue to harden it. He added a second service, then a constraint of finishing without looking at his notes, then a rule of narrating each decision aloud. The goal is always to keep the drill just beyond comfortable. A kata that never gets harder becomes the same comfortable repetition you were trying to leave behind.
Step 8: Log what moved
At the end of each session Lin added two lines to a notes file: what was weak and what changed. A fortnight later the file showed a clear progression, and it gave him material to talk through in his review. It also exposed a pattern: he often rushed the first step of every drill. That became tomorrow’s target.
Pitfalls Lin hit along the way
He kept a short list of the traps, and it is worth borrowing. In week one he drilled the part he liked, which was writing the build file, and avoided the part he feared, which was reading the build output line by line. In week two he let a session run to forty minutes and noticed that his last twenty minutes produced nothing he could repeat. In week three he asked a colleague “is this good?” and received a polite “looks fine”, which taught him to ask narrower questions.
Each trap has the same cure: tighten the target, shorten the box, sharpen the question. If you notice a drill turning into a pleasant evening of pottering, stop and rewrite the pass mark. The point of a kata is that it can fail, and a failed attempt shows you exactly where the next ten minutes should go.
Where an outside coach fits in
Eventually Lin outgrew the feedback available inside his team. Everyone around him shared the same habits. He looked for someone who could watch a drill and name the hidden problem. One place to start is a listing site where you can look up DevOps trainers, instructors and coaches across areas such as DevOps, SRE and DevSecOps, so you can tell which people claim experience in the specific skill you are drilling. Then ask any candidate to review a ten-minute drill rather than to deliver a lecture.
Step 9: Connect the kata to real work
A kata is only valuable if it transfers. Lin picked a real task from the backlog and applied the improved build structure to it. He also compared his plan against a deeper write-up on how detailed web project requirements are drawn up by engineers, adding one missing item to his own checklist. The loop then repeated: new weakness, new kata, small gain.
You do not have to practise everything. Three or four recurring weak spots drilled over a quarter usually matter more than twenty topics touched lightly.
A starter kata you can run tonight
- Write one sentence naming a skill you fumble repeatedly.
- Reduce it to an exercise that takes ten minutes or less.
- Write the pass mark before you begin.
- Time-box the attempt and record numbers or observations.
- Ask one narrow question of a person or a tool.
- Change one thing and run it again.
- Note two lines on what changed, then stop.
Lin’s four months of hours did not vanish; they gave him the vocabulary. But his improvement accelerated only after he stopped repeating and started aiming. The same shift is open to you, and it fits in the gap before dinner.

