Freelancers are often handed production access on the first day of an engagement and asked to “just make the change”. That moment carries more risk than any other part of remote DevOps work, because you are acting alone, in someone else’s environment, with a client who may not see the consequences until later. This audit gives you a pre-flight checklist for client production changes: a short, fixed sequence of checks to run before you touch anything you cannot cheaply undo. Print it, paste it into your notes app or keep it as a template. Use it every time, in the same order.
Scope matters, so state it plainly. This is not a methodology for the whole project. Discovery, design and testing have their own routines. The checklist covers the final minutes before a launch: a deployment, a configuration change, a database alteration, a DNS switch or a permissions update.
Section A: Authority and agreement
A1. Is the change written down and approved? A message in a chat thread counts if it names the change, the system and the person authorising it. Verbal approval does not. When disagreements arise later, a dated written instruction is the thing that protects both parties.
A2. Is the approver the right person? Someone friendly is not necessarily someone entitled to sanction a risky action. Confirm that the person who said yes owns the system or has been delegated authority. If you are unsure, ask once and note the answer.
A3. Is the timing agreed? Confirm the window with the client and check it does not overlap a sales event, payroll run, end-of-month close or planned maintenance on their side. Time zones cause more freelance trouble than most technical faults, so state the window in the client’s local time and in yours.
Section B: Safety net
B1. Is there a restore point? Identify the backup, snapshot or previous version you would return to. Confirm it exists and, where feasible, that it opens. If the client has no backup, say so in writing and agree how to proceed before continuing. Do not quietly accept responsibility for a gap you did not create.
B2. Can you describe the reversal in one sentence? “If this fails I will restore the previous configuration, which takes a few minutes and requires nothing from the client.” If you cannot say that, the change is a one-way door and deserves a conversation with the approver before you proceed.
B3. Have you recorded the current state? Take a note, export or screenshot of the settings you are about to alter. The before picture often resolves questions that arise a week afterwards.
Section C: Target and tooling
C1. Are you pointed at the right environment? Read the account name, project, cluster or hostname aloud and compare it with the plan. Freelancers who juggle several clients are especially exposed to this slip, because terminals, credentials and bookmarks pile up. A different background colour or a shell prompt that shows the environment name is a cheap defence.
C2. Are you using the right credentials? Confirm the identity you will act as and that it has only the access the change needs. If you are using an administrative role for convenience, switch to a narrower one for this task.
C3. Is your own setup stable? It sounds odd to audit yourself, but a laptop on low battery or a flaky connection can turn a ten-minute change into a half-finished one. Plug in, check connectivity and close unnecessary applications. Consider whether a second way of reaching the system is available if your first drops.
Section D: Visibility and people
D1. Is the client aware it is starting? Send a brief note before you begin. The people who may notice effects, such as support or operations, should not learn about the change from a customer complaint.
D2. Are alerts and dashboards quiet? Glance at whatever monitoring the client has. If something is already wrong, either deal with it first or record it so the existing problem is not blamed on your work.
D3. Who is reachable if it goes badly? Identify a client contact who can respond during the window and a way to reach them, whether that is a phone number or a chat channel they actually read. Do not run a risky change when nobody on their side can answer a question.
Section E: After the press
E1. What will you check, and for how long? Pick the handful of readings that will tell you the change worked, and note beside them the moment you would back out. Deciding afterwards tempts you to explain away early warning signs.
E2. When will you report? Promise a short completion note and send it. A plain message saying what changed, what was verified and what to watch protects your reputation as much as the change itself.
Using the audit in practice
Fifteen items sounds long, but with practice the whole sequence takes a few minutes, and most lines are answered at a glance. The discipline is in reading all of them, in the order given, every time, including when the change seems trivial. Trivial changes are where the habit is most likely to lapse, and where a wrong environment or missing backup is most likely to bite.
Keep a running copy per client. Clients differ in their windows, approvers and backup arrangements, so a small client sheet attached to the checklist saves repeating questions. It also demonstrates professionalism. Several freelancers report that clients react warmly to being asked these questions, because it signals care rather than bureaucracy. If you are building client relationships from scratch, the practical article on earning confidence with clients who have never worked with you before expands on this kind of habit.
A short story about the wrong terminal
A contractor who looked after the websites of several small businesses had two terminal windows open: one for a client’s staging server and one for the live site. Both prompts looked identical. She typed a command to clear a cache and reset a database table meant for testing, and pressed enter in the wrong window. The live shop lost its saved baskets for the afternoon. The restore took a while, the client was gracious, and the contractor was mortified.
Nothing about the technical skill was lacking. What failed was the absence of a pause. A single line reading “say the hostname aloud” would have exposed the mix-up. After that day she changed her prompt colours, added the checklist to every client folder and began pasting the target name into the change note before running anything. The cost was seconds per change; the benefit was never repeating that afternoon.
When to skip or shrink the list
Not every action needs the full audit. A read-only query or a change in a throwaway sandbox can skip most of it. The test is reversibility: if you can undo the action in seconds with no one noticing, run a lighter version. If undoing it would require a restore, an apology or an overnight session, run the full sequence. Be honest with yourself about which category you are in, because optimism is the usual reason people downgrade a risky step.
Adjusting the audit as you learn
After each engagement, spend a few minutes reviewing. Which item caught something? Which one never did? Add a line if something slipped through and remove one that has become reflex. Freelancers who maintain their templates this way tend to develop a distinctive, reliable method, and that consistency becomes part of what clients pay for. If you are also considering how side work interacts with a main job, there is a separate read about running a DevOps side business without neglecting the day job, which handles the scheduling side of that balance.
A glance at how other communities handle it
Careful routines before irreversible actions are not unique to DevOps. If you browse the CryptoBlockCoins blog on cryptocurrency and blockchain, you will find coin write-ups, tutorials and daily logs from a community that treats final actions seriously. You do not have to share any of its interests to notice the common thread: a fixed routine beats improvisation when mistakes cannot be reversed.
Take it away
Copy the five sections into a note today. Delete anything that does not apply to your usual work and add one line reflecting your own most recent scare. Then use it on your next production change, and tell the client what you did. A calm, repeatable launch routine is one of the few things that makes a solo operator look like a team.