{"id":1190,"date":"2026-11-16T16:10:00","date_gmt":"2026-11-16T16:10:00","guid":{"rendered":"https:\/\/pilottrainingus.com\/blog\/?p=1190"},"modified":"2026-11-19T11:10:00","modified_gmt":"2026-11-19T11:10:00","slug":"flight-school-infraops-case","status":"publish","type":"post","link":"https:\/\/pilottrainingus.com\/blog\/flight-school-infraops-case\/","title":{"rendered":"A Flight School&#8217;s Year of InfraOps: A Case Story in Three Gaps"},"content":{"rendered":"<p>The story below is invented, and the people in it are composites, but every mechanism it shows is real. A small flight school, which we will call Ridgeline Aviation, ran its scheduling portal, student records and weather dashboard on rented cloud capacity. Nobody there called what they did InfraOps. Yet the way they eventually repaired their habits matches what the Ops Atlas entry on InfraOps describes, and the evidence from the case is worth walking through slowly.<\/p>\n<h2>The first Friday: a safe change that was not<\/h2>\n<p>Ridgeline kept its infrastructure definitions in a shared repository. A volunteer developer proposed widening a storage rule so instructors could upload syllabus files. Two colleagues read the proposed difference, approved it, and went home. On Monday the upload still failed.<\/p>\n<p>The cause was banal. The approved change had been merged but never applied, so the live environment still carried the old rule. The Atlas lists it first among four mismatches between written intent and live reality: merged, not applied. For a week the team had believed a fix was live. Instructors, meanwhile, were emailing files to each other.<\/p>\n<p>Evidence point one: approval is a statement about intent, not about reality. A flight school understands this instinctively. A signed work order is not a serviced aircraft.<\/p>\n<h2>The second week: a quiet hand-made repair<\/h2>\n<p>On a busy weekend the dashboard stalled, and a part-time engineer opened the hosting console and raised the memory on one server. The portal recovered. The change was never written back into the repository.<\/p>\n<p>Three months later, a routine rebuild from code produced a smaller server, and the dashboard stalled again at the worst time. The Atlas labels this drift: the live system shifts and the paperwork stays put. It adds a sobering note about detection. A comparison that comes back clean is evidence only about objects the tooling already tracks, so a hand-built resource can sit outside every report. Ridgeline had one of those too, an old test database someone created for a demo.<\/p>\n<p>Evidence point two: the cost of drift is paid later, by someone else, at an inconvenient moment.<\/p>\n<h2>The measurements the team finally took<\/h2>\n<p>Ridgeline&#8217;s coordinator decided to measure before arguing. Borrowing from the Atlas, which proposes its own small set of indicators and is clear that they are suggestions rather than a standard, she counted four things over thirty days:<\/p>\n<ul>\n<li>how many components were described in code versus how many existed in total;<\/li>\n<li>how many separate times someone changed something by hand;<\/li>\n<li>how many applies failed partway;<\/li>\n<li>how old the oldest unresolved difference was.<\/li>\n<\/ul>\n<p>The numbers were unflattering, and that was useful. Roughly a quarter of what ran the school was not described anywhere. There were more manual changes than anyone had admitted. The Atlas&#8217;s own worked illustration makes the same point with different figures: a decent-looking coverage percentage can still hide a pile of resources no reviewer ever sees, and one stale difference can matter more than the headline.<\/p>\n<h2>Three changes, in order<\/h2>\n<p>The team resisted the urge to fix everything. They agreed on a sequence that mirrors the early part of the thirty-sixty-ninety-day outline in the Atlas.<\/p>\n<ol>\n<li><strong>Move the applying to a shared pipeline.<\/strong> Nobody applied from a laptop any more. The pipeline ran a plan on every proposal, posted the result for review, and applied only after approval.<\/li>\n<li><strong>Split the state.<\/strong> The record of what existed had been one large file. Dividing it by service meant two volunteers could work at once, and a mistake in the dashboard could not touch student records.<\/li>\n<li><strong>Run a scheduled comparison.<\/strong> Each night a read-only plan checked code against reality, and any difference opened a ticket with a named owner.<\/li>\n<\/ol>\n<p>One subtlety caught them out. They first saved the plan from the review step and re-used it hours later. Another change had landed in between, and the saved plan no longer matched. The Atlas warns that plans are perishable: applying one invalidates others made against the same record. The team switched to generating a fresh plan at apply time and comparing it to the reviewed one.<\/p>\n<h2>The scare that justified the guardrails<\/h2>\n<p>About then an engineer ran a clean-up command and typed one parameter incorrectly. It would have removed far more capacity than intended. Because the pipeline had a minimum-capacity floor, the run stopped. This resembles the episode the Atlas retells from February 2017, when an engineer&#8217;s command with one wrong input pulled out excess storage-service capacity, after which the provider changed its tooling to remove capacity more slowly and to enforce minimums.<\/p>\n<p>Evidence point three: guardrails are cheap rehearsals. A pilot reading about weather minimums, as described in <a href=\"https:\/\/pilottrainingus.com\/blog\/weather-study-tips-for-student-pilots\/\">advice for student pilots studying meteorology<\/a>, accepts that a limit set on a calm day protects you on a bad one.<\/p>\n<h2>What the plan files taught them about secrets<\/h2>\n<p>During a security review someone noticed that saved plan files were being uploaded as build artifacts. Anyone with read access to the repository could download them. The Atlas explains why that is a problem: state and plan files can contain secrets, and the role that merely plans still runs provider code, so it needs care close to that of the role that applies. Ridgeline stopped sharing plan files between jobs and tightened who could trigger planning on proposals from outside contributors.<\/p>\n<h2>Six months later<\/h2>\n<p>The rebuild drill, the last item in the Atlas&#8217;s maturity ladder, was the real test. The team tore down a copy of the whole school environment and rebuilt it from the repository alone. It took an afternoon, with two gaps: a firewall rule never captured and the forgotten demo database. Both went into code. A year earlier either would have been a surprise during an emergency.<\/p>\n<p>Look at what changed and what did not. Tools stayed the same. People stayed the same. What changed was that three copies of the facts, the code, the tool&#8217;s record and the live environment, began to agree, and that disagreements were noticed on a timer rather than by an angry user. The Atlas treats this lifecycle as the core of its subject, and its entry is a thorough read for anyone who wants the full vocabulary; you can find it via <a href=\"https:\/\/www.xopsschool.com\/infraops\/\">the Atlas&#8217;s InfraOps reference page<\/a>.<\/p>\n<h2>Takeaways for a training organisation<\/h2>\n<ul>\n<li>Treat an approved change as unfinished until the live environment confirms it.<\/li>\n<li>Count what exists outside your records; the number is rarely zero.<\/li>\n<li>Rehearse a rebuild, as you would rehearse an emergency procedure.<\/li>\n<\/ul>\n<p>For readers who want a structured route into this field, <a href=\"https:\/\/pilottrainingus.com\/blog\/your-path-to-mastering-infrastructure-engineering-with-devopsschool\/\">a learning path for infrastructure engineering<\/a> sits on this blog as well.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>A fictional flight school learns why approved changes, hand-made fixes and shared plan files drift apart, and how measuring them restored trust.<\/p>\n","protected":false},"author":6,"featured_media":1191,"comment_status":"","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[664],"tags":[],"class_list":["post-1190","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-aviation-tech"],"_links":{"self":[{"href":"https:\/\/pilottrainingus.com\/blog\/wp-json\/wp\/v2\/posts\/1190","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/pilottrainingus.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/pilottrainingus.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/pilottrainingus.com\/blog\/wp-json\/wp\/v2\/users\/6"}],"replies":[{"embeddable":true,"href":"https:\/\/pilottrainingus.com\/blog\/wp-json\/wp\/v2\/comments?post=1190"}],"version-history":[{"count":1,"href":"https:\/\/pilottrainingus.com\/blog\/wp-json\/wp\/v2\/posts\/1190\/revisions"}],"predecessor-version":[{"id":1301,"href":"https:\/\/pilottrainingus.com\/blog\/wp-json\/wp\/v2\/posts\/1190\/revisions\/1301"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/pilottrainingus.com\/blog\/wp-json\/wp\/v2\/media\/1191"}],"wp:attachment":[{"href":"https:\/\/pilottrainingus.com\/blog\/wp-json\/wp\/v2\/media?parent=1190"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/pilottrainingus.com\/blog\/wp-json\/wp\/v2\/categories?post=1190"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/pilottrainingus.com\/blog\/wp-json\/wp\/v2\/tags?post=1190"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}