Service board

The balloon lift desk

Six stations, one crew and one register. TRVLFT CO. runs every computer systems engagement through the same lift desk routine, from the first envelope check to the last chase crew walk across the field.

01 Envelope check desk

Envelope Check Desk

The envelope check desk is where every engagement begins, and it is the station we refuse to skip. Before any code changes hands, we lay the whole system out as if it were a fabric envelope on a long table and read it panel by panel. We map services, data stores, scheduled jobs, external integrations and the people who own each one. We look for the quiet damage that accumulates in a living system: duplicated helpers, half retired queues, credentials shared between services and test data that has crept into production.

The desk produces three artefacts. The first is a full envelope map, drawn so that a new engineer can follow it without a guided tour. The second is a gore register that lists every panel of the system with its owner, its test coverage and its current stress level. The third is a lift plan that states what should be folded, what should be patched and what should be left alone. We do not begin a rebuild without this desk, because a crew that has not read the envelope cannot know which seams are already thin.

Typical span: two to four weeks. Crew size: two engineers and one reviewer.

02 Basket load rounds

Basket Load Rounds

A basket is loaded in rounds, and each round is weighed before the next one starts. We use the same discipline for dependencies, migrations and releases. A basket load round is a scheduled walk through everything the system will carry into its next state, checking whether each item is documented, versioned, backed up and reversible. Nothing is added to the basket until the previous round has been signed off by the crew member on duty.

Rounds are deliberately repetitive. The same sheet is used each time, the same questions are asked in the same order and the same person is not allowed to sign two rounds in a row. This is how we keep a large migration from turning into a guess. If a load cannot be weighed, it does not board. The output of the rounds is a basket sheet that lists every dependency with its weight, its owner and the exact move that would remove it again if the flight has to be scrubbed.

Runs weekly during a build cycle, with a final weigh before any wind window.

03 Mooring drill weeks

Mooring Drill Weeks

Mooring is what keeps a balloon from wandering while the crew is still on the ground, and the drills are what make the mooring reliable under pressure. During a mooring drill week the whole team rehearses the moves for holding a system steady when a dependency goes silent, a database fails over or a third party starts returning malformed responses. Each drill is short, timed and followed by a five minute review that ends in one written correction to the runbook.

We run the same drill three times before we change it, because a drill that changes every time teaches nothing. The rounds cover isolated failure, partial degrade, slow degrade and total loss of an external service, and the crew must reach the ring, apply the hitch and confirm the state without consulting a document mid drill. What clients notice most is not the drills themselves but the evening review, where the crew corrects the plan while the failure is still fresh in memory instead of weeks later in a post mortem.

One drill week per quarter for standing pad clients. Four to six drills per week.

04 Burner audit days

Burner Audit Days

Every system has burners, the processes that can start themselves and spend resources without a human in the loop. A burner audit day is a full-day sweep of every source of heat in the stack: cron entries, queue consumers, webhook retries, autoscalers, background imports and anything else that fires on its own. For each burner we record when it fires, why it exists, what it touches and what happens if it fires twice by accident.

Audit days tend to produce two kinds of finding. The first is the burner that nobody can explain, which is usually a forgotten job that has been quietly rewriting a field for years. The second is the burner that fires far too often, usually a retry policy that was tuned for a slower network and never revisited. Both go into the same register with a named keeper and a date, and the register is reviewed at the next wind window before any release goes ahead.

One day per stack, repeated every six months or after any major integration.

05 Wind window sessions

Wind Window Sessions

A wind window is the period when conditions are right and the crew is ready, and nothing flies outside it. In systems work the window is a published change period with a hard start, a hard end and an explicit go or no-go call made at the gate. We publish windows weeks in advance, keep them as narrow as the work allows and open each session with a short read of the register so that everyone knows what changed since the last flight.

The session itself is calm. The crew has already walked the basket loads, so the window is only about executing a known plan and watching the burners. If a condition falls outside the plan, the lift is scrubbed and the reason is written down the same day. We treat a scrub as a successful outcome, because a cancelled flight on the ground costs far less than an unplanned descent. The session closes with a field note that records what flew, what stayed and what should change before the next window.

Windows published monthly. Go or no-go call made by the keeper on duty.

06 Chase crew evenings

Chase Crew Evenings

After the basket lands, the chase crew follows the flight path and collects what the wind scattered, so that the next morning starts on clean ground. In our practice the chase crew is the follow-up team. They walk the field after every window and close out the loose ends the flight left behind: stale feature flags, orphaned branches, temporary credentials, dashboards that still point at the old service and alerts that no longer matter.

The evenings also serve a second purpose. With the flight still fresh, the crew reviews the day together and corrects the runbook in the same sitting, rather than saving the lessons for a quarterly report that nobody reads. Clients receive a short field note the following morning that lists what was collected and what is still lying on the grass. Nothing is left open silently, and nothing is closed without a line in the register.

Held after every wind window, usually within forty eight hours of landing.

The process overview

Every lift desk engagement follows the same four stages, and each stage has a written exit condition that the crew must meet before moving on. The stages are sequential on purpose. A basket load cannot begin before the envelope check is signed, a wind window cannot open before the burner audit has closed and a chase evening cannot end before the register is current.

  1. Stage oneRead the envelope at the check desk and produce the map, the gore register and the lift plan.
  2. Stage twoWalk basket load rounds until every dependency is weighed, owned and reversible.
  3. Stage threeRun mooring drills and burner audits until the crew and the stack are both ready for a window.
  4. Stage fourFly only inside the published wind window, then send the chase crew across the field.

We are happy to explain how these stages fit a particular deadline, budget or regulatory constraint. The desk at TRVLFT CO. can be reached at escalate@trvlft.com or on +14325295260, and we answer scope questions in plain language.

Pick the first station

Most clients arrive with a specific worry, whether it is a migration that keeps slipping, an integration nobody owns or a team that has forgotten how its own systems behave under pressure. Tell the desk which station feels closest to your problem and we will say honestly whether the lift desk is the right shape for it.

Talk to the lift desk →

Before you call

Have a rough count of your services, a sense of which integrations are load bearing and the name of the person who signs for the biggest dependency. That is enough for a first conversation.

The pad is at 410 Hidden Village Ln, Holland - 43528-6608, United States (US), and visits are by appointment.