Skip to content

CONCEPT Cited by 1 source

Deploy-less operational lever

A deploy-less operational lever is any production- behaviour control whose invocation is a config flip (feature-flag toggle, parameter change, routing-rule edit) rather than a code change pushed through the deploy pipeline. The lever is wired in during a prior deploy; the invocation itself skips the deploy cycle entirely.

Why they matter

Incidents happen on human time scales. When a job is misbehaving at 2 am, the MTTR floor is set by whichever is fastest:

  • Code fix via deploy: minutes to hours (build + test + canary + rollout, or an emergency-bypass deploy with reduced safety).
  • Config flip via control plane: seconds (write a row, wait for propagation).

The deploy-less lever bounds MTTR at the propagation speed of the config substrate. The tradeoff is that the lever must be wired in during a normal deploy — if the code path to check the lever doesn't exist, there's nothing to flip.

Shape

Every deploy-less operational lever has four parts:

  1. A gate in the code path, installed on a prior deploy, that checks some config value at runtime.
  2. A fast-propagation config channel — feature-flag system, global config system, DB row, etcd key — that distributes changes to all evaluators within seconds.
  3. An invocation surface — Rails console, admin UI, kubectl edit, deploy bot command — that operators use to change the config value.
  4. A runbook that names the lever and prescribes when to flip it.

Relationship to other levers

  • Feature flag: general-purpose deploy-less lever, usually for release-engineering (staged rollouts).
  • Global feature killswitch: deploy-less lever specialised for rapid off.
  • Feature-flagged job enqueue rejection: deploy-less lever specialised for async-job class suspension.
  • Fast rollback: deploy-less lever specialised for undoing a recent change (may be the same substrate as the original deploy pipeline, just faster).

Seen in

  • sources/2026-04-21-planetscale-how-to-kill-sidekiq-jobs-in-ruby-on-rails — canonical wiki introduction. PlanetScale's Sidekiq+Flipper composition: the gate is the SidekiqJobsFlipper middleware (installed on a prior deploy), the config channel is Flipper's shared store (DB or Redis), the invocation surface is a Rails console running Flipper.enable("disable_invoice_job"). Operator disables a production job in seconds without a deploy — canonical small-scale instance of the pattern at the application-tier altitude.
Last updated · 378 distilled / 1,213 read