read-only Friday

expression

ops culture

The ops tradition of making no production changes on Friday, because weekends exist.

A deploy at five on Friday has a way of becoming a Saturday incident with half the team unreachable. Mature CI/CD teams argue the rule away - if deploys are that risky, fix the deploys - but the instinct it encodes is respect for blast radius and for people's time off.

Read-only Friday is the convention of not deploying changes late in the week, and the reasoning is entirely about recovery rather than about the change itself. A deployment that breaks on Tuesday afternoon meets a full team, an alert audience and a working day in which to fix it. The same deployment on Friday evening meets whoever is on call, at a weekend rate of attention, with the people who understand the change unreachable.

The practice is sensible and it is also a symptom worth reading. A team that cannot deploy safely on Friday cannot deploy safely at all; they have simply arranged for failures to land at a convenient hour. The underlying weaknesses are the same ones that make any deployment risky: no confident rollback, insufficient automated verification, changes large enough that failure is hard to localize, and a release process that depends on specific people being awake.

Which is why the mature position is to treat it as a temporary control while fixing the cause. Organizations with genuinely reversible deployments, progressive rollout and automatic rollback deploy on Friday without ceremony, not because they are braver but because the failure costs an alert rather than a weekend. Keeping the rule forever is choosing to live with the fragility rather than to remove it.

Also known as: no-deploy Friday

All glossary entries