When a worker disappears: leases, retries, and safe recovery
An expired lease does not stop a worker. Make recovery safe by guarding writes, identifying retries, and testing the handoff.
Field notes by Osvaldo Demo
I'm Osvaldo Demo. This is my engineering journal: software, infrastructure, and the decisions that make systems useful and dependable.
Backend architecture · Production reliability · Automation
An expired lease does not stop a worker. Make recovery safe by guarding writes, identifying retries, and testing the handoff.
The journal
The practice
Good engineering connects the application, the infrastructure, and the people who need both to work.
APIs, domain models, data pipelines, and background work designed to recover.
Infrastructure automation, delivery systems, migrations, and operational ownership.
Experiments, observability, and capacity decisions grounded in what the system does.
Something worth solving?