system-design – Page 13 – Türker Yıldırım

Reading from a replica in a transaction is not a read

NoteJune 22, 20222 min read

Opening a transaction pins the connection, so a read inside one goes to the primary regardless of what the router intended. The test-suite case is the one that…

A recording rule is a query you stopped rewriting

NoteJune 9, 20221 min read

The same ratio appears in a dashboard, two alerts and a report, and each copy can drift from the others. A naming convention with colons is the established…

The deploy that needed a database restore

Deep diveJune 7, 202215 min read

A migration that wrote bad data to 41,000 rows. Rollback restores the code and not the rows.

The queue table that became a bottleneck at four million rows

Deep diveJune 3, 202210 min read

A database-backed queue that worked for four years. SELECT FOR UPDATE SKIP LOCKED, and a lock wait timeout.

An error budget spent deliberately is the point of having one

NoteMay 31, 20222 min read

A budget that is never spent means the objective is too low, and a budget that is exhausted every month means it is too high or the system…

The report that ran on a replica and was three minutes stale

Deep diveMay 31, 202210 min read

Read routing added for capacity, and a report that disagreed with the dashboard. Two numbers, both correct, three minutes apart.

Style variations, and the theme that ships four designs

Deep diveMay 18, 202210 min read

6.0 in May, and styles/ as a directory of theme.json fragments. Four client sites that differ only in colour and type.

An SLO that survived contact with the business

Deep diveMay 17, 202214 min read

99.9% was written in a document in 2020 and had never been measured. Three teams with three definitions of “up”.

A multi-window alert stops firing when it recovers

NoteMay 11, 20221 min read

A long window detects a sustained problem and keeps firing for hours after it ends, which trains people to ignore the alert. The long window is doing the…

A deprecation with no measurement never completes

NoteMay 11, 20222 min read

Announcing a removal date without instrumentation means the date arrives and nobody can say whether removing it is safe. The brownout — disabling deliberately for a short window…