Bigtable Automation Wipes Everything
Chapter 7 - The Evolution of Automation at Google Failing to configure one subsystem, or configuring a system or component differently than other deployments, is a customer-impacting outage waiting to happen. In one case, a multi-petabyte Bigtable cluster was configured to not use the first (logging) disk on 12-disk systems, for latency reasons. A year later, some automation assumed that if a machine’s first disk wasn’t being used, that machine didn’t have any storage configured; therefore, it was safe to wipe the machine and set it up from scratch. All of the Bigtable data was wiped, instantly.
A screenshot of a Google chapter about automation shows a multi-petabyte Bigtable cluster being wiped because an automation system misunderstood why the first disk was unused. The joke is the horrifying contrast between a seemingly reasonable configuration assumption and the instant destruction of all stored data, making it a reaction to catastrophic software mistakes, outages, or overconfident automation.
web browser window / окно веб-браузераbrowser address bar / адресная строка браузераGoogle documentation page / страница документации Googleblue navigation bar / синяя панель навигацииhamburger menu icon / значок меню-гамбургераtext paragraphs / текстовые абзацыcomputer screen / экран компьютераBigtable cluster / кластер Bigtablehard disk / жёсткий дискserver machine / серверная машинаdata wipe / удаление данных
- @ithueti / 2348 2020-05-03
- Type
- image
- Size
- 717×445 · 68.9 KiB
- Mood
- darkly humorous
- Language
- en
- Added