Data and State for Site Reliability Engineers
Data and state looks at reliability when systems store information across many machines. It covers distributed consensus, replication and consistency trade-offs, data integrity and backups, along with testing and safe launches such as canarying. Stateful services are harder to keep reliable than stateless ones because data cannot simply be restarted away.
Stage: Engineering for Reliability · Part of the Site Reliability Engineer roadmap · Last reviewed 19 November 2026
What to learn in Data and State
- Explain replication and why consensus protocols like Paxos or Raft are needed
- Describe the trade-offs between strong and eventual consistency
- Design backups and verify that restores actually work
- Roll out changes with canary releases and staged percentage rollouts
- Test reliability with load tests and fault injection before launch
Data and State interview questions
Why is a backup not useful until you have tested a restore?
Backups can be incomplete, corrupted, encrypted with a lost key or too slow to restore within your recovery time. Only a regular restore test proves the data is usable and shows how long recovery really takes.
What is a canary release?
A new version is sent to a small share of traffic or servers first and compared against the current version on key metrics. If errors or latency worsen, it is rolled back with limited impact; otherwise the rollout widens in stages.
Where Data and State fits in the Site Reliability Engineer roadmap
Data and State belongs to the Engineering for Reliability stage (advanced level), which takes about 40 hours. Make systems resilient.
Other units in this stage
Back to the full Site Reliability Engineer roadmap