Beauty and the Beast: your application and distributed systems
summary
Explains concrete failure modes and remedies: elasticsearch and MongoDB return distinct not primary and auth errors, network blips can occur to or from the server, and retries require idempotency. Shows idempotent patterns (unique token then conditional increment) and optimistic locking with versioning for safe retries, describes MongoDB sessions for causal consistency, lists client options (sniffing, timeouts, retry settings), and demonstrates testing against a real networked cluster with dedicated benchmark boxes.
AI-written summary · about
chapters (17)
- 0:00Start
- 3:39Simple
- 6:17the Beast
- 9:49Consensus at work
- 11:57Applications of consensus
- 14:35Raft algorithm
- 15:10Elections
- 16:36Axes of consistency
- 17:58Causal consistency
- 23:34Redundancy
- 23:58Recovery
- 25:23Network errors
- 26:46Operation errors
- 32:22Idempotent operations
- 34:36Elasticsearch optimistic locking
- 37:24What does the client do in the case of
- 41:09Testing
related talks
53mKeynote - Jepsen 7: Anna Concurrenina
43mDistributed Systems
26mService Oriented Disasters
29mFollowing the Data: A Deeper Look into Distributed Applications
51mDesigning for Failure: Build Resilient Services
46mEdge Performance with In memory NoSQL
21mDatabase Reliability Engineering
01h 00mThe Secret Life of Distributed Systems