Serilog correctly configured for production: the Application Insights sink, enrichers for automatic context, levels per namespace, compact request logging, and the cost traps of excessive logging.
Citește mai mult
The three pillars of observability and the differences between them: logs answer "what happened," metrics answer "how often and how bad," traces answer "where it went through." When to use each and how they complement each other.
Citește mai mult
Real integration tests for Cosmos DB without a production account: Local emulator and in GitHub Actions as a service container, NUnit fixture with initialization, self-signed certificates, and timing traps.
Citește mai mult
Lost update: two processes read, modify, and save the same document — the last write overwrites the first. ETags and IfMatchEtag prevent the problem, with retry on PreconditionFailed and the Patch alternative for targeted operations.
Citește mai mult
When you need writes in multiple regions and what price you pay: consistency levels in a multi-region context, conflict resolution with Last-Writer-Wins and custom policies, plus configuring the .NET SDK.
Citește mai mult
Change Feed transforms the container into an event source: every insert and update becomes a trigger. Change Feed Processor in .NET, lease container, scaling across multiple instances, and practical cases.
Citește mai mult
How to choose the correct partition key in Cosmos DB: cardinality, distribution, and alignment with query patterns. Cross-partition queries, hierarchical partition keys, and concrete examples from an AI chat system.
Citește mai mult
The dual-write problem: the database and the message cannot be updated atomically. Outbox pattern with Cosmos DB and Service Bus: you save the event alongside the data, a dispatcher guarantees its publication, idempotent consumers close the loop.
Citește mai mult
