Java

How many instances did the calculation service have?

What happens at application startup if only the rollback block is changed in an already applied changeset?

Where is the Circuit Breaker attached and what happens when calling a method with TimeLimiter and fallback?

What is the feature of a composite index if the index is built on three columns, but the search is only performed on the third?

How does Liquibase validate the integrity of applied DDL changes?

Where was the Circuit Breaker in your system: at the input or on outgoing calls?

How to directly ensure that the entire business operation fits within 10 seconds?

What is a silent failure?

How to fix an orchestrator with two independent external calls to meet the 10-second business limit when service durations are 10 and 2 seconds?

Kafka, producer, traces, and basic metrics look normal, but there are no final messages/results. What are the next diagnostic steps?

What types of metrics are available in Prometheus and what are they used for?

Why are you leaving Gazprom Neft?

How does a Circuit Breaker help meet this time requirement?

Does the calculation service operate synchronously or asynchronously relative to the REST API?

Why does the first call still last 10 seconds and how do Feign, CompletableFuture, and TimeLimiter work in this scheme?

Describe a complex task on a project and the process of handling a request/event from start to finish: processing stages, integration, broker, and optimizations performed.

Messages stopped arriving at the calculation service. What steps will you take for diagnostics?

Tell me about yourself.

How to organize Counter metrics with retry and DLT so that repeated attempts do not lead to double counting of original messages?

How were integration tests organized: real containers, WireMock, or a full set of services? How did it look in the 'Otkritie' bank?