soaktest

Soak test, also known as endurance testing, is a type of performance testing in which a software system is subjected to a significant workload for an extended period to observe its behavior over time. The goal is to identify issues that only appear after prolonged operation, such as resource leaks, gradual degradation, data corruption, or stability problems under steady load.

A soak test examines how the system handles sustained use, including memory and resource utilization, file handles, database connections, and external dependencies. It is typically performed under near-production load, lasting from many hours to several days, and sometimes repeated in cycles.

During a soak test, metrics such as response time, throughput, error rate, CPU and memory usage, disk and network I/O, garbage collection behavior, and resource exhaustion thresholds are monitored. Data integrity and logging accuracy are checked, along with the system’s ability to recover from failures and maintain consistent state.

Common targets include web applications, databases, message queues, batch processors, and microservices architectures. It differs from load testing (which measures performance under expected peak conditions) and stress testing (which pushes a system beyond its limits to observe failure modes) by focusing on endurance and long-term reliability.

Best practices include defining clear acceptance criteria, using realistic and varied workloads, gradually extending test duration, monitoring at multiple layers, validating data integrity, and ensuring a rollback or restore plan. Tools such as JMeter, Gatling, Locust, and various monitoring suites are used to drive soak tests and collect observability data.