The fog computing paradigm is rapidly gaining popularity for latency-critical and bandwidth-hungry IoT application deployment. Meanwhile, MicroService Architecture (MSA) is increasingly adopted for developing IoT applications due to its high scalability and extensibility. For mission-critical IoT services in fog, reliability remains one of the most critical QoS requirements due to less dependability of fog resources. Granular microservices with independent deployment and scaling exhibit great potential in utilising resource-constrained fog resources to improve reliability through redundant placement. However, current research on service placement lacks reliability-aware holistic approaches that combine the MSA features and failure characteristics of fog resources under independent and correlated failures. Hence, we analyse MSA and formulate the reliability-aware placement problem by modelling composite services as k-out-of-n serial-parallel systems in a throughput-aware manner for placement under fog resource failures. Our proposed Reliability-aware Placement Method (RPM) is a hierarchical policy combining improved PSO and NSGA-II algorithms. We integrate it with Monte Carlo reliability calculations to produce redundant placements reaching a trade-off between reliability and cost. The performance results reveal that compared to the benchmarks, our algorithm shows significant improvements in reliability satisfaction (up to 25%) and time to first failure (up to 40%), thus providing a robust placement method.