In this paper, we propose a hybrid design-time and run-time framework for reliable resource allocation, i.e., mapping and scheduling of applications, in multi-core embedded systems with solar energy harvesting. Our framework is designed to cope with the complexity of an application model with data dependencies and run-time variations in solar radiance, execution time, and transient faults, with support for flexible schedule templates at design-time, and lightweight online adjustment mechanisms to monitor run-time dynamics and make adjustments to task execution strategy. Our experimental results indicate improvements in performance and adaptivity using our framework, with up to 29.5% miss rate reduction compared to prior work and 55% performance benefits from adaptive run-time workload management, under stringent energy constraints and varying system conditions at run-time.