Task Scheduling problem (TSP) in cloud computing is a critical aspect as diversified tasks from heterogeneous resources comes to cloud console. Mapping these diversified tasks to suitable virtual machines is challenge for the cloud service provider(CSP) to employ an efficient algorithm to tackle TSP. Ineffective scheduling lead to increase in makespan, failures which impacts reliability on CSP. Many authors developed various task scheduling mechanisms to tackle parameters makespan, execution time, energy consumption but very few authors addressed Rate of failures, reliability but there is need to optimize scheduling process in Cloud paradigm as it is a dynamic scenario. In this paper, a reliability aware task scheduler is formulated which calculates task priorities at task manager level to effectively schedule tasks. All priorities are fed to scheduler which is integrated with a deep Reinforcement learning model A3C which improved by adding RCNN to accelerate learning capacity and to extract features accurately mapping tasks to VMs according to their priorities. Simulations are carried out on Cloudsim using fabricated data distributions, real time worklogs. We evaluated our proposed RTSIA3C with baseline algorithms DQN, A2C. Results revealed that RTSIA3C outperformed over baseline approaches by minimizing makespan, rate of failures while improving reliability.