Contrastive learning technology has shown significant potential in alleviating the data sparsity problem in sequential recommendation. Nevertheless, existing methods mainly derive self-supervised signals from a single perspective or rely on abundant domain-specific auxiliary information to extract high-quality user representations, which either fail to make the most of supervised information or lack of flexibility to extend to more scenarios. In this paper, we propose a novel Multi-View Contrastive Sequential Recommendation (MVCSR) method for sequential recommendation scenarios with few auxiliary information. MVCSR utilizes the set-level, sequence-level and preference-level user information to derive valuable supervised signals and provides three effective contrastive learning strategies to improve the item representations, sequence embeddings and user preferences, eventually boosting the recommendation performance. Experimental results reported from four Amazon datasets have demonstrated the superiority of MVCSR compared with all the state-of-the-art sequential recommendation baselines. Moreover, we validate the effectiveness of each contrastive learning strategy and show our advantage at dealing with the recommendation for users with short sequences.