Existing entities and relations joint extraction methods mostly adopt an asynchrony framework, extracting entities and relations at the different time. Such asynchronous joint paradigm suffers from several issues: noisy intermediate redundant information induced, limited interaction among components, and exposure bias from training to inference. Synchronous joint extraction framework has no these issues, however, it has been challenged by overlapping relational triples problem, the main bottleneck preventing the widespread use of synchronous joint paradigms in community. We present a strategy of relation inquiry to break the impasse, empowering synchrony framework: on the ability of extracting overlapping triples, and more sufficient interaction learning for entities and relations models, producing few intermediate redundant information. Under relation inquiry strategy, original joint extraction problem is decomposed into four subtasks, leading to a novel synchronous joint extractor for entities and relations. Experiments on three public datasets demonstrate the effectiveness and robustness of our proposed synchronous joint extractor, outperforming all the classic baselines on different text domains and relation schema.