A resonant cantilever sensor is, for the first time, dressed in a water-proof 'diving suit' for real-time bio/chemical detection in liquid. The μ-'diving suit' technology can effectively avoid not only unsustainable resonance due to heavy liquid-damping, but also inevitable nonspecific adsorption on the cantilever body. Such a novel technology ensures long-time high-Q resonance of the cantilever in solution environment for real-time trace-concentration bio/chemical detection and analysis. After the formation of the integrated resonant micro-cantilever, a patterned photoresist and hydrophobic parylene thin-film are sequentially formed on top of the cantilever as sacrificial layer and water-proof coat, respectively. After sacrificial-layer release, an air gap is formed between the parylene coat and the cantilever to protect the resonant cantilever from heavy liquid damping effect. Only a small sensing-pool area, located at the cantilever free-end and locally coated with specific sensing-material, is exposed to the liquid analyte for gravimetric detection. The specifically adsorbed analyte mass can be real-time detected by recording the frequency-shift signal. In order to secure vibration movement of the cantilever and, simultaneously, reject liquid leakage from the sensing-pool region, a hydrophobic parylene made narrow slit structure is designed surrounding the sensing-pool. The anti-leakage effect of the narrow slit and damping limited resonance Q-factor are modelled and optimally designed. Integrated with electro-thermal resonance excitation and piezoresistive frequency readout, the cantilever is embedded in a micro-fluidic chip to form a lab-chip micro-system for liquid-phase bio/chemical detection. Experimental results show the Q-factor of 23 in water and longer than 20 hours liquid-phase continuous working time. Loaded with two kinds of sensing-materials at the sensing-pools, two types of sensing chips successfully show real-time liquid-phase detection to ppb-level organophosphorous pesticide of acephate and E.coli DH5α in PBS, respectively. The proposed method fundamentally solves the long-standing problem of being unable to operate a resonant micro-sensor in liquid well.