This paper presents an integrated virtual metrology (VM) and quality excursion detection framework for online implementation on a plasma etch tool. Traditional external metrology have inherent delays and are sparingly performed only on the most critical steps of manufacturing. To reduce metrology delay and limit the impact of quality excursion wafers, VM models can be used predict relevant quality outputs using process variables such as chamber pressure, gas flow rate, power level, and other plasma measurements. Although existing VM methods work well using off-line data sets, they face online implementation challenges, such as outliers and process drifts that degrade model accuracy. To develop an efficient and reliable framework for online VM monitoring, we combine data pretreatment using total partial least-squares projection, model updating with moving window PLS, and model feature selection using PLS variable importance in projections core filtering. The combined model can be implemented with low memory footprint and is also shown to be robust against drifts and disturbances. Modeling results from industrial data sets are shown to demonstrate the effectiveness of our approach.