Ordinal non-binary categorical variable with missing values

Hi all,

I have a predictor variable, Education, taking values from 1 to 5 (1=Primary school, 5=Master). I have a 30% of missing data (on a total sample of 2 millions observations) for this variable, which is categorical, ordinal and taking multiple (non binary) values. How should I handle it when using XGBoost?

Best,

Andrea