Normality Test

Do not ever bother to test normality of raw data. It's not valid. The normality assumption is for the normality of residuals. You need to do the ANOVA first then check normality of its residuals. Formal normality tests are poop. They are overly sensitive and will call any sufficiently large sample non-normal. So, do the ANOVA. Plot the residuals, if it doesn't look too wonky, go with the ANOVA.