So, does "food sample" mean different food types (e.g., poultry vs. egg), or does it mean you have seven total samples of one food type?
If different food types, how many sample batches/replications for each type did you create and monitor for the two days?
One food type/seven samples: dependent t-test would allow you to determine whether there was a significant change in measured CFUs from day 1 to day 2. However, with as few as seven samples, the statistical power of this comparison will not be very good, unless the magnitude of the effect is large.
Seven food types/unknown number of samples: A one-between (food type), one within (day) repeated measures anova would allow you to check for whether food type mattered, whether day mattered (both as main effects), and more important (as long as you had at least two samples per food type), whether there was a food type by day interaction. The same consideration about number of samples per food type and statistical power will apply.
If the chief question is equality of CFUs from day 1 to day 2, you could use:
1. Dependent t-test, pairing matched day 1 values to day 2 values for the 7 types.
2. A simple repeated-measures anova (types are the "cases", day 1 and day 2 represent the repeated measures dimension).
The two options are interchangeable; the squared t statistic for option 1 will be equal to the F-ratio for the test of days in option 2, with exactly the same probability.
As stated in my first post, more replications/samples of each food type would make the test of day differences more powerful (as well as offer a better chance at comparing types and whether type x day interaction exists).