This is an example from Ash's Probability and Measure Theory:
Let $\Omega$ be the rationals, $\mathcal{F}_0$ the field of finite disjoint unions of right-semiclosed intervals $(a,b]=\{\omega\in\Omega:a<\omega\leq b\}, a,b$ rational [counting $(a,\infty)$ and $\Omega$ itself as right-semiclosed]. Let $\mathcal{F}=\sigma(\mathcal{F}_0)$. Then, if $\mu(A)$ is the number of points in $A$ ($\mu$ is counting measure), then $\mu$ is $\sigma$-finite on $\mathcal(F)$ but not on $\mathcal{F}_0$
Actually, he gives the answer: since $\Omega$ is countable union of singletons, $\mu$ is $\sigma$-finite on $\mathcal{F}$. But every nonempty set in $\mathcal{F}_0$ has infinite measure, so $\mu$ is not $\sigma$-finite on $\mathcal{F}_0$.
It seems pretty clear after the 'But': the counting measure in each interval is infinite. But, I didn't understand the first sentence. How can this be possible? If I take any interval in $\mathcal{F}_0\subset\mathcal{F}=\sigma(\mathcal{F}_0)$, it is also in the $\mathcal{F}$. So, if it is infinite on the first one, shouldn't it be in the second one too? I'm confused about this concepts. Could anyone help me?
You are arguing that there are lots of sets on infinite measure in $\mathcal F,$ which is certainly true, but the author's point is that there are sets of finite measure (in particular, the singleton sets) in $\mathcal F.$ Thus, the rationals can be written as the countable union of sets of finite measure in $\mathcal F,$ but as you have noted, this is impossible with $\mathcal {F_0}$