Let $(X,\mathcal{M})$ be a measurable space. A measure $\mu: \mathcal{M} \to [0,\infty]$ is $\sigma$-finite if $X=\cup_i X_i$ where $X_i\in \mathcal{M}$ and $\mu(X_i)<\infty$ for each $i\ge 1$. A measure $\mu$ is said to be semi-finite if for any set $E\in \mathcal{M}$ with $\mu(E)=\infty$, one can find $F\subset E, F\in \mathcal{M}$ such that $0<\mu(F)<\infty$.
I read one result that and its proof
every $\sigma$-finite measure is semifinite.
The proof just takes $E=\cup_i (X_i\cap E)$ and takes $F:=X_i\cap E$, which satisfies our condition.
I am working on a similar question that if $\mu$ is $\sigma$-finite with $\mu(X)=\infty$, then for any $\epsilon>0$, there is a set $F\in \mathcal{M}$ so that $\epsilon<\mu(F)<\infty$.
I try to use a similar idea of above proof: Fix $\epsilon>0$. Let $E\in \mathcal{M}$ with $\mu(E)=\epsilon$. Consider the cover $E=\cup_i(E\cap X_i)$ and $X_i$ is as in definition of $\sigma$-finite. Let $E_j:=E\cap X_j$. Then $\mu(E_j)\le \mu(X_j)<\infty$.
However, we will get $0<\mu(E_j)<\epsilon$ for some $j$ with $E_j\in \mathcal{M}$...