This is question 10.2 in "Measures, Integrals and Martingales". The answer, which can be found here seems to contain a typo, and also I don't see why they pick such a complicated counterexample. I would like to know if the approach taken in the solutions would work, and also if my counterexample is sufficient.
The question: Let $(\Omega,\mathcal{A},P)$ be a propability space. Find a counterexample to the claim: every P-integrable function $u \in \mathcal{L}^1$ is bounded. [Hint: you could try to take $\Omega = (0,1)$, $P = \lambda^1$, and show that $u = 1/\sqrt{x}$ is Lebesgue integrable on (0,1) by finding a sequence of suitable simple functions that is above u on, say, $(1/m,1)$, and then let $m \to \infty$ using Beppo Levi]
Now in the solutions, they use $u(\frac{j+1}{n}) = \sqrt\frac{j+1}{n}$, which should be the inverse and thus the following convergence argument doesn't apply.
But apart from that, why not just take an atom A (set with no subsets in the $\sigma$-algebra), construct a measure $\mu$ such that $\mu(A) = 1$, $\mu(A^c) = 0$ and set $u$ to 1 on $A$ and to infinity on $A^c$? That would be sufficient to falsify the claim right?