I thought that since the function $f(x,y)=(x^2+y^2)^{-1}$ is not integrable in $\mathbb{R}^2$ (I mean away from the origin) the same integral but with counting measure may behave the same. I tried this approach (similar to polar coordinates in some sense) $$ \sum_{n,k\geq 1}\frac{1}{n^2+k^2}=\sum_{t\geq1}\frac{r(t)}{t} $$ where $r(t)$ is the function that counts how many ways an integer $t$ can be written as sum of two squares. Does it exists any esteem or asymptotic behaviour for this function? Or maybe a different way to solve the problem?
Thanks in advance!
You're definitely on the exact right track: It's not summable for the exact same reason that $(x^2 + y^2)^{-1}$ is not integrable (and that basically gives a proof: the counting measure just looks like a discrete version of the integral, and by breaking up the plane into a union of squares of side-length $1$ this argument can be made very precise).
As far as your question on the asymptotics of $r$, see this question here.