Your mistake is that you have the weight of all of the water in the sphere being lifted from every height in the sphere: in other words, the same weight is being multiple times.
Analyse the problem using things that are constant- that's one reason for emphasizing "Riemann sums" as the definition of integral: that's often how we set up integrals in applications.
As Unco suggeted, draw a circle representing the sphere. Put it in "standard position" in a coordinate system so the center is at (0,0) and the radius is 3: [itex]x^2+ y^2= 9[/itex]. Draw a horizontal line at any level in the circle, at y, representing one "layer" of water, with thickness "dy". All the water in that "layer" is at the same height, y, and so must be lifted the same distance, 3- y (notice that y may be from -3 to 3 so this distance is from 3-(-3)= 6 at the bottom of the tank to 3-3= 0 at the top).
Now, in 3 dimensions, that "layer" of water is a disk whith radius equal to the x coordinate where the line at y crosses the circle: [math]x^2+ y^2= 9[/math] so "radius squared" is [itex]]x^2= 9- y^2[/itex] and the area of the disk is [itex]\pi(9- y^2)[/itex]. The volume of the thin layer is [itex]\pi(9-y^2)[/itex] and since weight density of water is 1000(9.81)= 9810 Newtons/l3, the weight of that layer is [itex]9810\pi(9- y^2)dy[/itex] and, finally, the work done lifting that weight a distance 3- y is [itex]9810\pi(9- y^2)(3- y)dy[/itex].
You get work required to lift all of the water to the top of the sphere by "adding" the work done on every "layer" of water in the sphere. That is, you integrate that from y= -3 to y= 3.