It’s usually IMPOSSIBLE, especially if you want a good-sounding vocal track.
Learn to sing, or find a singer/mimic. 
"You can’t un-fry an egg or un-bake a cake, and you can’t un-mix a recording.
\
There is a simple “trick” of subtracting left from right to remove centered vocals (everything that’s identical and in-phase in the left & right channels). That can work “perfectly” if the vocals are centered, but it removes everything in the center (although you can filter to leave the centered bass.)
But, there is no similar simple trick for removing everything except the vocals. The more advanced methods of vocal extraction have the same problem of treating everything in the center the same, so you get vocals along with centered-lead instruments. Plus, the more-advanced FFT methods can give you artifacts (noise/distortion, etc.).
These things can be fun to play with but you rarely get good-usable results.
I imported mp3’s of two tracks of a song, the recorded ‘single’ and the ‘karaoke’ version. I did the invert thing on the instrumental and CTRL A and then Mix & Render and the track still sounds exactly the same?
That can work under ideal studio-conditions… But, the regular and karaoke versions are usually completely different recordings. And even if they aren’t completely different recordings, the mixing & mastering is probably different. So in the real world, subtraction rarely works. And, lossy MP3 compression further complicates things.
If you have the original-studio mix, and the same mix without vocals, it can be done. i.e. It can be done if you made the original recordings and mixed yourself. But, if you have the complete original mix and the music-only mix, you probably already have the vocal-only mix.
\
Here’s an example of what will work and what won’t work… If you record yourself saying “Hello”, you can take an inverted copy and mix to get absolute silence. You can even mix “hello” with background music, and subtract-out “hello” later.
But, if you record yourself saying “Hello” twice and subtract those two different copies, you’ll just hear yourself and your imaginary twin saying “hello” together… It will sound the same as a regular non-inverted mix…