To this patzer's eyes, it looks like you gifted your opponent the half-open a-file and then that knight outpost on b6, all on the side of the board where he already had a space advantage. Then, your king went mental and ran over to that side of the board to "help." Probably should have castled kingside around move 14 if not earlier.
EDIT: cxb was where it really all fell apart for you, hanging a pawn with check.
This one's over my head I'm sure. I'm definitely not an FM.
Moves 5 and 6 look like they're setting up castling kingside, but then you damage that by exchanging bishops and opening white's kingside rook. Why capture on move 7 if you didn't on move 6?
Would have been good to play h6 at some point to give your light-square bishop an escape square. Without it white can exchange his knight for your bishop whenever he wants.
With the queens staring at each other, I think move 9 should have been capturing the c pawn to prevent that push. Once the pawn pushed, I think retreating the queen to c7 would have been better than capturing and giving your opponent another rook file.
Move 17, smothering your rook with your other rook, is the most obviously bad move. They can't support each other from there, and the a rook can't avoid any kind of tactics. If you're setting up a pawn push, better to push immediately and move the rook later.
Don't know but white is probably not a great player either.
17...Rhb8 is very passive. You had to play 17...a6 to free Ra8 from babysitting pawn a7 and play a more active role.
In general White has played well, and it is hard to say which move was your exact mistake.
From strategical point of view:
White peaces were more active at the start and at the end of the game. Because:
1, You have exchanged your black colored bishop for his one. Which means that you have exchanged good bishop for a bad one. He has exchanged his good bishop for your Knight later in the game, but it was different - in closed endgames Knights are stronger than Bishops.
2, Pawns on a and h file were a weakness for Black - they were attacked and defended by the same number of pieces. The two half open files were an advantage for White. Due to them your rooks were fixed on A and H files.
3. In Rook endgame the side that attacks the opponent pawns has some advantage.
4. At the end of the game White has a way to progress, and have used it very well.
From Opening theory: In Queen Gambit there are two plans for White. The one of them is Minority attack. Minority attack means more or less advance of b pawn. You haven't done anything to prevent it in the early stage of the game.
You have played well.
Yeah, you had to retract your 17th move right away. Once you play 19... Ra7 you are in effect the exchange down.
I like 17... a6. If the knight plays for a4 you will move your queen rook then put your king on c7 and, although the hopper looks impressively posted, it's really just hitting air.
Incidentally, if the idea of 17... Rhb8 was to impede White's b5, then that does have a nice Petrosian-type feel to it. Trouble is, it takes two moves from White--b5 and bc--in order for it to become effective. If White had to make those moves in order to get play, it would be a cunning stratagem; but he does have other options, and the two rooks end up stumbling over each other a bit.
"Where did I lose this game?"