Arsh - Wednesday, August 16, 2017 1:56 PM
Good lord, NO! 😉 Not even as a last option. The type of fragmentation caused by shrinking an MDF requires that a defrag be done (it's not like the normal rather well behaved fragmentation you would get just from normal usage). The type of fragmentation done by a shrink WILL absolutely kill performance even for some of the small stuff. If you do index rebuilds to do the defrag, your MDF will grow by ~120% of whatever your largest index was... which could and probably will be the clustered index on your largest table. If that's a 30GB object, your MDF will likely grow by 36GB. If you do reorgs to fix the fragmentation, your log file could easily grow by that much or more depending on how often you do backups, etc.
There are ways to do a different type of "shrink" but, guess what? They also require extra disk space until all the data/index pages have been moved.
What really needs to happen is that Aaron's hardware group needs to work with him to make this possible. I know it's not free but buying extra hard disk space is a hell of a lot cheaper than spending a man-month two trying to figure out a solution that may simply fail or, worse yet, one that doesn't fail but isn't correct.
--Jeff Moden
Change is inevitable... Change for the better is not.