Limiting index rebuild time

Question

Is there a way to limit the time a rebuild index command will run so it will die at a specific time or after a given interval so as not to interfere with a scheduled job?
Joe Torre DBA III County of Riverside, CA

Answer 1

Joe,
Great, thank you for the information.
First, an index rebuild can be done online or offline... but either way, if it's killed then all of the work done is lost. It's all or nothing with a rebuild.
What I would suggest to you is two things:
1. Only perform maintenance on the indexes that absolutely need it. This can be done by investigating sys.dm_db_index_physical_stats()
2. During your maintenance windows (though it's an online operation so it could be done all day) reorganize the indexes. If you need to kill it, the progress stays. It can be used in an iterative fashion as the progress is not lost, even if it's killed.
When you notice you have a good amount of internal or external fragmentation even after reorgs are complete, that's when you will need to do a rebuild and if you can only get so many done at a time so be it.
One last thing I'd like to leave you with. Since T-SQL is completed step by step in a batch, you may want to create something using service broker that can rebuild the indexes. This way you can control how many outstanding items there are at once and take
into account hardware such as CPU and Disk IO, etc. The service broker part allows the procedures to be run asynchronously. That would allow you to potentially get through many more index rebuilds than running them all in one script, potentially, if your hardware
is beefy enough and the server activity is low.
I hope this helps.
Sean Gallardy | Blog |
Twitter