Thanks for the comment it’s always nice to hear that a blog post (even a 2 year old one) has helped someone along the way.

Personally I always leave auto create and update stats on unless there was a specific reason to turn it off. I’ve found that the “auto update stats” doesn’t catch everything I want which is why I do things manually as well but it DOES do some of what should be done and for that reason I leave it on.

Thanks again for reading!

I used to rebuild all indexes, then update stats (with fullscan). It was very time consuming, and somewhat redundant since the index rebuild updated index statistics anyway. Then one day I actually read up on the UPDATE STATISTICS command and found that I could use it to only update column statistics. Ah, the power of reading.

So now my weekly plan of attack is to rebuild indexes that are more than 20% fragmented and have more than 1000 pages, followed up immediately by updating column statistics only:

sp_MSForEachTable “UPDATE STATISTICS ? with fullscan, columns”

As Amar mentioned in the comments above, I still have auto update stats on. If I’m doing the index rebuild/update stats on a weekly basis, should I turn off auto update stats?

I ran the following:

Delete from dbo.tblIndexStatsMaintenance where [ID] > 2000

then ran:

DBCC SHOW_STATISTICS (‘tblIndexStatsMaintenance’, stats_NonIndexedData) WITH STAT_HEADER

which showed no change… (as expected) then ran:

EXEC sp_updatestats

and when I reran the SHOW_STATISTICS I saw the statistic was updated as expected.

I re-read your comment… and I missed rebuilding the index after the deletion of the records… I re-ran to match your use case exactly and low and behold I saw the same thing you saw.

Very weird and I can’t explain right now as to why this is. I re-ran the usecase both ways a few times and got the same results each time.

Hopefully someone else on the thread will be able to shed some light on this one. I’ll also throw it back out there to Twitter to see if we can get some thoughts on why this would happen.

Thanks for the comment!!

This is an example where obviously the stats should have been updated, but weren’t.

So should I do the sp_updatestats *before* the reindex on sql2000?

Also as Grant mentioned you’ll want to be cautious on how you update your statistics AFTER an index rebuild has taken place… cause if your sample size is smaller then a full scan then you might actually “get good statistics turned bad”

Hopefully this helps and I’ve re-ran my code/test shown above again just to be sure but that “demo” should give you a hands-on solution to play with.

Thanks!!

Ross