One of the best things about books online is the free code.

If you look up dbcc showcontig you will see a routine on the help page that retrieves the table output of the dbcc and then runs index defrag against any index fragged at or above 30%

this is easily modified to be a rebuild or a reindex.

I have a server where there is heavy file fragmentation but it is quite busy and cannot be down for long periods of time. Is there any way other than shutting down SQL Server that would allow me to run a defrag on the datafiles while the the database files are in use?

Hello!!!

I have been in trouble with my SQL Query that was causing time out on the UI...Investigation with DBCC showcontig reveiled that the Scandensity for these tables are 50% or less.

But to my surprise the Scandensity was not improving with REINDEXing and DEFRAG....(Number of rows in these tables are a few hundred only) . I am not sure why is this happening....

to some how escape I tried moving these tables to new ones and saw some improvement in the Scandensity (Query to a improved slightly).

But what puzzles me is why REINDEXing and DEFRAG did't work????

Also one more question....

If I make a backup of this database and restore into a new server will that create new data/index pages?? does that have the same fragmentation as when it was backed up??? or does the RESTORE just uncompress the old pages???

Thanks

Cheriyan.

DisKeeper claims to be able to defrag locked files.

I've used it a couple times in production.

A very nice feature it has is that it monitors the IO queue. If the queue rises, it backs off and allows the system more time on the disks. The effect is zero (Near zero) impact while doing a defrag. The con is your defrag could take forever. But you have the time...

On a whole, a good article.

One correction though.

When CREATE INDEX, WITH DROP_EXISTING is executed on a clustsered index, all non clustered indexes have to be rebuild.

A NC index contains keys to the clustered index, i.e. the data. When you rebuuild this, the data could, and probably will, move to another page. This would cause the key link to become invalid.

See BOL:

CREATE INDEX, DROP_EXISTING

"Because nonclustered indexes contain the clustering keys, the nonclustered indexes must be rebuilt when a clustered index is dropped. If a clustered index is recreated, the nonclustered indexes must be rebuilt to take the new set of keys into account"

I learnt this the hard way...

To try and answer your first question:

If you only have a few rows (couple could be a few) in a table, SQL creates those pages on an extent which is shared by other pages. An extent is 64k, page 8. If you have a table with narrow rows and only a couple hundred rows, the chance is good you using a mixed extent. With this in mind, SQL may not defrag it because a) it's so small, b) there's other data sitting on the extent.

Try adding a couple thousand rows to the table (Assuming you can) and see if that changes things.

See BOL, pages and extents for more info.

To your second question, yes.

When SQL does a restore, it restores all tables. One of them being sysindexes. Sysindexes determines on what pages / extents sit. DUring a restore, SQL acquires a chuck of disk space. It then creates all the extents / pages. Once done, it starts writing out the data page by page. (or extent by extent?).

Restoring a backup will only give you contiguous OS space, not within the SQL files.