Go Premium for a chance to win a PS4. Enter to Win

x
  • Status: Solved
  • Priority: Medium
  • Security: Public
  • Views: 389
  • Last Modified:

to compensate for not being able to drop NC index online

when dropping an NC index, because you are not able to keep the index online, what are the other alternatives, to avoid for the incoming queries to be locked out?

thanks
0
anushahanna
Asked:
anushahanna
  • 9
  • 6
  • 2
5 Solutions
 
rajeshprasathCommented:
ONLINE = ON

ON
Long-term table locks are not held. This allows queries or updates to the underlying table to continue.

Specifies whether underlying tables and associated indexes are available for queries and data modification during the index operation. The default is OFF.

check the following link,
http://msdn.microsoft.com/en-us/library/ms176118.aspx
0
 
anushahannaAuthor Commented:
that is available only for clustered index; what options are advised for NC index?
0
 
rajeshprasathCommented:
sorry, online = ON option works only with clustered index.

Only a clustered index can be dropped online.

Non Clustered index cannot be dropped online.
0
Technology Partners: We Want Your Opinion!

We value your feedback.

Take our survey and automatically be enter to win anyone of the following:
Yeti Cooler, Amazon eGift Card, and Movie eGift Card!

 
anushahannaAuthor Commented:
>>Non Clustered index cannot be dropped online.

any alternatives to avoid the locking? (when user queries hit the underlying table)
0
 
dqmqCommented:
Please clarify.  Dropping an NC is pretty fast and should be infrequent.  So, why the concern about locks for the duration?  

The larger concern, in my mind, is an excution plan that relies on the NC index is no longer valid after the index is dropped.  At best, the SQL requires re-optimization; at worst, it will perform awful after the re-optimization.

 
0
 
anushahannaAuthor Commented:
dqmq, while restoring a database in another box, there is no need for a bunch of NC indexes in the new environment, and they all get dropped together at one time, before users start hitting the database tables again. in that short time between the indexes being dropped, if a query hits the tables under the index, it causes endless locks..
0
 
dqmqCommented:
Do you see the apparent contradiction:

"before users start hitting the database tables...if a query hits the tables..., it causes endless locks..."

What exactly is the sequence of events, for example:

Backup Database (Old)
Restore Database (New)
Users Query (New)  
Drop All Indexes (New)  

0
 
anushahannaAuthor Commented:
Backup Database (Old)-Box A
Restore Database (New)-Box B
Drop All Indexes (New) -BoxB
Users Query (New)  --While Step #3 is not yet finished, people start step #4, and hence the locks.
0
 
dqmqCommented:
So, you restore the indexes in step 2 THEN drop them in step 3?   does that make sense?
0
 
anushahannaAuthor Commented:
yes; because backup/restore is a package- we do not say what we need or not; so we get them all in, and then remove what we do not need...
0
 
dqmqCommented:
I can accept at face value that you do not "need" the indexes. But, aside from some disk storage, what harm do they cause?  I could understand if you did not want to take the time to load the indexes, but you load them then drop them.  Seems like the long way around the block.  Further, compiled execution plans that came along with the restore and rely on the dropped indexes will be rendered invalid.  

One of the classical uses for filegroups is for organizing backup/restore processes into usefult subsets.  You may want to consider that for isolating the indexes that you don't want.
0
 
anushahannaAuthor Commented:
dqmq, thanks for your wise input. I will take your word, and focus on fg.

but in the current setup, would you think there are any alternatives to avoid the locking?
0
 
dqmqCommented:
My first response is to stop dropping the indexes.  And I don't say that to be funny. Weigh the cost of dropping indexes (high for now) with the benefit (not very much, it would seem) and see where that assessment takes you.  At a minimum, you might spread out index dropping or defer till a time locking is not such an issue.

Having said that, if you want to pursue it further, here's my quick synopsis.  Dropping NC indexes is a relatively quick operation and should not involve long term locks.  Actually, I question whether it involves locks on the underlying table, at all!  So start with identifying exactly what resource is getting locked, which process is locking it and which processes are getting blocked.  It very well may be that queries are blocking the drop index command and that in turn has a cascading effects.   If so, another strategy might be to hold off running the queries until the NC indexes are dropped.

Also, the fact that there is some "locking" relationship between the NC index and the queries makes a suggestion that queries are dependent on the index in some way.  So, beware of that!



0
 
anushahannaAuthor Commented:
dqmq,

>>Dropping NC indexes is a relatively quick operation and should not involve long term locks

I put the DB on single user mode and did it, still took an hour. (90 indexes in all).

is the size of the index a factor in this?
0
 
dqmqCommented:
Wow! Apparently index size does matter.  I had one more thought.  You might see what happens with a two step process:

single user mode
alter index indexname disable
alter index indexname disable
...
multi user mode/allow queries
drop index indexname
drop index indexname
....
0
 
anushahannaAuthor Commented:
Thanks for the idea. Will try and let you know.
0
 
anushahannaAuthor Commented:
this was quiet efficient. Thanks dqmq for the patient guidance.
0

Featured Post

Free Tool: IP Lookup

Get more info about an IP address or domain name, such as organization, abuse contacts and geolocation.

One of a set of tools we are providing to everyone as a way of saying thank you for being a part of the community.

  • 9
  • 6
  • 2
Tackle projects and never again get stuck behind a technical roadblock.
Join Now