Viewing 15 posts - 7,396 through 7,410 (of 7,597 total)
An occassional highly unusual exception doesn't validate the horrible idea in general of identity as the default for a clus key.
May 31, 2012 at 2:30 pm
True, there could be special situations.
you could potentially have a very large and active table that is in a database mirrored over a slow wan connection.
If you're trying to mirror...
May 31, 2012 at 2:14 pm
Other than logging-type tables, where the clus key value is essentially irrelevant, I can't imagine any case where identity would be the proper clus key on business data.
Sure, for some...
May 31, 2012 at 2:05 pm
There are vastly better ways to stress that in general a table needs a clustered key than to have the damaging and false notion that it should be an identity...
May 31, 2012 at 1:57 pm
While I may agree, I have to play devils advocate, why not cluster on the identity column?
Because the overwhelming majority of queries will specify the datetime.
I wish I had a...
May 31, 2012 at 1:49 pm
As long as you do things in the proper sequence -- change AD first, then use the SQL tools, not the Windows tools, to change the account pwd -- you...
May 31, 2012 at 1:40 pm
*Cluster* the table on the add date/time (rather than on, say, an identity column, which may still be a good idea to have on this table, just don't cluster by...
May 31, 2012 at 1:18 pm
Create a logon trigger, and capture everything you can -- app_name, host_id, host_name (most people don't bother to fake this) and original_login (just in case).
You could of course then also...
May 31, 2012 at 1:15 pm
How about something like this?:
SELECT
/*@date1 = */ last_friday,
/*@date2 = */ DATEADD(DAY, -07, last_friday),
/*@date3 = */ DATEADD(DAY, -14, last_friday),
...
May 31, 2012 at 1:02 pm
Interesting. I used a start-up proc to set the flag, so it was almost immediately on.
May 31, 2012 at 9:34 am
It claims it does, but seems to work only under certain conditions.
May 30, 2012 at 4:40 pm
Not a lot to go on, but I'll take a shot:
SELECT col1, col2
FROM dbo.tablename
GROUP BY col1, col2
HAVING COUNT(DISTINCT col1) = (SELECT COUNT(DISTINCT col1) FROM dbo.tablename)
May 30, 2012 at 4:14 pm
I would think that object name clashes would be a potential concern. If they occurred, resolving those could theoretically involve a LOT of work.
May 30, 2012 at 3:04 pm
Tables with fewer than eight pages can see increased fragmentation when rebuilt, because SQL may spread the table across more extents.
Personally I don't like SQL's mixed extents, but I can't...
May 30, 2012 at 3:01 pm
Viewing 15 posts - 7,396 through 7,410 (of 7,597 total)