Comments for SQL Fascination

Comment on Oracle : Duplicate GUID values being returned from sys_guid() when run in parallel by Lapuinka

Lapuinka — Tue, 22 May 2012 20:32:45 +0000

Ok my name is Fabio Lapuinka I’m a DBA and .NET Specalist too. I’m having the same problem

Comment on Subtle change to ::fn_dblog in SQL 2008 by Mendolis

Mendolis — Thu, 17 May 2012 22:00:04 +0000

To satisfy your curiosity regarding the question “Why would there is so much interest”. We are looking into this for the possibility of incremental data captures. Sure, you have replication, DML triggers and Change Data Capture. However, this is a vendor supplied database. By contract, if we make ANY change to the database (even non-destructive changes) they cancel our 5 year support contract. All of the methods I mention above modify the database in some way… and, yes, the vendor states that marking the table as a ‘publisher’ for replication is in volation. Sad but true

Comment on Dynamic Partitioning : Wishlist by Gerzson

Gerzson — Fri, 04 May 2012 09:13:11 +0000

I encountered a real word dynamic partitioning situation. We have a fact table in a data warehouse which stores huge amounts of fact data by month. The current month’s data changes frequently and accessed frequently, the previous months data never changes but accessed frequently and the remaining data also never changes and accessed infrequently.

We access the fact table from a partitioned SSAS cube. The Cube has 3 partitions: Current, Previous and Historic data. To be able to process the current data fast the fact table also have to be partitioned the same manner if possible. Current, previous and historic partitions would be nice to create and maintain shifting data from one to the other month by month.

Comment on Oracle : Recursive Common Table Expressions and Parallelism by Andrew Stein (@AndrewStein)

Andrew Stein (@AndrewStein) — Sat, 14 Apr 2012 14:59:50 +0000

I take my hat off to you Andy. As always, excellent work finding these Oracle problems…

Comment on Which User Made That Change? by Sten Miles

Sten Miles — Sun, 18 Mar 2012 17:40:51 +0000

Great little article. I am using Change Data Capture (CDC), which does not have built in auditing, and your solution allows me to add the user specific data. Something that countless articles have described as not possible but is with the availability of the transaction ID at login time. As a note, I am using SQL 2008 EV but my issue may be covered now in the 2012 version.

Comment on SQL Server Denali – Paging by SQL Server Blogs and news for the week 07/01/11

SQL Server Blogs and news for the week 07/01/11 — Mon, 05 Mar 2012 19:24:59 +0000

[...] SQL Server Denali – Paging – A closer look at this new feature courtesy of Andrew Hogg. [...]

Comment on Decoding a Simple Update Statement Within the Transaction Log by Muhammad Imran

Muhammad Imran — Mon, 06 Feb 2012 14:01:35 +0000

Hi Andrew,
I decoded both of them (LOP_MODIFY_ROW,LOP_MODIFY_COLUMNS ).Here is the source code with the solution.

http://raresql.com/2012/02/01/how-to-recover-modified-records-from-sql-server-part-1/
http://raresql.com/2012/02/01/how-to-recover-modified-records-from-sql-server-part-2/

Imran

Comment on Oracle : Duplicate GUID values being returned from sys_guid() when run in parallel by AndrewS

AndrewS — Sun, 22 Jan 2012 17:21:49 +0000

Feeling your pain here…

Comment on Dynamic Partitioning : What objects are using a Partition Schema? (SQL Tuesday #002) by SQLMD

SQLMD — Fri, 20 Jan 2012 18:21:58 +0000

Excellent – you saved me a bunch of time trying to determine if all our partition schemas are being used.

Comment on How to Remember the Next Used Filegroup in a Partition Scheme by Andrew Hogg

Andrew Hogg — Tue, 03 Jan 2012 12:10:11 +0000

Think the style might of got updated, used to wrap the code – fixed it up for now so that all the code is readable.

Comment on How to Remember the Next Used Filegroup in a Partition Scheme by Dave B.

Dave B. — Thu, 29 Dec 2011 23:45:16 +0000

Great article! I cannot see all of your code because some of it is chopped-off.

Comment on How to Remember the Next Used Filegroup in a Partition Scheme by Ahmed B.

Ahmed B. — Mon, 12 Sep 2011 22:36:39 +0000

the only post where I find answer to my question: how to figure out the filegroup flagged as NEXT USED. Thanks for the wonderful article.

Comment on Dynamic Partitioning : What objects are using a Partition Schema? (SQL Tuesday #002) by Dev Nambi (@DevNambi)

Dev Nambi (@DevNambi) — Thu, 01 Sep 2011 15:59:32 +0000

Nicely done! This just saved me about an hour’s worth of digging, since I was trying to figure out how to drop partitioning from a table (including all of the partition scheme objects).

Comment on Finding Next Identity Value, and a Wild Goose Chase. by Martin

Martin — Sun, 15 May 2011 13:14:02 +0000

The relevant columns in the system tables are sys.syscolpars and idtval, I have successfully (just for “fun”) changed the identity property of columns using these on a test instance in single user mode.

This doesn’t get updated in real time however. If you were to do

CHECKPOINT
SELECT * FROM sys.syscolpars WHERE id = OBJECT_ID(‘YourTable’)

You should see the idtval value change.

Comment on Subtle change to ::fn_dblog in SQL 2008 by Microsoft – SQLServer – Page Splits | Daniel Adeniji's Trail

Microsoft – SQLServer – Page Splits | Daniel Adeniji's Trail — Sat, 29 Jan 2011 17:49:03 +0000

[...] Subtle change to ::fn_dblog in SQL 2008 http://sqlfascination.com/2010/07/15/subtle-change-to-fn_dblog-in-sql-2008/ [...]

Comment on Which User Made That Change? by SQL Server Blogs and news for the week 14/01/11 | John Sansom - SQL Server DBA in the UK

Fri, 14 Jan 2011 10:33:46 +0000

[...] Which User Made That Change? – Paranoid about what your developers have been up to in your environment? You should be Here’s one method that you can use to keep an eye on things. [...]

Comment on SQL Internals Viewer by SQL Server Blogs and news for the week 03/12/10 | John Sansom - SQL Server DBA in the UK

Fri, 03 Dec 2010 12:06:58 +0000

[...] SQL Internals Viewer – Here’s a brief look at this cool tool. [...]

Comment on SQL Server Hash Partitioning by Andrew Hogg

Andrew Hogg — Sat, 02 Oct 2010 17:35:02 +0000

I’ll send you an email so you can send me the code to look at, will be easier to diagnose.

Comment on SQL Server Hash Partitioning by developer

developer — Sat, 02 Oct 2010 07:39:50 +0000

So does that mean the underlying structure of a hash-partitioned table cannot be modified later, as it has a schema bound function and an attached computed column?

What I observe is that other columns can be modified at the cost making the computed column nullable .Is there any possible issues that might arise from leaving the computed column as nullable? (It will be only in the schema,in reality there will be data inserted to this column for sure!)
Hope its not going to affect partition elimination of tenant based horizontally partitioned tables in any way.

Comment on SQL Server Hash Partitioning by Andrew Hogg

Andrew Hogg — Fri, 01 Oct 2010 14:41:13 +0000

If I am guessing your situation correctly, the hash function is schema bound and attached to the computed column, and you are now attempting to change the schema on a schema bound object – which is not permitted. The point of the schemabound function is that it can rely on the underlying structure of the table will not change.

Comment on SQL Server Hash Partitioning by developer

developer — Fri, 01 Oct 2010 12:01:43 +0000

I’m trying to implement Partition by modulus and right now facing a peculiar thing!! I’m not able to change “Allow nulls” property of any column in a partitioned table without automatically converting the computed partition id column as nullable. And can’t modify this column later!! why is this so..? Right now my tables does not contain any data..

Comment on SQL Server Hash Partitioning by Andrew Hogg

Andrew Hogg — Thu, 30 Sep 2010 11:17:22 +0000

If they do not use the same partition scheme (what I would call, partition aligned) then the rolling window effect is not possible, whilst the data could be rolled out, the index would still need a massive deletion of entries to be correct, so it will refuse. To make the index partition aligned the partition key must be somewhere within it, but does not have to be the first field, if anything placing it as the first is likely to create a less efficient index. In most cases I would expect to see the partitioned key field as the last field within the partition aligned index.

Comment on SQL Server Hash Partitioning by developer

developer — Thu, 30 Sep 2010 04:53:27 +0000

Thanks!! And I would like to add one more question. How much does creating an index using the same Partition Scheme and Partition Key, help?

Comment on SQL Server Hash Partitioning by Andrew Hogg

Andrew Hogg — Wed, 29 Sep 2010 12:17:49 +0000

With the partition field being used within the where clause, partition elimination is at least possible – which can provide a performance benefit. If you fail to use the partition key as part of the query criteria, then all of the partitions will have to be searched to check for matching data. The real key to the performance gain from partitioning is the cheap cost of ageing out older data.
In terms of improving the performance, partitioning is not a silver bullet, you still have to be mindful of your storage, query plans and indexing strategies.

Comment on SQL Server Hash Partitioning by developer

developer — Wed, 29 Sep 2010 10:14:34 +0000

Hi,
How does partitioning actually improve query performance..? Is it enough that we include the partitioning column in the WHERE clause or is there something more to be done to improve query performance on a partitioned table?

Comment on SQL Server Hash Partitioning by Andrew Hogg

Andrew Hogg — Mon, 13 Sep 2010 14:13:58 +0000

SQL Server 2008 does not natively support composite partitioning, which Oracle does with the partition / sub-partition feature. Oracle supports 9 different composites I believe. It is not equivalent, but you can certainly place a view across multiple partitioned tables, physically creating multiple partitioned tables to represent the sub-partitioning and the view sitting across them all to represent the top level partition. The problem will come in the maintenance and the limit of 1000 partitions that is still in place.

Comment on SQL Server Hash Partitioning by developer

developer — Mon, 13 Sep 2010 11:43:30 +0000

Hi..
Do u know whether SQL server 2008 supports composite partitioning??
Thanks in advance..:)

Comment on Rolling a Partition Forward – Part 2 by andrewhogg

andrewhogg — Wed, 25 Aug 2010 07:16:38 +0000

Users and partitions have nothing in common, unless you are planning to use partitioning for multi-tenancy, which I do not believe is a good use of partitioning. More details here : http://sqlfascination.com/2010/01/05/

Comment on Guidance on How to Layout a Partitioned Table across Filegroups by andrewhogg

andrewhogg — Wed, 25 Aug 2010 07:12:03 +0000

It will depend on the variance and size of the data rows, depending on that variance I would take the disk space penalty, or opt for all partitions into a single filegroup and spread the filegroup’s files across all the disks. That technique has disavantages when it comes to the backup strategy, and will not perform as well since the filegroup will fragment over time.

Comment on Rolling a Partition Forward – Part 2 by developer

developer — Wed, 25 Aug 2010 05:18:21 +0000

Hi,
Can you please suggest what partitioning strategy is to be used for managing a large no:of user accounts?

Comment on Guidance on How to Layout a Partitioned Table across Filegroups by aparna

aparna — Wed, 25 Aug 2010 05:11:38 +0000

Hi,
What is the best strategy of partitioning if the quantity of data in each partition might vary considerably?

Comment on Finding Next Identity Value, and a Wild Goose Chase. by Matthijs

Matthijs — Tue, 27 Jul 2010 21:58:34 +0000

I just saw your post on stack overflow, and it was good. Found this blog in the same google. Also good is that you worked on answering the question, instead of questioning why the question was asked – if you catch my drift.

In my own case I was checking up reports from a user on missing records. I could see in the table no gaps in the ID order, so that told me the records had probably not been added and then physically deleted (there’s no ID re-set in this system). But how would I know if the missing records had been added at the “end” of the table or not? I needed to know the next ID, without actually making a record, to know if there was a “gap” at the “end” of the table.

Comment on Immutable Primary Key vs Immutable Clustered Key by andrewhogg

andrewhogg — Tue, 20 Jul 2010 19:33:50 +0000

You are right in that as a generic concept and the relevance of application to other databases / logical design it would provide more hits on a search engine for good reason – I am more interested in the physical implementation of a table design and impact on the database performance specifically from the SQL Server perspective – the choice of the clustering key is a key decision when implementing the logical design / data model.

The concept of a clustered table is not unique to SQL Server, Oracle’s Index Organised Tables would be akin and the same concepts could be explored there, but I’ve not tried to reverse a redo log as yet.

Comment on Immutable Primary Key vs Immutable Clustered Key by Kamal

Kamal — Tue, 20 Jul 2010 19:16:28 +0000

Hi Andrew,
thanks for this post!
I think the immutability of the Primary Key gets more attention because the reason behind it is logical (and not DBMS specific), whereas for the Clustering Key there is a physical reason (SQL Server specific).
I wouldn’t compare / relate the two reasons. When you model your data you might desire an immutable PK even before knowing what DBMS you’re going to use…

Cheers

–
Kamal

Comment on Guidance on How to Layout a Partitioned Table across Filegroups by andrewhogg

andrewhogg — Thu, 24 Jun 2010 12:29:16 +0000

Creating the initial partition script is not a problem, what you are probably having trouble with the the rolling of the partition window as a new month appears and an old month is to go. I have written a guide covering the harder parts of that process in two parts (part 1, part 2 )

Comment on Guidance on How to Layout a Partitioned Table across Filegroups by Marcelo

Marcelo — Wed, 23 Jun 2010 21:16:05 +0000

Hi
do you have the sample script to do this partition in months?

Comment on Query Parameterization and Wildcard Searches by Mark Pilkington

Mark Pilkington — Wed, 09 Jun 2010 09:34:43 +0000

Excellent…

will have to give this a try and see the results. We seem to have had an increase in ‘%?%’ like queries just lately!

thanks andrew

Comment on Locating Table Scans Within the Query Cache by pinaldave

pinaldave — Thu, 11 Mar 2010 00:23:39 +0000

Thanks,

This is for sure enhancement to original script.

Kind Regards,
Pinal

Comment on Dynamic Partitioning : What objects are using a Partition Schema? (SQL Tuesday #002) by Adam Machanic : T-SQL Tuesday #002: The Roundup

Adam Machanic : T-SQL Tuesday #002: The Roundup — Mon, 08 Feb 2010 19:16:12 +0000

[...] Hogg contributed a post in which he describes the trickery required to query the various DMVs in order to properly do dynamic partitioning. In the end, five views are needed to satisfy what at [...]

Comment on Finding Next Identity Value, and a Wild Goose Chase. by faheem

faheem — Wed, 20 Jan 2010 13:13:41 +0000

very nice

Comment on What is the SQL Server 2008 DateTime2 Internal Structure? by andrewhogg

andrewhogg — Tue, 13 Oct 2009 09:30:32 +0000

Your absolutely right, I’ve ended up copying the wrong thing out of the Query window and duplicated a result that should be different. I’ve updated the post with the correction.

Comment on What is the SQL Server 2008 DateTime2 Internal Structure? by BEL8490

BEL8490 — Tue, 13 Oct 2009 09:14:46 +0000

Thanks for this clarifying insight.

I would suggest two minor editing corrections:

1. The code at the beginning of the article does not have a declaration for the variable @bin1 and should be corrected as follows:
set @bin = CONVERT(varbinary(max), @dt)

2. At about halfway down the following example

’0180/06/06 00:00:00′ => 0x070000000000FFFF00
’0180/06/07 00:00:00′ => 0x070000000000FFFF00

should be corrected as:

’0180/06/06 00:00:00′ => 0x070000000000FFFF00
’0180/06/07 00:00:00′ => 0×070000000000000001

as clearly two different dates can not have the same internal presentation.