Part Time Database Administration

As many of you know I started a new part time database administration company a few months back and its been going great.  If you're in the need for a database administrator but can't justify hiring an FTE please check us out.  http://www.yourparttimedba.com/p/part-time-sql-server-database.html

-Jeff

How to tell what tables in a database are compressed

Run this query and then run the output of the query (remember to drop off the last union).


select 'select''' + name + ''' as dbnme, st.name, st.object_id, sp.data_compression_desc FROM ' + name + '.sys.partitions sp INNER JOIN ' + name + '.sys.tables st ON st.object_id = sp.object_id WHERE data_compression <> 0 union'
from master.sys.databases

Query to identify stored procedure table dependencies

If you're looking to identify what stored procedures have a dependency on a table the query below will help you to find them.  If your stored procedures span multiple databases this query will still pick up dependencies, however you will need to run it in each individual database.

SELECT d.referencing_id,

    QUOTENAME(s.name),

    QUOTENAME(o.name),

    QUOTENAME(d.referenced_database_name),

    QUOTENAME(d.referenced_schema_name),

    QUOTENAME(d.referenced_entity_name),

    d.referenced_id,

    o.type_desc

FROM sys.sql_expression_dependencies AS d

INNER JOIN sys.objects AS o ON d.referencing_id = o.[object_id]

INNER JOIN sys.schemas AS s ON o.[schema_id] = s.[schema_id]

How to identify a lead blocker in SQL Server

If you’ve ever logged into your SQL Server database only to be greeted by greeted by hundreds or thousands of active spids all of which are being blocked there’s an easy way to identify which blocker is the lead blocker by using the query below.  In some cases there may be multiple lead blockers, this indicates that there are several spids running which aren’t being blocked by any other processes.  To find the one causing the most harm by looking at the last_batch, cpu, and physical_io. 

select blocker.spid, last_batch, cpu, physical_io

from master.dbo.sysprocesses blocker

inner join master.dbo.sysprocesses victim on blocker.spid = victim.blocked

where blocker.blocked = 0

How fast is my SQL Server database growing?

Do you know how quickly your database is growing?  If not you can get a pretty good idea based on how the size of your full backups have grown over time. 

Run the script below, and then run the output of that script and you will see the growth of your backup day by day.  Putting the output into Excel can help to generate a graph for this.  Overall it’s a quick and dirty way to get some idea of how quick you’re growing.

use master

go

DECLARE @dbname as sysname

DECLARE @select_string as varchar(4000)

DECLARE @for_string as varchar(4000)

DECLARE @init as bit

DECLARE curs CURSOR FOR

SELECT name FROM sysdatabases

OPEN CURS

FETCH NEXT FROM curs INTO @dbname

SELECT      @select_string = 'SELECT DISTINCT backup_date',

      @for_string = 'FOR database_name IN (',

      @init = 0

WHILE @@FETCH_STATUS = 0

BEGIN

      IF @init = 0

      BEGIN

            SET @init = 1

            SELECT      @select_string = @select_string + ', [' + @dbname + '] as ' + '"' + @dbname +'"',

            @for_string = @for_string +  @dbname

      END

      ELSE

      BEGIN

            SELECT      @select_string = @select_string + ', [' + @dbname + '] as ' + '"' + @dbname +'"',

                  @for_string = @for_string + ', ' + @dbname

      END

      FETCH NEXT FROM curs INTO @dbname

END

CLOSE curs

DEALLOCATE curs

PRINT @select_string

PRINT 'from

      (select isnull(backup_size / (1024 * 1024 * 1024), 0) as backup_size,

      database_name,

      cast(cast(backup_finish_date as varchar(12)) as datetime) as backup_date

      from msdb.dbo.backupset

      where type = ''d'') p

pivot

      (

      sum(backup_size)'

PRINT @for_string

PRINT ')) as pvt 

order by backup_date'

Odd Behavior Using Partitioned Tables in a Replicated Database With SQL Server 2008

I ran into an odd problem today troubleshooting an error trying to switch a partition in SQL Server 2008.  The error read:
  
  Msg 137, Level 15, State 2, Line 1
  Must declare the scalar variable "@VARIABLE".

This was a particularly odd error since the variable was indeed declared and the process worked properly in the non production environment.  After troubleshooting for a while I found that the problem was that you cannot use a variable to specify the partition as in the statement below if you have your database published for replication

  ALTER TABLE PartitionTable SWITCH PARTITION @partition TO NonPartitionTable

Microsoft's workaround to this was to use dynamic SQL.  The full KB article on the problem can be found here: http://support.microsoft.com/kb/2002474

Create and attach script for a database prior to detaching it

If you are going to detach a database in SQL Server that you’re planning on attaching later save yourself some trouble and generate the attach script first.  Running the code below will generate a script to be used to attach your database at a later time (assuming that you keep the database files in the same location that they were in when you detached it).  Make sure you’re in the database you’re going to detach when you run this script

declare @dbname as sysname

declare @sql as varchar(1000)

declare @n as integer

declare @filename as sysname

set @dbname = db_name()

declare curs cursor for

select rtrim(filename) from sysfiles

open curs

fetch next from curs into @filename

set @n = 0

print '--ATTACH SCRIPT FOR ' + @dbname

print 'exec sp_attach_db @dbname = N''' + @dbname + ''','

while @@fetch_status = 0

begin

      set @n = @n + 1

      print '@filename' + cast(@n as varchar) + ' = ''' + @filename + ''','

      fetch next from curs into @filename

end

close curs

deallocate curs

Identify Duplicate Indexes

Indexes make things quicker in the databases but there’s not much uses for having two of the same indexes.  If you have duplicate indexes they’re likely having a negative effect on your database since they’re causing the following problems:

·         Maintaining duplicate indexes increases your overhead associated with inserts, updates, and deletes. 

·         Reindexes/Reorgs of duplicate indexes takes valuable time out of your maintenance window.

·         Increase storage costs since you have two of the same thing

You can identify duplicate indexes in your database by running the following query (note: you must be in the database you wish to run it on).

with AllIndexes (objectname, indexname, keyno, columnname)

as

(     select so.name as objectname, si.name as indexname, sik.keyno, sc.name as columnname

      from sysindexes si

      inner join sysobjects so on so.id = si.id

      inner join sysindexkeys sik on sik.id = so.id and sik.indid = si.indid

      inner join syscolumns sc on sc.id = so.id and sc.colid = sik.colid

      where so.xtype = 'u')

select objectname, indexname, [1] fld1, [2] fld2, [3] fld3, [4] fld4, [5] fld5, [6] fld6, [7] fld7, [8] fld8, [9] fld9, [10] fld10

into #temp

from (select objectname, indexname, keyno, columnname from AllIndexes) t

pivot (max(columnname) for keyno in ([1], [2], [3], [4], [5], [6], [7], [8], [9], [10])) as pvt

select t1.objectname, t1.indexname, t2.indexname

from #temp t1

inner join #temp t2 on

      t1.objectname = t2.objectname and

      t1.indexname < t2.indexname and

      isnull(t1.fld1, '') = isnull(t2.fld1, '') and

      isnull(t1.fld2, '') = isnull(t2.fld2, '') and

      isnull(t1.fld3, '') = isnull(t2.fld3, '') and

      isnull(t1.fld4, '') = isnull(t2.fld4, '') and

      isnull(t1.fld5, '') = isnull(t2.fld5, '') and

      isnull(t1.fld6, '') = isnull(t2.fld6, '') and

      isnull(t1.fld7, '') = isnull(t2.fld7, '') and

      isnull(t1.fld8, '') = isnull(t2.fld8, '') and

      isnull(t1.fld9, '') = isnull(t2.fld9, '') and

      isnull(t1.fld10, '') = isnull(t2.fld10, '')

drop table #temp

Find Fragmented Indexes in SQL Server

Identifying the most fragmented indexes in a particular database can be found by running the below query.  Only indexes in the active database will be analyzed.  The results will be ordered by the most fragmented table being listed first.

declare @table_index table (table_name sysname, table_id int, index_name sysname, index_id int, fragmentation float, pagecount int)

declare @table_id as int

declare @index_id as int

declare @fragmentation float

declare @pagecount int

insert into @table_index (table_name, table_id, index_name, index_id)

select o.name as table_name, o.object_id as table_id, isnull(i.name, 'HEAP') as index_name, i.index_id

from sys.objects o

inner join sys.indexes i on o.object_id = i.object_id

where o.type = 'u'

declare curs cursor for

select table_id, index_id

from @table_index

open curs

fetch next from curs into @table_id, @index_id

while @@FETCH_STATUS = 0

begin

      select @fragmentation = avg_fragmentation_in_percent,

      @pagecount = page_count

      from sys.dm_db_index_physical_stats (DB_ID(), @table_id, @index_id,null,'limited')

      update @table_index

      set fragmentation = @fragmentation,

      pagecount = @pagecount

      where index_id = @index_id

      and table_id = @table_id

      fetch next from curs into @table_id, @index_id

end

close curs

deallocate curs

select *

from @table_index

order by fragmentation desc

Identifying the largest HEAPS in your database

Running the below script will identify list off all HEAPs in your database (largest HEAPs will be listed first).

DECLARE @dbname as sysname

DECLARE @sql as nvarchar(1000)

DECLARE @heaps as table (dbname sysname, objectname sysname, create_dte smalldatetime, size_mb int)

DECLARE curs CURSOR FOR

SELECT name FROM master.sys.databases

WHERE state = 0 --online

and name <> 'TEMPDB'

OPEN curs

FETCH NEXT FROM curs INTO @dbname

WHILE @@fetch_status = 0

BEGIN

      SET @sql = 'USE ' + @dbname + char(10) + 'SELECT db_name() as dbname, o.name, o.create_date, (ps.in_row_used_page_count * 8096) / (1024 * 1024) as size_mb

      FROM sys.indexes i

      INNER JOIN sys.partitions p on p.object_id = i.object_id and p.index_id = i.index_id

      INNER JOIN sys.dm_db_partition_stats ps on ps.partition_id = p.partition_id

      INNER JOIN sys.objects o on o.object_id = i.object_id

      WHERE i.type = 0

      and o.type = ''U'''

      INSERT INTO @heaps (dbname, objectname, create_dte, size_mb)

      exec sp_executesql @sql

      FETCH NEXT FROM curs INTO @dbname

END

CLOSE curs

DEALLOCATE curs

SELECT *

FROM @HEAPS

order by size_mb desc

Monitoring Deadlock in SQL Server

 SQL Server can passively write deadlocks to the log.  To do this you can turn on two traces:

                DBCC TRACEON(1222,-1)

DBCC TRACEON(1204,-1)

Trace 1204 Returns the resources and types of locks participating in a deadlock and also the current command affected

Trace 1222 Returns the resources and types of locks that are participating in a deadlock and also the current command affected, in an XML format that does not comply with any XSD schema.

For more information on trace flags refer to http://msdn.microsoft.com/en-us/library/ms188396.aspx

How to get the runtimes of individual steps of an SSIS package

Open up the SSIS package which you want to add logging to and follow these steps:

·         Right click anywhere in the design screen and click on “Logging”

·         In the “Configure SSIS Logs: PROJECT NAME”  on the “Providers and Logs” tab select “SSIS Log Provider for SQL Server” and Click “Add”

·         In the newly added item in the list box choose or add a new configuration for the log

·         In the Details tab select the “OnPostExecute” and “OnPreExecute” events.

·         Click OK to save your changes

The next time you run your SSIS package a new table named sysssislog will be created.  By running the query below you will be able to view the start and end times of the different pieces of your SSIS package

declare @start_of_run as smalldatetime

set @start_of_run = '01/02/2013 14:47'

select pre.source, pre.starttime, post.endtime, isnull(cast(DATEDIFF(second, pre.starttime, post.endtime) as varchar),'In Progress') as run_time_seconds

from sysssislog pre

left outer join sysssislog post on pre.source = post.source and pre.executionid = post.executionid and post.event ='onpostexecute'

where pre.starttime > @start_of_run

and pre.event = 'onpreexecute'

order by pre.starttime