This procedure should do what you want:
create procedure dbo.dbrestore
@dbname nvarchar(50)
as
declare @fileno integer
declare @dumpdevice nvarchar(50)
declare @mediasetid integer
declare @position integer
select top(1) @dumpdevice=a.logical_device_name, @mediasetid=b.media_set_id, @position=b.position from backupmediafamily a inner join backupset b on a.media_set_id=b.media_set_id
where b.database_name='backuptest' and b.type='D' order by a.media_set_id desc, b. position desc
declare restore_cursor cursor for
select position from msdb.dbo.backupset where database_name=@dbname and type='D' and media_set_id=@mediasetid and position=@position
union
select max(position) as position from msdb.dbo.backupset where database_name=@dbname and type='I' and media_set_id=@mediasetid and position>@position
union
select position from msdb.dbo.backupset where database_name=@dbname and type='L' and media_set_id=@mediasetid and position>@position
and last_lsn>(select max(last_lsn)from msdb.dbo.backupset where database_name=@dbname and type='I' and media_set_id=@mediasetid and position>@position)
order by position asc;
open restore_cursor
fetch next from restore_cursor into @fileno
while @@FETCH_STATUS = 0
begin
--comment the print statement and uncomment the exec statement to run
--exec('restore database ['+@dbname+'] from '+@dumpdevice+' with norecovery, replace, FILE='+@fileno+';')
print ('restore database ['+@dbname+'] from '+@dumpdevice+' with norecovery, replace, FILE='+convert(varchar(4),@fileno)+';')
fetch next from restore_cursor into @fileno
end
close restore_cursor
deallocate restore_cursor
--exec('restore database ['+@dbname+'] with recovery;')
go
execution goes:
get the name of the dumpdevice based on the supplied database name. I'm assuming that they are all on the same device. The top is to work around having multiple media sets per database if using 'with format' for the initial DB, and to cope with the possibility of multiple full backups in the device
Then build up a union query consisting of:
The last full database backup (type=D) in the device
the last differential backup (type=I and has the greatest LSN (Log Sequence Number)) taken after the last full db backup.
the transaction logs (type=L) taken after the above diff.
returning just the position value, which maps to the FILE value in this case (http://msdn.microsoft.com/en-us/library/ms186299.aspx)
Loop through the result restoring them individually, and then a final restore with recovery to finish it off.
Call it as:
exec dbrestore 'db2restore'
I've tried it on a test db and it appears to work happily, and comparing it to the files SSMS wants to restore for the same operation it looks the same.
Provided as is, there's probably typos, I've probably missed a possible boundary condition, and there's no error checking.
Carlo,
do you have other ideas?
You jus took a backup, so when you write the execute SQL Statement part of your job reference the MSDB backup history information tables to pull out the latest full and any other diffs/logs/etc that you would like to apply. The data can be found in msdb under the dbo schema and will give you the file names, etc, so there will be no need to pass anything and you can make it work should you want to do it from the last full or from the last full with the rest of the backup chain.
http://msdn.microsoft.com/en-us/library/ms188062.aspx
Best Answer
Assuming you run the same version of SQL Server, Yes. What you're looking for is called RESTORE WITH STANDBY:
Every time you apply another log backup users will need to be disconnected briefly. This is how Log Shipping works, and so long as they send you all the transaction log backups, you don't need to restore the weekly full. Just apply the next log backup in the chain.