DELETE WHERE NOT EXISTS

I was putting together a procedure earlier for a fairly common ETL need -- remove records from one table that no longer exist in another table.  A simple WHERE NOT EXISTS, right?   Well, I caught myself in a syntax error, so I thought I would post it here for all of you.

In this example, we want to delete from CustomerSales where CustomerID no longer exists in the Customers table:

             DELETE dbo.CustomerSales a

             WHERE NOT EXISTS(

                    SELECT 1 FROM dbo.Customers b

                    WHERE a.CustomerID = b.CustomerID)

                    

Pretty close, but no cigar.  Try that, and you'll receive this error:

                   Msg 102, Level 15, State 1, Line 3
                       Incorrect syntax near 'a'.

This is the correct syntax:

             DELETE a

             FROM dbo.CustomerSales a

             WHERE NOT EXISTS(

                    SELECT * FROM dbo.Customers b

                    WHERE a.CustomerID = b.CustomerID)

Hopefully useful for somebody!

Msg 1422, The mirror server instance is not caught up to the recent changes to database....

I ran into this error a couple weeks ago when I was applying a service pack to a couple of mirrored SQL Servers. I had just patched the Mirror and resumed the mirroring session, but when I tried to perform a manual failover of the Mirror to the Principal, I received this message:

We'll see this if we try to failover the mirroring session when the databases are not in a SYNCHRONIZED state.  Makes sense.. if you take a look at this post, you'll see the step before the manual failover was to resume the Mirror session.  It just wasn't completely synchronized yet.

To confirm how far behind the mirror was, I ran this query on both the Principal and the Mirror.  

     SELECT

         DB_NAME(database_id) [Database],

         mirroring_failover_lsn

     FROM

         sys.database_mirroring

The 'mirroring_failover_lsn' is the log sequence number that the Mirror partners use as the point of reconciliation. I waited a couple minutes and tried again.  The failover completed successfully because the LSN was now the same on both the Principal and Mirror instance.

See this for more detail on sys.database_mirroring and the LSN:

      https://msdn.microsoft.com/en-us/library/ms178655.aspx

And this for instructions on performing the manual failover via tSQL:
     https://msdn.microsoft.com/en-us/library/ms179481.aspx

Add DEFAULT constraint, or add new column with DEFAULT

On many occasions over the years, I've enabled a DATETIME attribute on a table, but allowed SQL to populate it for me, with a GETDATE() DEFAULT constraint.  This can be very helpful for monitoring application data flow, and for troubleshooting or debugging.  To know exactly when a record was written, is a very good thing, I can assure you.

You may already have a date field, but there is no constraint.  Just use this to add the constraint to your table:  

   ALTER TABLE [dbo].[YourTable]

   ADD CONSTRAINT [df_YourTable_DateField]

   DEFAULT (GETDATE()) FOR [DateField];

Or, this one can be used to both add the new field, and enable the constraint:

   ALTER TABLE [dbo].[YourTable]

   ADD InsertTime DATETIME NOT NULL

   CONSTRAINT [df_YourTable_InsertTime] DEFAULT {GETDATE());

* I usually use 'InsertTime' or 'DateCreated' to name attributes like this, because I think it very intuitively suggests what the value is used for. 

Error # was raised, but no message with that error number was found in sys.messages

I received this alert from one of my customer's servers today:

   Msg 18054, Level 16, State 1, Line 1

   SQL Server Alert System: 'Severity 016' occurred on \\SQLSERVER11

   DESCRIPTION: Error 778441, severity 16, state 1 was raised, but no message
   with that error number was found in sys.messages. If error is larger than 50000,
   make sure the user-defined message is added using sp_addmessage.

Odd.  I didn't know who or what is trying to raise an error with a msg_id that doesn't exist yet, so I ran this to search all database objects for any reference to 778441:

     SELECT o.name, o.id, c.text, o.type

     FROM sysobjects o RIGHT JOIN syscomments c 

       ON o.id = c.id

     WHERE c.text LIKE '%778441%'

It returned to me this trigger:

If we look at the trigger definition, you can see where the unknown msg_id's are being called:

   CREATE TRIGGER [dbo].[tblOrderItems_ITrig] ON [dbo].[tblOrderItems] FOR INSERT AS

   /* PREVENT INSERT IF NO MATCHING KEY IN 'tblItems' */

   IF (SELECT COUNT(*) FROM inserted) !=

      (SELECT COUNT(*) FROM tblItems, inserted WHERE (tblItems.ItemID = inserted.ItemID))

       BEGIN

           RAISERROR(778441, 16, 1)

           ROLLBACK TRANSACTION

       END

   /* PREVENT INSERT IF NO MATCHING KEY IN 'tblOrders' */

   IF (SELECT COUNT(*) FROM inserted) !=

    (SELECT COUNT(*) FROM tblOrders, inserted WHERE (tblOrders.OrderID = inserted.OrderID))

       BEGIN

           RAISERROR(778450, 16, 1)

           ROLLBACK TRANSACTION

       END

   

Take a look in sys.messages to confirm that they don't exist:

   

     SELECT * FROM sys.messages WHERE message_id IN (778441,778450)

What is the fix?  I just run this to create two new user defined error messages using the numeric identifiers in the trigger:

     USE master; 

     GO 

     EXEC sp_addmessage

        @msgnum = 778441, @severity = 16, @msgtext = 'There is no matching key in

        tblItems, the insert may not be performed.', @lang = 'us_english', 

        @with_log = 'TRUE';

     EXEC sp_addmessage

        @msgnum = 778450, @severity = 16, @msgtext = 'There is no matching key in

        tblOrders, the insert may not be performed.', @lang = 'us_english', 

        @with_log = 'TRUE';

With these new error messages, the msg_id numbers are now known, and the sev 16 alert goes away.  

See this for more details on sp_addmessage:

    https://msdn.microsoft.com/en-us/library/ms178649.aspx

How to apply a service pack to Mirrored SQL Servers

I patched a set of Mirrored servers last week, and I wanted to post the sequence of events here. Hope to help somebody else who needs to do the same.  To minimize downtime, I used a 'rolling' upgrade.  Very simply, this is the upgrade of the Mirror instance, followed by a manual fail-over, and then the upgrade of the new mirror server, formerly the Principal instance.

Say the servers are SERVERA and SERVERB and the mirrored database is JTSdr.  These are my steps to prepare for, and then to apply SQL Server 2008 sp4:

Prepare:

1.       Script dependencies to file from both instances

a)      Configuration settings

     SELECT * FROM sys.configurations ORDER BY name;

b)      Startup parameters

     SELECT registry_key,value_name,value_data

     FROM sys.dm_server_registry

     WHERE registry_key LIKE '%MSSQLServer\Parameters';

c)       Linked Server definition, if exists

     SELECT a.* FROM sys.servers a LEFT OUTER JOIN sys.linked_logins b

       ON b.server_id = a.server_id LEFT OUTER JOIN sys.server_principals c

         ON c.principal_id = b.local_principal_id

d)      SQL Server Agent Job definition

      SQL Server Agent\Jobs\ right click each job\Script Job as CREATE To File

e)      ALL SQL Server logins, with corresponding database users

           ...back soon on this one

f)       Mirror Session definition 

  SELECT d.name, d.database_id, m.mirroring_role_desc,

  m.mirroring_state_desc, m.mirroring_safety_level_desc,

  m.mirroring_partner_name, m.mirroring_partner_instance,

  m.mirroring_witness_name, m.mirroring_witness_state_desc

  FROM sys.database_mirroring m JOIN sys.databases d

    ON m.database_id = d.database_id

  WHERE mirroring_state_desc IS NOT NULL

Perform: 

1.         Connect to Principal, pause mirroring session:

ALTER DATABASE JTSdr SET PARTNER SUSPEND

2.         Backup JTSdr database on Principal instance, SERVERA

BACKUP DATABASE JTSdr TO DISK = 'E:\backups\JTSdr_20160601.bak'

WITH FORMAT;

3.         Connect to Principal, run CHECKDB on JTSdr, SERVERA

              USE JTSdr;

       DBCC CHECKDB

4.       Apply the patch to the Mirror, SERVERB

5.      Connect to Principal, resume mirroring session

    ALTER DATABASE JTSdr SET PARTNER RESUME

6.       Connect to Principal, perform manual fail-over so Mirror assumes Principal role

    USE master;

    ALTER DATABASE JTSdr SET PARTNER FAILOVER

7.       Connect to Principal, pause mirroring session

                  ALTER DATABASE JTSdr SET PARTNER SUSPEND

8.       Apply patch to the new Mirror, SERVERA

9.       Connect to Principal, resume mirroring session

    ALTER DATABASE JTSdr SET PARTNER RESUME

10.   Analyze logs, validate patch application success

Review SQL Server Error Log, Windows Application Event Log, and a Summary.txt and Detail.txt file at this location:        

      C:\Program Files\Microsoft SQL Server\100\Setup Bootstrap\Log

Assuming success, the Final result in the Summary.txt file be one of these:

    Passed

    Passed but reboot required, see logs for details

Of course, SELECT @@VERSION  will now return this:

Microsoft SQL Server 2008 (SP4) - 10.0.6000.29 (Intel X86)   Sep  2 2014 22:55:13 Copyright (c) 1988-2008 Microsoft Corporation  Standard Edition on Windows NT 5.2 <X86> (Build : ) 

Query to list schedule details for SQL Server Agent Jobs

Kinda big, but this is a helpful query for listing scheduling details for SQL Server Agent jobs.  I'm doing a multi-table join between sysjobs, sysjobschedules and sysschedules, returning JobName, Enabled, Frequency, NextRunDate, RunTime and ScheduleDetails.

Here's a slice from my customer's data:

Let me know what you think.

  -- return agent job schedule details

  USE msdb;

  SELECT

    sj.Name [Job Name],

    CASE sj.enabled WHEN 1 THEN 'Yes' WHEN 0 THEN 'No' END [Enabled],

    CASE syssc.freq_type

      WHEN 1 THEN 'Once'

      WHEN 4 THEN 'Daily'

      WHEN 8 THEN 'Weekly'

      WHEN 16 THEN 'Monthly'

      WHEN 32 THEN 'Monthly relative'

      WHEN 64 THEN 'When SQLServer Agent starts' END [Frequency],

    CASE syssc.active_start_date WHEN 0 THEN NULL

      ELSE SUBSTRING(CONVERT(VARCHAR(15),syssc.active_start_date),1,4) + '/' +

       SUBSTRING(CONVERT(VARCHAR(15),syssc.active_start_date),5,2) + '/' +

       SUBSTRING(CONVERT(VARCHAR(15),syssc.active_start_date),7,2) END [NextRunDate],

    CASE LEN(syssc.active_start_time)

      WHEN 1 THEN CAST('00:00:0' + RIGHT(syssc.active_start_time,2) AS CHAR(8))

      WHEN 2 THEN CAST('00:00:' + RIGHT(syssc.active_start_time,2) AS CHAR(8))

      WHEN 3 THEN CAST('00:0' + LEFT(RIGHT(syssc.active_start_time,3),1) +':' +

 RIGHT(syssc.active_start_time,2) AS CHAR (8))

      WHEN 4 THEN CAST('00:' + LEFT(RIGHT(syssc.active_start_time,4),2) +':' +

 RIGHT(syssc.active_start_time,2) AS CHAR (8))

      WHEN 5 THEN CAST('0' + LEFT(RIGHT(syssc.active_start_time,5),1) +':' +

 LEFT(RIGHT(syssc.active_start_time,4),2) 

         +':' + RIGHT(syssc.active_start_time,2) AS CHAR (8))

      WHEN 6 THEN CAST(LEFT(RIGHT(syssc.active_start_time,6),2) +':' +

 LEFT(RIGHT(syssc.active_start_time,4),2) 

         +':' + RIGHT(syssc.active_start_time,2) AS CHAR (8)) END [RunTime],

    dbo.udf_schedule_description(syssc.freq_type,

         syssc.freq_interval, 

         syssc.freq_subday_type,

         syssc.freq_subday_interval,

         syssc.freq_relative_interval, 

         syssc.freq_recurrence_factor,

         syssc.active_start_date,

         syssc.active_end_date, 

         syssc.active_start_time,

         syssc.active_end_time) [ScheduleDetails]

  FROM

  dbo.sysjobs sj LEFT OUTER JOIN dbo.sysjobschedules sjs

           ON sj.job_id = sjs.job_id INNER JOIN dbo.sysschedules syssc

             ON sjs.schedule_id = syssc.schedule_id LEFT OUTER JOIN (

                 SELECT job_id, MAX(run_duration) [run_duration]

                 FROM dbo.sysjobhistory

                 GROUP BY job_id

                ) q1

                 ON sj.job_id = q1.job_id

  WHERE

       sjs.next_run_time = 0

UNION

  SELECT

     sj.Name [JobName],

     CASE sj.Enabled    WHEN 1 THEN 'Yes' WHEN 0 THEN 'No' END [Enabled],

     CASE syssc.freq_type

         WHEN 1 THEN 'Once'

         WHEN 4 THEN 'Daily'

         WHEN 8 THEN 'Weekly'

         WHEN 16 THEN 'Monthly'

         WHEN 32 THEN 'Monthly relative'

         WHEN 64 THEN 'When SQLServer Agent starts' END [Frequency],

     CASE sjs.next_run_date WHEN 0 THEN NULL

         ELSE SUBSTRING(CONVERT(VARCHAR(15),next_run_date),1,4) + '/' +

          SUBSTRING(CONVERT(VARCHAR(15),next_run_date),5,2) + '/' +

          SUBSTRING(CONVERT(VARCHAR(15),next_run_date),7,2) END [NextRundate],

     CASE LEN(sjs.next_run_time)

         WHEN 1 THEN CAST('00:00:0' + RIGHT(next_run_time,2) AS CHAR(8))

         WHEN 2 THEN CAST('00:00:' + RIGHT(next_run_time,2) AS CHAR(8))

         WHEN 3 THEN CAST('00:0' + LEFT(RIGHT(next_run_time,3),1) +':' +                   

            RIGHT(next_run_time,2) AS CHAR (8))

         WHEN 4 THEN CAST('00:' + LEFT(RIGHT(next_run_time,4),2)  +':' +         

            RIGHT(next_run_time,2) AS CHAR (8))

         WHEN 5 THEN CAST('0' + LEFT(RIGHT(next_run_time,5),1) +':' +

            LEFT(RIGHT(next_run_time,4),2) 

                +':' + RIGHT(next_run_time,2) AS CHAR (8))

         WHEN 6 THEN CAST(LEFT(RIGHT(next_run_time,6),2) +':' +

            LEFT(RIGHT(next_run_time,4),2) 

                +':' + RIGHT(next_run_time,2) AS CHAR (8)) END [RunTime],

         syssc.freq_interval, 

         syssc.freq_subday_type,

         syssc.freq_subday_interval,

         syssc.freq_relative_interval, 

         syssc.freq_recurrence_factor,

         syssc.active_start_date,

         syssc.active_end_date, 

         syssc.active_start_time,

         syssc.active_end_time) [ScheduleDetails]

   FROM

      dbo.sysjobs sj LEFT OUTER JOIN dbo.sysjobschedules sjs

        ON sj.job_id = sjs.job_id INNER JOIN dbo.sysschedules syssc

          ON sjs.schedule_id = syssc.schedule_id LEFT OUTER JOIN (

               SELECT job_id, MAX(run_duration) [run_duration]

               FROM dbo.sysjobhistory

               GROUP BY job_id

               ) q1

              ON sj.job_id = Q1.job_id

   WHERE

      sjs.next_run_time <> 0

   ORDER BY

      RunTime