Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

How to Partition a Table by Month ("Both" YEAR & MONTH) and create monthly partitions automatically?

I'm trying to Partition a Table by both Year and Month. The Column through which I'll partition is a datetime type column with an ISO Format ('20150110', 20150202', etc).

For example, I have sales data for 2010, 2011, 2012. I'd Like the data to be partitioned by year and each year be partitioned by month as well. (2010/01, 2010/02, ... 2010/12, 2011/01, ... 2015/01...)

E.X:

Sales2010Jan, Sales2010Feb, Sales2011Jan, Sales2011Feb, Sales2012Dec, etc.

My Question is: is it even possible? If it is, how an I automate the process using SSIS?

like image 903
Amr Tharwat Avatar asked Aug 10 '15 15:08

Amr Tharwat


People also ask

Can you partition by date in SQL?

You can partition on DateCreated column after you make it a clustered index.

How do I partition a table in SQL?

To create a partitioned table, you follow these steps: Create file groups that hold the partitions of the table. Create a partition function that maps the rows of the table into partitions based on the values of a specified column. Create a partition scheme that maps the partition table to the new filegroups.


1 Answers

SSIS is an ETL (extract, transform, load). This is not what you want to do. You just need to create DDL statements dynamically .

I work with quarter below but it works as well with 1, 2 or X months if you want.

If you want to partition the table, you first need to create the file, filegroups and partionned table and set the partitionning manually

Creation of N+1 partitions for 2015 Q1 (before, Q1 and after Q2) on a table with an int identity PK and a datetime2 partitioned column. Update it to add months, make it monthly or whatever you need...

  • First create N files groups:

    Alter Database [Test] Add Filegroup [Part_Before2015]
    Go
    Alter Database Test Add Filegroup [Part_201501]
    Go
    Alter Database Test Add Filegroup [Part_201504]
    Go
    
  • Add a file for each filegroup:

    Alter Database [Test] Add FILE ( NAME = N'Part_Before2015', FILENAME = N'...\Part_Before2015.ndf' , SIZE = 5120KB , FILEGROWTH = 1024KB ) TO Filegroup [Part_Before2015]
    Alter Database [Test] Add FILE ( NAME = N'Part_201501', FILENAME = N'...\Part_201501.ndf' , SIZE = 5120KB , FILEGROWTH = 1024KB ) TO Filegroup [Part_201501]
    Alter Database [Test] Add FILE ( NAME = N'Part_201504', FILENAME = N'...\Part_201504.ndf' , SIZE = 5120KB , FILEGROWTH = 1024KB ) TO Filegroup [Part_201504]
    
  • Create a partition function on a datetime2 type (or date or even datetime):

    Create Partition Function RangePartFunction (datetime2)
    as Range Right For Values ('20150101', '20150401') 
    
  • Create a partition scheme using the partition function on each filegroup (N+1):

    Create Partition Scheme RangePartScheme as Partition RangePartFunction
    To ([Part_Before2015], [Part_201501], [Part_201504])
    
  • Create the partitioned table on its partition scheme:

    Create TABLE [PartitionTable] (id int identity(0, 1) not null, date datetime2 not null, text char(8000))
    On RangePartScheme (date) ;
    
  • Add a Clustered index on the partitioned column and partition scheme:

    Create Clustered Index IDX_Part On dbo.PartitionTable(date) 
        On RangePartScheme (date);
    
  • Add a PK to the id column:

    Alter Table dbo.PartitionTable Add COntraint PK_Part Primary Key Nonclustered(id, date);
    

Build the query used to add extra file groups after the right boundary and split the last partition

  • Review partition scheme extension and partition function split
  • Review DMV used
  • Review all of this and how to use it to create dynamic SQL

    Declare @currentDate datetime2
    Declare @endDate datetime2 = '20160701' -- new end date
    Declare @dateAdd int = 3 -- Add 3 month = 1 Quarter
    
    -- Get Current boundaries 
    Select @currentDate = DATEADD(MONTH, @dateAdd,Cast(MAX(value) as datetime2)) From sys.partition_range_values as r
        Inner Join sys.partition_functions as f on r.function_id = f.function_id
    Where f.name = 'RangePartFunction'
    
    -- Get all quarters between max and end date
    ; with d(id, date, name) as (
        Select 0, @currentDate, Convert(char(6), @currentDate, 112)
        Union All
        Select id+1, DATEADD(MONTH, @dateAdd, date), Convert(char(6), DATEADD(MONTH, @dateAdd, date), 112)
        From d Where d.date <= @endDate
    )
    Select * From (
        Select id = id*10, query = 'If Not Exists(Select 1 From sys.filegroups Where name = ''Part_'+name+''')
            Begin 
                Print ''Create Filegroup [Part_'+name+']''
                Alter Database [Test] Add Filegroup [Part_'+name+']
            End
            GO'
        From d
        Union All
        Select id*10+1, 'If Not Exists(Select 1 From sys.sysfiles Where name = ''Part_'+name+''')
            Begin 
                Print ''Create File [Part_'+name+'.ndf]''
                Alter Database [Test] Add FILE ( NAME = N''Part_'+name+''', FILENAME = N''C:\DB\MSSQL11.MSSQLSERVER\MSSQL\DATA\Part_'+name+'.ndf'' , SIZE = 5120KB , FILEGROWTH = 1024KB ) TO Filegroup [Part_'+name+']
            End
            GO'
        From d
        Union All
        Select id*10+2, 'Print ''Add Range [Part_'+name+']''
            Alter Partition Scheme RangePartScheme Next Used [Part_'+name+']
            Go'
        From d
        Union All
        Select id*10+3, 'Print ''Split Function ['+Convert(char(8), date, 112)+']''
            Alter Partition Function RangePartFunction() Split Range ('''+Convert(char(8), date, 112)+''');
            Go'
        From d
    ) as q order by id
    

the output of this query is a list of SQL queries that must be run in order.

Execute the dynamic SQL

  • It can be executed manually (copy and past in SSMS)
  • It can be executed in a while loop or with a cursor which will executed each row of the ouput table one by one (use sp_executesql)

Automation

  • Create a SQL Server Job which excute SQL queries: run the query used to create the dynamic SQL, save its output to a table variable and then execute each statement with a loop/cursor

If you want to run it monthly and make sure the next 12 months are always created, use this Set @endDate = DATEADD(MONTH, 12, getdate())

Finally

  • It will output 4*N rows for the N missing quarters between the last boundary of the function and @endDate:

    • Create Filegroup
    • Create File on Filegroup
    • Extend the range of the partition Scheme
    • Split the range of the partition function
  • You can run it row by row with a cursor or a while loop or you can just copy and paste it in SMSS.

  • It can be automated with a job as well ie. @endDate = DATEADD(MONTH, 3, getdate() will create the next 3 months
  • Change @dateAdd to 1 if you want monthly partitions
  • Add you own columns or checks

Link

Create job = https://www.mssqltips.com/sqlservertip/3052/simple-way-to-create-a-sql-server-job-using-tsql/

sp_executesql = https://technet.microsoft.com/en-us/library/ms188001%28v=sql.110%29.aspx

While loop = https://dba.stackexchange.com/questions/57933/can-exec-work-with-while-loop-of-cursor

like image 157
Julien Vavasseur Avatar answered Sep 21 '22 16:09

Julien Vavasseur