"Primary Filegroup is Full" in SQL Server 2008 Standard for no apparent reason

Anton,

As a best practice one should n't create user objects in the primary filegroup. When you have bandwidth, create a new file group and move the user objects and leave the system objects in primary.

The following queries will help you identify the space used in each file and the top tables that have highest number of rows and if there are any heaps. Its a good starting point to investigate this issue.

SELECT  
ds.name as filegroupname
, df.name AS 'FileName' 
, physical_name AS 'PhysicalName'
, size/128 AS 'TotalSizeinMB'
, size/128.0 - CAST(FILEPROPERTY(df.name, 'SpaceUsed') AS int)/128.0 AS 'AvailableSpaceInMB' 
, CAST(FILEPROPERTY(df.name, 'SpaceUsed') AS int)/128.0 AS 'ActualSpaceUsedInMB'
, (CAST(FILEPROPERTY(df.name, 'SpaceUsed') AS int)/128.0)/(size/128)*100. as '%SpaceUsed'
FROM sys.database_files df LEFT OUTER JOIN sys.data_spaces ds  
    ON df.data_space_id = ds.data_space_id;

EXEC xp_fixeddrives
select  t.name as TableName,  
    i.name as IndexName, 
    p.rows as Rows
from sys.filegroups fg (nolock) join sys.database_files df (nolock)
    on fg.data_space_id = df.data_space_id join sys.indexes i (nolock) 
    on df.data_space_id = i.data_space_id join sys.tables t (nolock)
    on i.object_id = t.object_id join sys.partitions p (nolock)
on t.object_id = p.object_id and i.index_id = p.index_id  
where fg.name = 'PRIMARY' and t.type = 'U'  
order by rows desc
select  t.name as TableName,  
    i.name as IndexName, 
    p.rows as Rows
from sys.filegroups fg (nolock) join sys.database_files df (nolock)
    on fg.data_space_id = df.data_space_id join sys.indexes i (nolock) 
    on df.data_space_id = i.data_space_id join sys.tables t (nolock)
    on i.object_id = t.object_id join sys.partitions p (nolock)
on t.object_id = p.object_id and i.index_id = p.index_id  
where fg.name = 'PRIMARY' and t.type = 'U' and i.index_id = 0 
order by rows desc

I found that this happens because: http://support.microsoft.com/kb/913399

SQL Server only releases all the pages that a heap table uses when the following conditions are true: A deletion on this table occurs. A table-level lock is being held. Note A heap table is any table that is not associated with a clustered index.

If pages are not deallocated, other objects in the database cannot reuse the pages.

However, when you enable a row versioning-based isolation level in a SQL Server 2005 database, pages cannot be released even if a table-level lock is being held.

Microsoft's solution: http://support.microsoft.com/kb/913399

To work around this problem, use one of the following methods: Include a TABLOCK hint in the DELETE statement if a row versioning-based isolation level is not enabled. For example, use a statement that is similar to the following:

DELETE FROM TableName WITH (TABLOCK)

Note represents the name of the table. Use the TRUNCATE TABLE statement if you want to delete all the records in the table. For example, use a statement that is similar to the following:

TRUNCATE TABLE TableName

Create a clustered index on a column of the table. For more information about how to create a clustered index on a table, see the "Creating a Clustered Index" topic in SQL

You'll notice at the bottom of the link that it is NOT noted that it applies to SQL Server 2008 but I think it does

In my experience, this message occurs when the primary file (.mdf) has no space to save the metadata of the database. This file include the system tables and they only save their data into it.

Make some space in the file and the commands works again. That's all, Enjoy

our problem was that the hard drive was down to zero space available.