visual studio - SQL Server Project Publish to a Docker Hosted Instance - SSDT and DefaultDataPath

Question

I am having a VERY difficult time publishing a pre-existing SQL Server project to a Docker hosted instance of SQL Server.

What I am attempting to do is make a clean pipeline for a Docker hosted instance to use in testing a SQL Server project, which of course starts with doing it first by hand to understand all the steps involved. The SQL Server project itself has been around for many years, and has no problems deploying to SQL Server instances hosted on Windows boxes.

As near as I can tell, the issue comes while SSDT is generating the SQL Server deployment script itself. In a normal deployment to a Windows hosted SQL Server, the generated script starts out with some :setvar commands, including:

:setvar DefaultDataPath "C:Program FilesMicrosoft SQL ServerMSSQL11.MSSQLSERVERMSSQLDATA"
:setvar DefaultLogPath "C:Program FilesMicrosoft SQL ServerMSSQL11.MSSQLSERVERMSSQLDATA"

However, when publishing to a Docker hosted instance of SQL Server, and the same deployment process, the SQL script has:

:setvar DefaultDataPath ""
:setvar DefaultLogPath ""

The 1st thing this deployment does is to alter the database by adding in an additional data file, e.g.:

ALTER DATABASE [$(DatabaseName)]
ADD FILE (NAME = [ARCHIVE_274A259D], FILENAME = N'$(DefaultDataPath)$(DefaultFilePrefix)_ARCHIVE_274A259D.mdf') TO FILEGROUP [ARCHIVE];

The Docker based deployment then craps itself because the file path is (obviously) invalid.

In researching this problem, I've seen MANY solutions which hand-edit the generated deployment SQL script, and manually set the "proper" values for DefaultDataPath and DefaultLogPath ... and even one solution that ran the generated Sql through some sort of post-processor to make that same edit in a programmatic way with string replacement. This does work, but is less than optimal (especially in an automated build/test/deploy pipeline).

I've checked in the Docker instance itself, and its mssql.conf file does have defaults defined:

$ cat /var/opt/mssql/mssql.conf
[sqlagent]
enabled = false

[filelocation]
defaultdatadir = /var/opt/mssql/data/
defaultlogdir = /var/opt/mssql/log/

Can anybody shed light on why these are not being picked up by the SSDT process of generating the deploy script?

I spent a few days trying various workarounds to the problem ...

1. Defined the DATA and LOG directories in the Docker "run" command, but this had no effect on the gnerated Sql deploy script, e.g.: -e 'MSSQL_DATA_DIR=/var/opt/mssql/data/' -e 'MSSQL_LOG_DIR=/var/opt/mssql/log/'
2. Configure the Sql Project with SQLCMD Variables. This method could not override the DefaultDataPath or DefaultLogPath. I could add new Variables, but those would not affect the file path of the ALTER DATABASE command above.
3. Tried a Pre-Deployment script specifically tailored to override the values of DefaultDataPath and DefaultLogPath. While this technically CAN override the default values, the Pre-Deployment script is included in the generated Sql deployment script after the ALTER DATABASE commands to add data files. It would effectively work for the rest of the script, just not the specific portion that was throwing the error on initial deployment of the database.

At this point I feel there is either a Sql Server configuration option that I am simply unaware of, or possibly a flaw in SSDT which is preventing it from gathering the Default Path values from the Docker Sql Server instsance. Any ideas?

question from:https://stackoverflow.com/questions/65848030/sql-server-project-publish-to-a-docker-hosted-instance-ssdt-and-defaultdatapat

Answer 1

Waitting for answers