Running MSSQL service in Bitbucket Pipeline

[Todd Schiller]

The official MSSQL Linux Server Docker image expects at least 2GB of RAM: https://hub.docker.com/r/microsoft/mssql-server-linux/

It appears the Pipelines can only allocate at most 1GB to a service, however.

While we can run tests against MSSQL in Pipelines, tests that load a large amount of data using bulk inserts result in the server terminating the connection to the client.

I suspect the service is killing the connection due to lack of memory. All subsequent connections to the service then fail with a "Login timeout expired" message. I can't verify this since no error information is appearing in the pipeline log for the service

Is there any way to relax the memory ceiling on services? It doesn't look like the memory requirement for MSSQL will be relaxed any time soon: https://github.com/Microsoft/mssql-docker/issues/91

[Attila Szeremi]

I have since figured out that you can give more memory to a service by adding "memory: <amount in MB>"

Example:

definitions:
  services:
    sqlserver:
      image: microsoft-mssql-server
      memory: 2048
      variables:
        ACCEPT_EULA: Y
        SA_PASSWORD: <YourStrong!Passw0rd>

[Orlando Galashan]

I don't found memory in the documentation. But it works

[Orlando Galashan]

A very simple solution: https://github.com/justin2004/mssql_server_tiny

service:
  mssql:
    image:
      name: justin2004/mssql_server_tiny
    variables:
      ACCEPT_EULA: Y
      SA_PASSWORD: changeIt!

[unhammer]

It works! Kind of. I can send queries/statements to it with tsql.

But I'd also like to run a db restore before my integration tests run – is there a way to get files into the docker image used by the service? If I apt install docker io && docker ps -a I get

Cannot connect to the Docker daemon at tcp localhost 2375. Is the docker daemon running?

So how can I make a file available to the service?

Is the build directory automatically available to the service in some path (like with --mount in docker)? Is there a way to expose volumes under services in bitbucket-pipelines yml?

(sorry for the weird punctuation, this forum keeps complaining that I'm posting links without using the link function :<)

[Orlando Galashan]

You can generate your own Docker image based on this image and in these image include the test data your needed.  Or you add a step in the pipeline that populate the Database with data.

[unhammer]

I tried

definitions:
  caches:
    stack: ~/.stack
  services:
    mssql:
      image:
        name: justin2004/mssql_server_tiny
      memory: 2048 # default: 1024
      volumes:
      - /opt/atlassian/pipelines/agent/build:/stuff
      variables:
        ACCEPT_EULA: Y
        SA_PASSWORD: lsakjfd

and in the pipeline step script:

- printf 'USE master\n%s\ngo\n' "RESTORE DATABASE CI FROM DISK='/stuff/bak.bak' WITH MOVE 'CI' TO '/var/opt/mssql/data/CI.mdf', MOVE 'CI_log' TO '/var/opt/mssql/data/CI_log.ldf'" |tsql -oq -U SA -S "$HOSTNAME" -p 1433 -P lsakjfd

but it fails with "Cannot open backup device '/stuff/bak.bak'. Operating system error 2(The system cannot find the file specified.)."

I think this question  is related.

Guess I may have to look into making a private docker image if there's no way to pass files into services :-/

[unhammer]

One workaround is to install mssql manually in the main pipeline. Probably won't work for everyone, as you'll need pipeline-specific install steps, but worked for us at least (on a debian stretch image) with some .deb-fiddling.

Relaxed memory ceiling on containers would make things a lot simpler though :-/

[Attila Szeremi]

Great, so how are we supposed to test things in MSSQL then if the amount of memory available to services is not enough?

[Matthew Hillier]

I've hit the same issue, and it looks like the SQL Server container is only ever granted 1GB memory, no matter the pipeline memory configuration. :(