2025-latest ships with outdated template databse (version 984)

[mcintoshk]

The mcr.microsoft.com/mssql/server:2025-latest image ships system template databases at internal version 984, but SQL Server 2025 expects version 998. On every fresh container start, SQL Server copies the template files and runs 14 upgrade steps (984 → 998) before accepting connections:

  Setup step is copying system data file 'C:\templatedata\master.mdf' to '/var/opt/mssql/data/master.mdf'
  Database 'master' running the upgrade step from version 984 to version 985.
  Database 'master' running the upgrade step from version 985 to version 986.
  ...
  Database 'master' running the upgrade step from version 994 to version 995.
  (and so on through 998)

This adds 30+ seconds to every container startup, which is particularly impactful in CI/CD environments where containers are ephemeral and recreated per job. It is worse on older hardware. GitLab CI's default 30-second service health check timeout is not enough, causing jobs to fail before SQL Server is ready.

Expected behavior:

The template databases in /templatedata/ should already be at latest/expected version.

Environment

• Image: mcr.microsoft.com/mssql/server:2025-latest
• SQL Server version: 2025
• Platform: Linux (Ubuntu 24.04)
• Gitlab: run as a service for tests

Workaround:
Get image, run image (to update it) and then use that updated image.

[CVET-TJGutjahr]

The 2025-latest image in general seems to have a myriad of issues including #948 that also only happens on 2025.

@mcintoshk for your issue, while it shouldn't have to be done, have you considered creating a small docker image based on 2025-latest where you pre-upgrade those databases and commit that image to a registry you or your company owns? At least until Microsoft addresses some of these issues?

Just trying to think of ways to help as I haven't seen much response from Microsoft here... :(

@amitkh-msft Is there someone you can raise these tickets with so that someone is aware or at least have someone respond that they know of these issues? Sorry to tag you but you've been the only person committing to the master branch of this repo in quite some time.

[mcintoshk]

Yeah I had considered it, just was going to see if I would get a response or not before dealing with it on our side. Considering it has been a week and nothing, I will end up doing that this week and see if if I run into any other issues besides #951 / #954.