Our Azure Managed Grafana instance was updated to 9.1.8.1, this broke our dashboard because how the Azure Monitor datasource references resources changed fundamentally, and the old way was deprecated (and apparently they just decided to break it based on my experience). While this is annoying that it was a change with major breaking changes to components that just about everyone using AMG is probably using (i.e. the Azure Monitor queries), and is going to require a major refactor of my dashboards.
I get that the point of a managed service is that it's kept relatively up to date without my intervention, and I appreciate that, but to have an update break everything I've built without notice makes it hard for me to continue to trust the managed service. I also recognize that this is a young product offering, and I appreciate that it's iterating quickly to bring in new features, but this is a GA product now, so being suddenly surprised that my dashboards are broken and need a major refactor when I come back from the weekend was a nasty surprise. There has to be some combination of better testing to identify breaking changes between versions, advance warning of breaking changes, explicit opt-in to versions with breaking changes, etc.
The bright spot in this is that the new version adds support for the Resource Graph query type for template variables... but it has a major bug where the UI picker for the subscription doesn't work, this bug is fixed in 9.2, which has been released.
I guess my questions here are two-fold...
1) Are there plans to allow users to pin a version of grafana in the case of breaking updates, similar to how AKS clusters can specify the version that they're running. Or are there plans in place to help customers anticipate and prepare for breaking changes in the future?
2) What is the timeline for 9.2 to be rolled out to managed instances? I feel like the killer feature I got out of this whole headache is mostly negated by a bug that makes it incredibly difficult to use.