-
Notifications
You must be signed in to change notification settings - Fork 8.3k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[DISCUSS] Visualization saved object migration plan #29382
Comments
Pinging @elastic/kibana-app |
@stacey-gammon Option 3 sounds reasonable to me too, especially since the code to do that in-memory migration would be reusable when it came time to do the "real" migration. Though I'd be interested in hearing @ppisljar's thoughts on all of this. |
currently, we build expression from the current saved object on-the-fly, everytime you render your visualization.. Pretty much option 3). the plan was to do this conversion with the migration API, everytime you add/create a saved object. the reaons why we want new saved object:
i still think this is not a breaking change ... nothing will break :) except downstream projects, that would be loading saved objects from kibana and expecting them to be of specific format .... I think thats fine ... i mean plugin author anyway will need to change the plugin every minor. |
oh. and no, this has nothing to do with new editor, its purely about visualizations, what they were, what they are right now... before they were.... just some json config pretty much, describing how to render and what to render we can always convert from this migration api could ALWAYS convert from json to expression ... no issues there. |
Breaking a saved object api request would be a breaking change, so it should be avoided in minors. The saved object API is a feature of Kibana. The incremental approach is an interesting one. I assume this transformation is being applied at read time in the UI and subsequently persisted when saved. Effectively that means that the existing saved object format is supported at the API level... to an extent. That’s an interesting gray area I hadn’t considered. Can we prompt the user when they go to save a legacy object? Some sort of “saving this object will automatically ‘upgrade’ it behind the scenes to a new more powerful API format, confirm and remember this choice” modal? |
And also, can someone clarify the details about why we can’t do the entire migration to the new format in 7.0? |
Seems like we could use some documentation regarding what format / interface changes constitute breaking changes. |
If there is a multi-user instance of Kibana, what happens when one of the users chooses "save in the new format" option? What level of privileges does the user need to save as new format? Also, can we introduce the new format alongside the old one in a minor and say the old one is deprecated? This way old content continues to load and new content is saved in the new format. But you can't move objects from the new version of Kibana into the older versions (which I think is fine) |
why we can't do it in 7.0: well, we don't have time (or resources).
well ... we are not breaking API request, api will still look the same regarding multi users: this would be for whole kibana and for all users. We are talking about moving to new saved object format for visualizations, when we introduce that (in some version), that will be global, for everyone |
Closing as confirmed with Jim and Court this is acceptable to do in a minor |
The original plan was for #21645 to be introduced in a minor but that wasn't the intention of the migration API - to allow for big changes to objects to be introduced in a minor.
Let's talk about why that migration is required. In the issue above it's written:
but I think we already have now visualizations working with the pipeline in their old saved object format.
@ppisljar do you mind filling in some details more specifically on that migration plan? I could be wrong but I think it might be mainly for the New Visual Editor and so we don't have to support two ways of doing everything. Is that accurate?
I see some options being:
Cons: maintenance overhead and worse UX with them not being interchangeable.
Cons: syncing two sources of truth.
Maybe 3. is a reasonable option? I don't really know, I've only thought about this for about 30 minutes, so I could be missing a lot of context.
Interested to hear others input.
The text was updated successfully, but these errors were encountered: