What is the best practice to change the definition of a datamodel

[marcokrueger]

hi,
I have a heavy datamodel which is used by several dashboards. The acceleration takes more than one hour so when I change the definition of the model, the dashboards are not usable for at least one hour.
What is the best way to manage such a case without any model downtime ?

Like a pre-acceleration of the new model and than switching the two models, the old and the new one?

best regards Marco

[rsennett_splunk · Accepted solution]

I think you're half way there. In order to enact this plan with near-zero downtime, you'd need to build a macro for that datamodel name so that all the dashboards use something like `beefy_DM_all_users` who's value holds the name of your datamodel. That way, if you clone the datamodel (or import it ) and make your change, then accelerate... once it's ready to go, you just change the value of the macro to point to your new datamodel ie, beefyDMallUsers16 and all the dashboards now point to the new one.

[marcokrueger]

Thank you for the input. 🙂 What can I do for Users using pivot and saved dashboards via the webapp so dont use macros?

[rsennett_splunk]

Pivot is going to be a puzzle...
Because it starts with the datamodel and then it's attached to it.
Now you can either enact or instruct people that they must replace the datamodel name (in the search string) with your macro but that'll happen over and over. (dashboard, edit panels, click the pivot icon, edit search string.
(this is just text in the dashboard... so theoretically, you could run a clean up script to overwrite the name of the dashboard with the macro...

but that might be overkill

So I think you might have to consider that any changes would have to happen off hours... and you might just be stuck with the one datamodel being edited and accelerated. 😕