Hi,
I'm trying to build a simple visualization that lists the top N significant terms in a data table and then include it in a dashboard. I'm doing this on an analyzed text field for a relatively large index that frankly stretches our hardware to the max. The issue is that the aggregation and visualization on the dashboard works perfectly fine for any number of queries and filters that shrinks the dataset significantly. However, the default "start" for the dashboard is the entire large index and sometimes users will actually want to see our dashboard visualizations for the entire index before filtering down.
So the problem is that running significant terms on the entire index takes too long and times out. I can set the timeout rate higher, but frankly I just don't want the dashboard to take forever to load. However, because the significant terms is only one part of the dashboard, having that one visualization time out automatically times out every other visualization on the dashboard even though the other ones run quickly.
I can think of a couple theoretical solutions that I'm not sure exist:
-
Is there a way to make the dashboard visualizations load separately? That is, is there a way to set it such that the other visualizations on the same dashboard still load even if one of the visualizations times out? Or alternatively, is there a way to "switch off" one of the visualizations on a dashboard when the index is too large without filters?
-
Is there an implementation of the sampler aggregation in Kibana? I'm guessing this would significantly reduce the load on the aggregation.
Are there any other potential non-hardware solutions to this problem?