While looking into building an SLO dashboard for the logging cluster I wasn't able to find bucketed latency metrics for elasticsearch.

Looking a bit further it seems the elasticsearch exporter doesn't provide this by itself, so an alternate approach would likely be needed. One option could be to enable terse query logging for every query and parse that into metrics with mtail, but I'm not sure off hand what if any side effects that might cause. Other options may be available as well, will need to do some research and testing.