How to minimize indexing impact on search operations?

Hi all, I'm wondering if anyone has any recommendations on how to minimize the performance impact of indexing on search operations. My use-case is very search-heavy, but indexing is also happening fairly constantly - mutating random documents in the index, not generally inserting new ones. It is not a big deal if modifications take a couple minutes to show up in the index, but it is big deal if search operations run slowly. Ideally we want all search operations to complete in <200ms, but we're seeing some fairly erratic unpredictable performance when there is indexing happening. I'm planning on playing around with the index.refresh_interval setting, to see if it has any impact on query performance (but the docs seem to indicate it mostly impacts indexing performance). I'm also considering throttling the indexing requests from the client side so they come at a more consistent rate and are less bursty...but I really don't know if that would have any tangible impact (it is using the bulk API). We're fairly new to elasticsearch and don't really fully understand all the internals and what all the considerations, any insights are appreciated. We're using AWS's managed elasticsearch (for now), so our ability to tune the server/node configs is very limited...hopefully that's not a problem :slight_smile:

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.