Explain difference between Kibana "time ellapsed" and query "took"

In Kibana dev-console, I see a big difference between ES query took time and Kibana ellapsed time.

Why ?

Hi @ebuildy - this post is explaining what "took" is measuring: Query timing: 'took' value and what I'm measuring

In short it's how long the internals of Elasticsearch spent time on the request - everything outside of the cluster (JSON serializing, transmitting data, ...) is not included.

Thanks you !

We found latest Kibana (7.9.X) really slower than before, is it normal ?

I'm not aware of anything - this might be worth opening an issue in the Github Kibana repo

I really cant believe JSON serialization takes more time than the query itself .....

We can confirm, Kibana 7.4 is faster than Kibana 7.9.2.

7.4 is using "_msearch"
7.9 is using async search, which is horrible (in our case).

We setup a 7.4 cluster, reindex data from our 7.9 cluster, re-recreate Kibana stuff, then load page. Kibana 7.4 dashboard is loading 10x faster.

Is is possible to disable the new async search?

You should be able to switch back to the old way of data fetching by setting courier:batchSearches in advanced settings to true.

Thanks you, I will.

To illustrate my misunderstand about Kibana timing, here 4 differents timing information:

The real ES query takes 600ms, but I receive the result after 5 secondes.