High-Availability Query Performance

vkruoso · February 14, 2020, 6:14pm

HA-2 vs HA-1: besides the fact that when a query runs in a HA-2 cluster, all nodes CPU seems to be equally used, is the workload really distributed and all leafs participate in it?

Recently I moved from HA-2 to HA-1 in a cluster for testing, and the performance is much improved with the same table structure and data. Does HA-1 cluster distribute the query load more efficiently?

nikita · February 15, 2020, 12:20am

What version are you on?
Reads should be the same on HA-2. Writes should be a bit slower on HA-2 because we duplicate data

vkruoso · February 16, 2020, 2:28pm

We are running 7.0.9.

I understand. It may have been the fact that the HA-2 cluster were very close to the memory limits all the time and that was affecting performance. The HA-1 cluster have lots of extra memory to work on so that probably helps. Does that make sense?