Manticore cluster

Hi all,

We have implemented a Manticore cluster with 2 nodes (each node in a separate datacenter). We have noticed a significant increase in time of reindexing large volumes of data.

For example the time to reindex a volume of 200K records (data being inserted in batches of 5K records) has increased for about 5 times with this setup vs standard setup with indexes detached from cluster.

I am aware there should be expected an increase in write time when using a cluster but I would not expect such a significant difference.

It this something normal? Are there any Galera parameters we should consider tunning to optimise speed in this setup with nodes in different datacenters?

Thanks a lot!

you could increase nodes count in the datacenter there you populate your data to make sure the most of quorum nodes share the same network and use different gmcast.segment for nodes in the different data center to minimize the communication. Or you might populate your data first then add that index into cluster to issue SST and transfer data once

Thanks for suggestions, will try them