We need to understand what is the requirement for 'Cassandra/ElasticSearch clusters' where we have DMA’s in 3 different locations.
Is having all the nodes in one location (main site) fine, or we need to have a node per location?
What is the best option here ?
In our compute requirements we do not recommend latency higher than 50 ms:
Reading up on Cassandra documentation they only talk about using the same rack <1ms
General System/OS checks for DSE/Cassandra clusters :: DataStax Enterprise Capacity Planning Guide
In there you can also find some calls you can do to verify if the slow network is causing issues
I didn't check for Elastic but i would assume it is very similar to avoid high latency between the nodes since this would delay any read/write actions significantly
so therefor i would say try to keep all nodes together, in case DataMiner can't push fast enough due to the latency, this will likely result in a SLDataGateway leak where in turn can easily identify with a logcollector package + SLDataGateway dump which parameters are not able to cope
Likely you will also be able to set an alert on a network metric to identify when the network can't cope.
The latency from the main site to the other site is 119 ms.
Within the main site itself the latency is <1ms.