...
Monitoring cluster description
The monitoring cluster has a virtual IP (VIP) that Each monitoring node runs HAproxy and Keepalived with a virtual IP to create a highly available load balancer. HAproxy listens on port 36639/tcp to balance across all Emmett instances, which listen on port 36638/tcp, and KairosDB instances. The virtual IP will be active on one node at a time only. If a node fails, Keepalived will switch to another node.
To access Watchtower (Emmett), the Abiquo components ( API and RS) Remote Services will use the VIP. All HAproxy instances balance across all Emmett and KairosDB instancesvirtual IP. Emmett also uses VIP to access KairosDB, so it will always reach an HAproxy that will be able to connect to a KairosDB backendthe virtual IP to access KairosDB.
Each monitoring node runs all the Watchtower components:
abiquo-emmett
abiquo-delorean
kairosdb
cassandra service
Cassandra is configured as 1 cluster with 3 monitoring nodes, with replication factor set to 3, so all nodes have the same data.
The Watchtower MySQL database is on the Datanode server, together with the other Abiquo databases.
All monitoring nodes can be in the same physical datacenter, so the data replication is faster. Each node will hold the same data, so if one of the nodes fails, there will be an exact copy on the other ones.
With the Watchtower MySQL database on the central database server, it is easier to configure the Emmett and Delorean database server because it will point to the central Datanode.