Purpose of adding Openshift intrastructure nodes

Question

I can see often Openshift architectures including "Infrastructure Nodes" in the picture. What is the difference between these nodes and the masters ? I understand that Infrastructure Nodes typically host non-user pods like ha-proxy but do we need it just for that ? Put it simply, can't the ha-proxy stuff be installed just on the master nodes ? Thanks

Kyoko

Answer 1

The purpose of the Masters is to provide the API and Controllers services and, if possible, make them non-schedulable. Those masters will take care of:

• Handling requests from clients (Nodes, Users, Admins and other infrastructure systems deployed on Openshift)
• Run the Scheduler and Replication Controller
• Provide Client tools (oc and oadm)
• etcd Datastore (if not installed separately in different servers)

Then, all your non-user pods like ha-proxy router, as you say, and docker-registry, registry-console, logging-ops, metrics or whatever you consider infrastructure services (gitlab, nexus and so forth) can be deployed on your infra-nodes.

Aurora

Thanks Ruben - could you also elaborate on what is the benefit of segregating the nodes which hosts non-user pods from the non-schedulable master nodes. – Succinctorium

@Succinctorium as you see in my comment, there are multiple services that are critical for the cluster to be able to take care of the applications running. If for some reason an application is overloading the node, the API and Controllers won't be affected. That's also why it is advisable to slit infrastructure from business applications as well. – Aurora

Answer 2

In addition to @ruben-romero 's answer there is also a difference in subscription and licensing cost of openshift infrastructure nodes.

vCPU consumption of infra nodes does not count towards your usual vCPU count which is used to calculate subscription costs.

Southern