Infra Scheduler Operator
The purpose of this operator is to segregate OpenShift infrastructure workloads from normal tenant workloads.
The intention is to automate the structures around KCS Infrastructure Nodes in OpenShift 4
Scratch notes
createInfraPool
validateInfraPool
deleteInfraPool
Managed by MachineSets?
-> Yes, createInfraMachineSets with infra label
Pick default AZ MachineSet definitions to copy: which AZs? (default all AZs with replica of one or more, up to three nodes)
How many of each AZ ?
Compute Profile ?
Apply by T&T ?
-> add Taints to MachineSets
Scale up N replicas of N MachineSets in each AZ
remove worker label on each node
validate end state - reconcile on this.
-> No, user pick nodes,
assignNodeLabel infra,
removeNodeLabel worker,
Apply by T&T?
-> setNodeTaint
Managed by defaultnodeselector?
-> setDefaultNodeSelectorOnScheduler
for each infra workload:
-> setWorkloadNodeSelector
Apply by T&T ?
-> setWorkloadToleration
Managed: true/false
Method: taints,defaultnodeselector
MachineSetDefs:
AvailabilityZones: by cloud provider; list
Replicas:
Nodes: list of nodes
(either Nodes or MachineSetDefs to be defined)
Isolate: true/false, determine whether to remove worker label
workloads:
- IngressController
- Monitoring
- Logging
- Registry
- ServiceMesh components
- Operators
also create a controller to modify inflight pod specs of any known infra workloads to apply at this particular point (prevents need for changing by using CRDs for each operator CRD)