Cannot specify different interfaces for different nodes to run 'tcpcloud' driver
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
os-faults |
Won't Fix
|
High
|
Anton Studenov |
Bug Description
In tcpcloud baremetal environments can be used hybrid deploy (control plane is deployed on VMs, computes and some other services can be deployed on baremetal nodes).
In this case, some nodes will have interfaces eth0/1 , some nodes ens2/3 , some nodes ens3/4 , some - p4p1/p4p2, some - enp8s0f0/enp8s0f1 and more interfaces.
Another issue is: on the same interface can be several different IP addresses. For MK24 it is normal to have a "static" address and a VIP address on the same interface.
os_faults should use a different approach to get the IP addresses of the nodes instead of using IPs for interfaces.
For example, solution can be like this:
$ salt '*' pillar.get _param:
, but the pillar name can be different from model to model used in salt, so it should be also specified for os_faults.
Changed in os-faults: | |
assignee: | nobody → Anton Studenov (astudenov) |
Changed in os-faults: | |
status: | New → Confirmed |
Changed in os-faults: | |
status: | Confirmed → In Progress |
importance: | Undecided → High |
MK24 is not relevant any more. Moving the bug into Won't Fix