-
Bug
-
Resolution: Done
-
Normal
-
None
-
None
-
False
-
-
False
-
?
-
?
-
OSPRH-9864 - Support Distributed ctlplane nodes across low latency L3 interconnected DataCenters
-
?
-
?
-
None
-
-
-
Important
Somewhere in https://github.com/openstack-k8s-operators/ci-framework/blob/00e228e93ad5624da4dc8eea577b33192ba3beaf/roles/libvirt_manager/tasks/generate_networking_data.yml#L156C15-L156C52 the "public" network is implicitly added to all nodes.
When attempting to set up a virtual lab environment where EDPM compute nodes are not connected to the "public" network (L2) but the intent is to allow traffic via routing this is not desired.
Issue:
leaf1s:
network-template:
range:
start: 100
length: 5
networks:
trunk1: {}
internalapi1:
trunk-parent: trunk1
tenant1:
trunk-parent: trunk1
storage1:
trunk-parent: trunk1
leaf2s:
network-template:
range:
start: 100
length: 5
networks:
trunk2: {}
internalapi2:
trunk-parent: trunk2
tenant2:
trunk-parent: trunk2
storage2:
trunk-parent: trunk2
With VMs definition:
leaf1: start: false manage: false amount: 2 disk_file_name: "blank" disksize: 20 memory: 4 cpus: 2 nets: - trunk1 leaf2: start: false manage: false amount: 2 disk_file_name: "blank" disksize: 20 memory: 4 cpus: 2 nets: - trunk2
Results in error:
TASK [networking_mapper : Call the networking mapper networking_definition={{ _cifmw_networking_mapper_definition }}, interfaces_info={{
cifmw_networking_mapper_ifaces_info |
default(omit)
}}, search_domain_base={{
cifmw_networking_mapper_search_domain_base |
default(omit)
}}, interfaces_info_translations={{
cifmw_networking_mapper_interfaces_info_translations |
default(omit)
}}, full_map={{ cifmw_networking_mapper_full_map | default(omit) }}] ***
Tuesday 03 September 2024 12:05:52 +0200 (0:00:03.685) 0:04:10.658 *****
fatal: [hypervisor-1]: FAILED! => changed=false
field: null
invalid_value: null
message: ocpbm contains ranges that collides
parent_name: null
parent_type: null
range_1:
end_ip: 192.168.32.104
network: 192.168.32.0/20
start_ip: 192.168.32.100
range_2:
end_ip: 192.168.32.104
network: 192.168.32.0/20
start_ip: 192.168.32.100
It is attaching both leaf1 and leaf2 VMs to "public" network (`ocpbm`) - and the range with start 100 and length overlaps.
Expected - since leaf1 and leaf2 VMs are not supposed to share any L2 network, this should work. But public network is implicitly added, causing the conflict.
