https://rancher.com/ logo
Join Slack
Powered by
This message was deleted.
# harvester
a
This message was deleted.
b
vm networks
are used by VM network interfaces - vm networks are in turn attached to a 
cluster network
(in practice a vlan-aware linux bridge) which is connected on each node to a 
network config
(in practice a linux bond)
l
Thanks! hmm, Say I need to add a new vm vlan 100 for new vms in this tagged vlan. I first created a new cluster network, naming it as 
cvlan100
, then when I tried to create a new vm vlan 
vmvlan100
, and trying to select 
cvlan100
in the dropdown of 
Cluster Network
, it only showing 
mgmt
as a valid option, and cvlan403 is gray out, with 
(Not Ready)
text near it. Is this expected?
just noticed there’s this ‘Create Network Config’ button in UI. But when I go to 
Uplink
, it won’t show up the 2 NIC ports already being used for 
mgmt
network (during initial installation), these 2 NICs were selected during initial installation. Does this mean it requiring lots of NIC ports, and each nic port could only be used for one vlan?
I have 3 nic ports for the physical hosts. 2 10g NIC ports, and one 1g NIC port. I used 2 10g NIC ports for initial installation, as bonding. Now it only showing this 1g NIC port as a possible selection when I try to create new vm vlan.
Does this mean I will have to keep using the default 
mgmt
cluster network for all things later?
b
the cluster network needs a network config, either targeting all nodes (if interfaces are identically named) or one for each node - this consumes the selected physical interfaces, just like the management bond consumed the two interfaces chosen for it
I’m not too familiar with all the different bonding modes - there might be one that lets you select just the single 1Gbit interface for a “bond”
l
I don’t want to use the 1g port. Is it ok to share all things with default 
mgmt
cluster network? any concerns in doing that other than it would get more traffic later? I assume 
mgmt
should be able to support vlan tagging?
I think the main reason for cluster network use case if to get different physical network ports, and configure those to another vlan just for longhorn storage replication later.
b
You can run everything over the mgmt bond “just fine”, but I think you’ll have issues with at least live migration (hopefully someone will correct me if I’m wrong)
l
yeah makes sense. in VMware, we could have dedicated vmotion ip/vlan for live migration traffic separation, and decicated vsan ip/vlan for storage replication traffic separation .. I am not sure if possible to separate vmotion traffic in harvester case.
12 Views
Open in Slack
PreviousNext
Powered by