https://rancher.com/ logo
Join Slack
Powered by
Questions around defining Storage Network on Harve...
# harvester
s
Questions around defining Storage Network on Harvester 1.6.0. I set up a new Cluster Network (VLAN) that is separate from my VM network. I bound it to different interfaces. On the switches I assigned a new VLAN and pass all of that information to the Advanced -> Settings -> Storage Network. However, when I enable it, I get an error that "IP allocation failure for Longhorn Pods". The pods get a new Multus interface bound to them, but there is no communication between them over the storage network. I'm guessing that the switch configuration may be incorrect, but I'm at a loss right now. Any thoughts? Please and thank you!
b
Probably worth asking what version of Harvester this is
s
It's up top - 1.6.0
b
From what I see on the diagram there, it's using multus or something to set up a network over the network (like how it does for the management network.
s
Yep - that's it
Correct. I see the Multus interface get added successfully to the pod.
I can shell into the longhorn-instance-manager pods and do 
ip addr show
and see both interfaces.
b
So you have your VLAN you created. That's coming into your host on what interface?
It's tagged traffic and not untagged right?
s
We have been trying both tagged and untagged. Here's what I have in Harvester
default = VM
Bound to different interfaces for each
image.png
Where the vlan in the Storage Network settings uses the storage network and the VLAN set up on the switches
b
So if you have 
esp-storage
it should be tagged vlan for those interfaces.
s
Yes
That's what my original thought was.
I can see both interfaces added to the Longhorn Manager
image.png
b
You should probably also double check that your IP range doesn't have any conflicts with any of the other vlans ranges.
Because I think that's the multus network.
s
They do not overlap.
b
Are you getting an IP for the nodes on that vlan?
s
No - different VLAN for node and VM IPs
b
I think you still need an IP for the node on the VLAN 117 so multus knows how to talk to each other
s
Oh
How do you set that up in Harvester networking CRDs/UI?
b
so like running 
ip -br -c a
should show a VLAN 117 ip from when you log into the actual node.
Do you not have a dhcp server?
(I've never done this before, it's just what I think is missing)
s
We do not for this environment. Is that required?
b
I think so? There might be a way to set it manually.
s
I see the 10.0.17.0/24 IPs getting assigned to the pods
b
Well for sure there is, wicked settings in /etc/
You can probably make it manual if you needed to
s
You are right that I don't have a 117 VLAN IP on my nodes
b
I think the IPs for the pods would, but it needs a way to contact the other nodes, which is why it isn't working.
s
image.png
b
I don't think the flannel stuff will work without a node IP.
Like you see that on the mgmt-br
but there's no bridge or IP for your other interfaces.
s
Ok, this helps. We'll keep looking on our end. I know we can ping the pods from an instance that has a 117 IP so, like you said, maybe the nodes do need an IP assigned there.
Thanks!
Circling back to update on this. Turned out to be that I needed to "kick" the Longhorn pods a bit and get them all to restart before enabling the Storage Network worked. I had no VMs running and no PV/PVCs. I terminated all Longhorn pods and allowed them to restart and that finished the transition to a dedicated Storage Network.
Open in Slack
PreviousNext
Powered by