I am doing the first installation of the Cloudstack HCI device. I have 4 nodes, all connected to the same VLAN. I performed the installation of the OS and all 4 nodes received IP from DHCP and all respond on the network. 3 nodes are connected to a switch and the fourth node is connected to a second switch. The 3 nodes that are connected to a switch appear in the initial scan. The fourth node that is connected to a second switch does not appear in the scan to compose the cluster, even though the switches are interconnected and all nodes “see” each other in the same VLAN/Subnet.
I would love to do a POC considering LINBIT as an alternative to vSphere, but I started off on the wrong foot!
Any ideas on how I can start debugging the problem?
Thanks for your help. What a shame to hear this. An optimized OpenStack deployment option with support for large environments would definitely be a great option for our company. We have a 50 host cluster with vSphere and LINBIT would be an excellent option if it could scale. It’s really a shame but even the POC ends up being unnecessary.
The Linstor and DRBD can scale to many nodes but the Cloudstack HCI appliance is being developed too.
We are developing this feature where we can add more hosts additionallyu and we should be able to add more hosts in the releases sometime soon.
Its my suggestion that you install separate Cloudstack and use Linstor as the backing storage. You could contact our Architects and they will help you design this.
We’d be happy to walk you through a more traditional CloudStack approach essentially consisting of:
CloudStack nodes
LINSTOR
LINSTOR’s CloudStack plugin
The HCI appliance is a neat idea, but LINSTOR + CloudStack has been a great combination long before the HCI appliance ever existed. We also have some Ansible Playbooks laying around that can spin up a CloudStack and LINSTOR environment pretty quickly.