Hi, I'm hoping that someone can offer some suggestions for getting an unusual configuration of CloudStack working. We're using CloudStack for managing VM's that are used for training purposes. We've developed code in MOODLE that controls VM states and snapshots and maps VM's to students. When students enter a particular course MOODLE allocates an available VM, restores it to the appropriate snapshot, spins up the VM and then provide the student with an RDP file to access the VM. It also monitors for idle VM's, spins them down and returns them to the "training pool".
It's been working well for years and we haven't updated anything in that time. We recently decided to upgrade all of the systems to supported versions and have run into some trouble with Cloudstack networking. In the original system the networking was configured so that CloudStack was hidden behind cloudbr0, effectively using cloudbr0 as a router as well as the cluster management server, with rules in iptables to forward external RDP connections to CloudStack VM's. Since there is no need for this system to be on an actual cluster, just using cloudbr0 worked well. Also, since this system is portable the NIC is DHCP, meaning that the NIC ip address can't be used for the cluster manager ip address. I'm trying to upgrade to Rocky Linux 8 and CloudStack 4.20 and discovered that cloudbr0 can't be used for the management server. I went into the source code and discovered that there are checks in there to ensure that it's a physical device and that its link state is up. So I tried a few things: - Configured the active NIC to use DHCP and gave it a secondary static ip address that was in the same subnet as cloudbr0, and then used it as the management server ip. That almost worked. CloudStack started up and the CloudStack service VM's came online and were handed out appropriate ip addresses, and I could spin up client VM's. And I could see that the VM vnet devices were bridged to cloudbr0. But none of the VM's had any network connectivity. DHCP didn't work. I tried to set static ip addresses but the links in the VM's didn't appear to come up and I couldn't ping anything (not cloudbr0 or any other VM). I adjusted the ingress and egress rules to allow all traffic. - Bridged the active NIC to cloudbr0 but this appears to have caused it to lose DHCP. - Set up the unused NIC with a static ip address, bridged it to cloudbr0, and specified its ip address for the cluster manager. But because it's not in an up state, CloudStack won't use it. So I'm kind of stuck now and can't think of how to configure the networking. Any suggestions? Thanks, Dave
