Is that an issue ?

Oh so replication can’t work

got it

But this is not the issue per se

As you've seen the VMs seems to work fine as they don't need the triple replication to work correctly in order to boot. However Longhorn or Harvester seems to put a huge strain on the disk (for trying to achieve replication or whatever), which make the cluster provisioning fail (as it is trying to write a lot at once on the already busy disk). I'm facing the same issue and I'm trying to solve it. Can you try to create a single VM and test the disk performance ?

b.t.w cloud provider should be None, correct ?

I don’t see any issue with disks

that is during provisioning

Can you try

dd if=/dev/zero of=/tmp/test bs=512 count=1000000

on your VM ?

I’m using bridge mode so I guess I can’t SSH directly to the hosts as they don’t have an IP

which could be the issue as well, maybe this is a network issue

With an ubuntu cloud you should inject some credentials with cloud-init, either an ssh key if you can access it from outside or a password if you can only access it via the Harvester console, user is ubuntu

will add password + SSH to cloud init later

I’m suspecting network issue, How can I know this is actually working?

I think this is network issue, can you share how you configured it on your side ?

for example if I’m manually creating vm, what settings should I use, considering I only have a single management network on harvester

Waiting for node to get IP This is probably the issue…

Please see logs from the vm attached, something is not working with network

If you've configured another network than the management one, you should make sure that this network is correctly configured (i.e. with dhcp or static mapping enabled somewhere). Sometimes the ip don't come up on the interface because harvester is using qemu-agent to get the ip and if it's not installed correctly in the VM it can fail. You can check up the dhcp logs to find the ip of your VM and try to ssh into it