LINSTOR has issues creating new resource definitions during full clone
operations. Switching to linked clones (full_clone = false) should avoid
this issue as it uses snapshots instead of creating new disk resources.
Also removed replicate parameter as LINSTOR handles replication automatically
through its resource groups.
Storage configuration changes:
- Add full_clone=true for better LINSTOR compatibility
- Add replicate=1 to disk config for k3s servers on LINSTOR
- Configure etcd-witness to use local-lvm storage on thinkpad
- Add etcd_witness_storage_pool variable with local-lvm default
Fixes:
- etcd-witness now uses local storage since thinkpad is LINSTOR controller only
- k3s-server-1 and k3s-server-2 use LINSTOR replicated storage on acemagician/elitedesk
- Explicit replication flag helps LINSTOR create resources correctly
Update target_node values to match actual cluster node names:
- pve1 → acemagician (192.168.100.10)
- pve2 → elitedesk (192.168.100.20)
- pve3 → thinkpad (192.168.100.30)
This fixes the "no such cluster node" error during deployment.
- Change disk slot from number (0) to string ("scsi0")
- Change disk type from "scsi" to "disk"
- Move cores and sockets into cpu {} block
These changes are required for Proxmox provider v3.0.2-rc05.
- Add required 'id' argument to network blocks
- Change iothread from number (1) to boolean (true)
These changes are required for compatibility with provider v3.0.2-rc05.
The v3.0 provider is not yet available as a stable release in the
OpenTofu/Terraform registry. Downgraded to v2.9 which is the latest
stable version.
Also fixed minor yamllint issues in flux.yml:
- Added space after comment marker
- Removed trailing blank line
