For vBMC only there is no constraint on the VM needed configuration. But there is some specific needs for a VM to be able to PXE-boot by mean of HPCM. This part is detailed here.
The VM to create for the HPCM software should have:
- at least 2 network interfaces
- one for the public network (the house network),
- one for the HPC admin (aka "head" network)
- at least 100 GiB of disk per slot (if you want to play with two slots set a virtual disk of at least 250 GiB)
- 16 GiB or RAM seems OK, I use 32 GiB but have more than 16 GiB unused so far (all monitoring feature are setup, but no SU_Leaders configured)
- 8 cores seems OK, more if possible to speedup compression operation (image creation for example)
- a virtual CD/DVD Drive to boot on the ISO image during the HPCM installation process
- UEFI Boot is mandatory (Proxmox will require the creation of a tiny virtual disk to store the EFI parameters)
Here below the VM configuration I used for HPCM VM
Installing HPCM using the provided image with the embedded OS (here RockyLinux 8.10) is straight formward, as described in the HPCM documentation. The VM environment does not bring any issue relative to that task.
The problems arise with the compute node setup for PXE-boot ability:
Create a new VM using Proxmox, select an hypervisor (here proxmox-2, eventually assign it to an existing Resource Pool):
Next we do not need any OS (HPCM will provide the OS/image by mean of PXE boot)
Then select the BIOS OVMF to get UEFI boot:
Also add an EFI Storage and uncheck the Pre-Enrolle Keys box, check the Qemu Agent box as the default images HPCM provides already provide a qemu-guest-agent, which will let you transparently interact from Proxmox with the running operating system (including cleanly shutting down the OS from Proxmox):
You don't need any disk (unless you want to setup a SU_Leader, though disks can be added later once the VM has been created), thus let's delete the preconfigured disk for now and click "next":
Choose as much CPU socket/core as you want (depending on what workload you want to have running on you virtual HPC cluster):
Choose as much RAM as you want (and have) still depending on the workload you expect to deploy in your virtual HPC cluster
Last set the first Network card you want, the other NIC(s) will be added later.
Obvious note: that the "head" Virtual Network should be connected to this VM (through this NIC or another we will add later) for HPCM to provide PXE boot support.
Complete the VM creation process.
Last tune this hardware set as follows:
- a Virtio RNG (Random Number Generator) is mandatory for PXE Boot to work
- a serial port can be added and be used as the console output of the system (optional) see Console on Serial
- one or more additional network interface(s). For example, a High Speed Network emulation
- the CD/DVD drive can be removed it will not be used for a compute node bootstrapped by HPCM
| Prev | top | Next |
|---|