What performance am i looking at per proxmox server, i want to follow your lead in getting something like 1 VM per project going that seems ideal. We will be running maybe at most 6 projects at once but Ill need some headroom. maybe a 32 core threadripper with 256Gb of ram should be plenty?
I cant see another way of working going forward, and I have a crazy idea with 20Gbit/s thunderbolt networking with mac Studios.
Looking at upgrading to a Qnap with thunderbolt ports to get 4 mac studios going for flame.
it seems quiet economical vs 25GbitE anyhow, would be great to hear what you think.
We run our S+W servers in an LXC container and that is much more light weight than a full VM. We run on an old Dell R430. Modern machine are so much faster.
You probably don’t even need as much CPU as you are proposing. Lots of RAM is good though if you are going to use built in ZFS on Proxmox as underlying storage for the LXC/VM though. We also have an offsite instance of Proxmox Backup Server that we backup the VMs to every hour.
i am wondering if docker could be a option vs lxc/proxmox just because I am way more familiar with that, ah yes another rabbit hole to go deep into but I like that.
anyhow so I follow the adsk guide on installing it?
→ setup rocky8 lxd container (do I need 8.7 specifically I guess thats a yes?)
→ Move parts/folders/configs needed from a full autodesk-flame ISO install over so that DKU can be installed?
→ install DKU with novdrivers option
→ install flame as project server
in there setup s+w partition to be on my fast nas, neverminding the speed between the project server and the nas as there is no data transfered through the VM. (the adsk guide is weird on that as it claims I need 50Gbit/s throughout for 8 clients, but it all sounds like they setup the project server as a Nas also?
Docker does not run natively on Proxmox, you would have to run it with a VM and then that just seems kinda redundant and lame. LXC is the way to go or normal VM.
for project server I run with the latest rocky 8.X. dnf -y update starting with a Minimal install.
Just use the install_projectserver script with the nvinst option, it should take care of the rest. No DKU.
I have my Proxmox host on 10gig, but only metadata data passes thru the project server with the proper settings. Shared=True on everything and mounted identically.
The ADSK guides are rather in-efficient and basically as always with Flame, instead of being optimized, just throw more silicon at it.
Also, I put the /opt/Autodesk folder on its own 1TB drive slice. That way if I need to blast the OS, I don’t have to worry about my Flame stuff. That also give plenty of space to run the fucking .clip library backups and all the other file spam that S+W creates.
thats good to know! I tried wirh the installer and was complaining about missing nvidia libraries - ill give it another go without DKU. agaik the guide is claming rhe nonvidiadriver option is for the dku or whatever…
Been running the full-VM just to try out rhe concept abd yea I do want this going forward, thats awesome.
still hardcore in the middle of deciding on hardware.
like it works now, its fine - but my disk based synology is kinda hitting a wall a little bit, even with nvme cache.
Its not that easy scaling up production like this, more people, more flames, more hardware… more ac, more power…
but hey definetely going down this route if I get another minute ill do the container setup again, just for testing running a single vm and its
been fine
