I have an instance of proxmox with a VM for Truenas (12gb allocated - non-ballooning) and a VM (fedora 16gb allocated - non-ballooning) for containers with PCI pass through of my GPU. This leaves 4gb for proxmox.
I think I may have some memory pressure here and am wondering if I should add a 16gb kit i already own to the system that is not matched to a currently installed 32GB kit. I have seen some instability that I am trying to troubleshoot as well where VMs seem to shut down on their own daily.
The system currently has 32GB of ram at 3200 but i also have a 16gb 3000 kit laying around already, Should, I try to use both kits at lower speeds? Is it worth it to manually tune the kits or just run at jedec speeds?
• ASUS TUF B450M-PLUS GAMING AM4 (iNIC Realtek RTL8111H)
• AMD Ryzen 7 2nd Gen - RYZEN 7 2700 (Zen+) 8-Core 65W
• Crucial 32GB Ballistix DDR4 3200 MHz (2 x 16GB)
• Gigabyte GeForce RTX 2060 6GB - GVN2060D66GD - passed to container VM.
• HBA - IBM ServeRAID N2215 IT mode (equivalent to LSI 9302-8i) SAS3008 SAS-3 chipset PCIe 3.0x8 lanes (in physical PCIe 2.0x8).
• 4x Seagate 12TB IronWolf SATA NAS HDD CMR RAIDZ1 connected to HBA – Main Pool - passed via IDs to truenas VM (I don’t have a way to do hardware pass through of this and the GPU with the motherboard unfortunately).
• Sabrent 512MB m.2 boot disk with VMs etc. on ZFS pool.
• Patriot 1TB P220 SATA SSD (backup target for boot proxmox and VMs which is replicated to truenas)
I would try to track down the instability first. Then I would try to somehow quantify the memory pressure issue. It’s hard to say sometimes what is is the bottleneck.
You could always just throw it in and see what happens, but adding more variables when you already have uncertainties might just make it more confusing.
For instability, I always like to check thermals make sure the memory sticks aren’t overheating
Thanks, that does sound like good advice. I tried to setup temperature monitoring and found that the VFIO.conf file was all messed up and was passing a few random items that should not have been passed to VFIO so i cleaned that up. I will monitor it for a bit and do more trouble shooting if the issue continues.
I have reserved 8GB for proxmox and that seemssss… to be helping. But so far proxmox has not impressed with its reliability.
BTW - the VFIO correction seemed to help but it wasn’t enough. I found some out of memory errors, especially after a crash when trying to reboot the VM.
If anyone stumbles upon this in the future, the Proxmox instance has been stable for a few weeks now.
For context, my setup uses snapshot-based backups using the Proxmox backup server installed directly on Proxmox and there are a couple VMs that only run on demand. My understanding of the snapshot-based backup process first boots the VM (if not already booted) and then generates a snapshot. After which the VM reverts to its prior state (e.g., shutdown). The most consistent crashes happened around the time of backup. However, other crashes occurred as well with them being more frequent when the system was up for longer.
It appears that the issue was memory pressure and the solution was to reserve 8GB of RAM for Proxmox. This appears to be stable now. Long story short minimum RAM for proxmox is probably 8GB.