Recent content by syfy323

No issues, just a restart of the replication service. What I do not understand is that the status query works using shell and failes when run by pveproxy (when the webif needs the data). This will most likely be the same issue.

Anyone?

Hi! I am managing a cluster running PVE v6, which currently includes two servers. My third node has been replaced but I am unable to join it. When I access hv102 using the GUI on hv101, I get "Connection error 596: Broken pipe". If I try to join hv103, I get this: root@hv103 ~ # pvecm add...

https://bugzilla.proxmox.com/show_bug.cgi?id=2569

Indeed, this needs to be automated :) In the past I had to just rename the RBDs to the vm scheme and they appeared as unused. Thanks for clarification!

Hi! Is it safe to generate /etc/pve/qemu-server/XXX.conf files on my own? I have to migrate an OpenNebula-Cluster with more than 400 VMs, host by host. It would be much simpler if I can create the VM.conf on my own and place it on the host (mac address, vcpu config, etc.). The storage is Ceph...

Ok, I understand. AFAIK Proxmox only supports this. If your NFS server is based on Linux, you can hardlink the file to a new export instead and mount that.

You just need to mount the NFS share using Proxmox. The only restriction is, that the directory is "flat", in terms of all ISOs need to be directly in this mount and not located in subdirectories.

I don't think there will be a stable solution anytime soon. Kernel NFS has a lot of issues and deadlocks, at least in VM context. The reason for faster NFS might be "async" mode which is very unsafe, just like the "writeback unsafe" option in proxmox. If oyu don't mind the risks you can also...

The kernel nfs-client and server has a lot of flaws. Thats also the reason, large setups use nfs-ganesha instead of the kernel server.

Mos Most if not all CPUs work with virtualization. I'm running CPUs <3GHz in production, so yes, it works. It depends on how much you want to overprovision your CPUs and how your load is distributed (idle VMs vs. high-load VMs.).

High frequency is always good but if you are running an application that is incapable of using multiple cores, a high frequency is the only way to get performance. Software that utilizes more cores is much more efficient, much more than high frequency.

https://packages.debian.org/buster/net-tools

Für HA-Zwecke und Live-Migration ist das sowieso zwingend notwendig. Das sollte man immer im Hinterkopf behalten, falls man das mal irgendwann verwenden möchte.

Makes sense. I know the script on github but wanted to make sure, I don't run mitigations on the KVM/Qemu layer that neither the host, nor the VM needs. Thanks for clarification!