Search results

A vnet in sdn is a bridge, configured at datacenter level then deployed locally on each host. That's why it's the same permission name.

maybe you have your proxmox host ip on a specific vlan ? (eth.X) , and vm on a non-vlanware bridge with same vlan ?

@tuxis Are you looking for speakers ? Maybe I could make a talk on my current work on SAN snapshot support. (I have already done a conference in French, slides are almost ready, I just need to translate them to English)

services restart occur on logrotate, but they are no impact on the vms. (the "server" in the log, is the service, not the host) /etc/logrotate.d/pve /var/log/pveproxy/access.log { rotate 7 daily missingok compress delaycompress notifempty...

go to the network zone on the left tree, then add permission on the whole zone or on a specific bridge/vlan

you need to add PVESDNUser role to your bridge

you need HA, if you want to auto restart vms of a dead node to another node. you can install corosync qdevice on your truenas as 3rd node. https://pve.proxmox.com/wiki/Cluster_Manager#_corosync_external_vote_support

no, qemu can't emulate missing cpu flags. (vm will not start if you choose a virtual cpu model newer than your physical cpu model)

you need to enable vlan aware on your vnet (in advanced options)

This is the way.

host ip is in the same subnet and same vlan than the vm ?

please test with pve8 && recent qemu, I remember than some iothreads crash has been fixed since. (and pve7 is EOL anyway)

until you have infiniband switches && application using infiniband, no, you don't need it ^_^

mdev devices are different than pci functions, so you can't see them with lspci, but you can see them with "mdevctl list" you simply need to pass the mdev (nvidia-xxx) in the pci passthrough gui

nf_conntrack_allow_invalid:1 in host.fw

ah, in english, great. I'll try to come this year :)

you can't use physical output. vgpu is mostly use for gpu compute (ai, machine learning,...). or you need a fast network display protocol like for cloud gaming.

the problem with "upgrade" is that it's only upgrade installed packages. if an update of a proxmox package have a depend on a new lib not yet installed, it could break it. (This is different than debian, where the package list is fixed over the whole cycle. Proxmox can add new features and...

memory is not reserved at vm start (until you define static memory hugage in vm conf directly). so it can be dynamically allocated to a different vm. then, if a vm is reserving a memory page, it's reserved. Note that windows is allocating all memory pages with 0 at at boot. (so it's...

Well, I don't known how far are theses sites, but you can't have too much latency (3~5ms) for 1 cluster. (and keep quorum (a majority of nodes up) if you have if 1 site down). The vcenter approach is pdm. (separate clusters by site), and should be the design for your setup normally.