Search results

I know that difference but in backup process incremental snapshot using zfs send / zfs receive time is also the same. Maybe PBS/PBS Client dev should use other magic spells for backuping LXC?

Well - doing incremental backup I see that comparing backups VM(1TB) and LXC(800GB) speed there is 10x performance gap (or even more! - doing backup for LXC is significantly slower even if it's smaller) I did some test zfs send/zfs receive to remote host not differs so much. Local storage...

[ 10.984549] VFIO - User Level meta-driver version: 0.3 [ 11.469416] systemd-journald[873]: Received client request to flush runtime journal. [ 12.249335] power_meter ACPI000D:00: Found ACPI power meter. [ 12.269933] power_meter ACPI000D:00: Ignoring unsafe software power cap! [...

Sorry. Dictionary in my Phone made bad things sometimes. I mean PLP technology - capacitor on SSD Drive. That allows to scream by ssd "write done" before it is done because it save. Check pvestat command on both drives.

Most important if You use ZFS: buy SSD with PLOP (performance 2-2.5x !!!)

https://bugzilla.proxmox.com/show_bug.cgi?id=3148 for --with-local-disks

This issue is only valid when there is a 2x Mellanox ConnectX-3 cards. I changed the cards to ConnectX-4 - all works like a charm.

I tried to juggle cards with different PCI-E - it didn't help. When I turn on second network device (no matter whitch) it gives me an error: genirq: Flags mismatch irq 16. 00000000 (vfio-intx(0000:88:00.0)) vs. 00000000 (vfio-intx(0000:05:00.0)) It looks like some race condition bug when I...

When I am trying to start VM with passtrough two network devices there are errors in DMESG: root@rtx-proxmox:~# dmesg | grep 'Flags mismatch' [ 99.938242] genirq: Flags mismatch irq 16. 00000080 (vfio-intx(0000:81:00.0)) vs. 00000000 (vfio-intx(0000:08:00.0)) [ 302.258175] genirq: Flags...

The other thing I tried is add another virtual machine and passthrough second device at the same time. It doesn't worked. Do this matter if the devices got the same IRQ's: 88:00.0 Ethernet controller: Mellanox Technologies MT27500 Family [ConnectX-3] Physical Slot: 4 IOMMU group: 100...

No - I will try. No. All of the available updates are applied.

Yes - I checked it twice. Name of module is like blacklist module inside conf file. I attached my lspci for more info. Most important thing: 1. Kernel driver in use: vfio-pci Kernel modules: ixgbe 2. Kernel driver in use: vfio-pci Kernel modules: mlx4_core So that means that is...

@oguz did You read one mail infromation from my post about that VM is starting when only one device (no matter whitch) is set for passthrough? In my /etc/modprobe.d/ I've got: -rw-r--r-- 1 root root 16 Jun 1 11:36 ixgbe.conf -rw-r--r-- 1 root root 20 Jun 1 11:36 mlx4_core.conf -rw-r--r--...

I tried to passthrough PCI device alone: I tried to passthrough another PCI device also alone: Then I tried to Passthrough together two devices: And I've got: kvm: -device vfio-pci,host=0000:81:00.0,id=hostpci1.0,bus=pci.0,addr=0x11.0,multifunction=on: vfio 0000:81:00.0: Failed to set...

Networking stack is complex - adding additional layer for this would block the oportuninty of tuning in many ways. So I think I probbably use IOMMU instead of SRV-IO and not to use virtio as a bridge to network nic. OS: Proxmox 7.1 --> Debian 11

I will do tests between LXC and KVM with passthrough. I could do also (virtio vs passthrough) but You show me the way to simply simulate 100.000 connections with different src/dst ips....

@tom I plan not to use virtio nics because of performance (I need multiqueue and offload from my hw nic). So there is not a good idea to do that - i will use VirtIO nic for management nics.

I would like to make a router for a 10G+ bandwidth (BGP+QoS+Firewall). Of course I will passthrough the network cards inside to the VM/Container. What is better for that purpose: Container or Virtual machine?

I mean: Why the Nesting feature is not enabled for newly created containers? (and not why the "unprivileged container" option is set by default) :)

Yes that helps.