Search results

rbd image 'vm-104-disk-1': size 60 GiB in 15360 objects order 22 (4 MiB objects) snapshot_count: 0 id: 433efbc6fb9b03 block_name_prefix: rbd_data.433efbc6fb9b03 format: 2 features: layering op_features: flags...

root@gpu01:~# ceph versions { "mon": { "ceph version 14.2.2 (a887fe9a5d3d97fe349065d3c1c9dbd7b8870855) nautilus (stable)": 3 }, "mgr": { "ceph version 14.2.2 (a887fe9a5d3d97fe349065d3c1c9dbd7b8870855) nautilus (stable)": 1 }, "osd": { "ceph version...

please try pc-q35-3.1 instead of q35 or i440fx if you are using PVE 6.0

I am running PVE 6.0 and on this pve node, I have already started 2 vm with rbd disks to my ceph storage, However, sometime if I need to start third VM using rbd disk, PVE will error with following msg. It only can be solve if I reboot the pve node and then I can start 3 VM or more using rbd...

FYI for PVE 6.0 live migration with local-disks and with replicated volume is still not possible root@gpu01:~# qm migrate 800 gpu02 --online --with-local-disks --force 2019-10-01 03:27:01 use dedicated network address for sending migration traffic (10.10.1.51) 2019-10-01 03:27:01 starting...

ceph osd down 4 ceph osd rm 4 ceph auth del osd.4

It still require kvm to send discard and ceph is capable of handling discard command and pass it to the corresponded osd so enable discard should be enough.

1035.8 mb/s is your throughput.. 1666.53 mb/s for seq single client is very nice I am getting my results from ceph manager dashboard

rados bench

with 80 OSD you should able to achieve 11 GiB/s, I have 96 OSD setup accross 8 nodes and it is capable of achieving following throughput

machine type needs to be pc-q35-3.1 please remove args: -machine type=q35,kernel_irqchip=on you need to pass pcie 04:00 entire card do you have iommu enabled in /etc/default/grub does your cpu and motherboard support VT-d

Please check my settings - with red line highlights I am using PVE 6.0 with NV 2080ti (you need to roughcast entire pcie card including audio and usb port)

I am using PVE 6.0 with NV 2080ti please check my setting with red line highlights You need to passthrough entire PCIe device including audio and also usb

pve5.4 or pve6.0? a good place to start with https://pve.proxmox.com/wiki/PCI(e)_Passthrough

this can be done using lxc.net.[ i ].type=phys

JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this) at /usr/share/perl5/PVE/CLI/pvesh.pm line 125

I keep getting pvesh error complaining json malformat /usr/share/perl5/PVE/CLI/pvesh.pm @ line 125 sub proxy_handler { my ($node, $remip, $path, $cmd, $param) = @_; my $args = []; foreach my $key (keys %$param) { next if $key eq 'quiet' || $key eq 'output-format'; # just to...

yes I have same error and I realize that I need to move cloud-init drive to local storage and not put them on share storage, then, live migration/HA will work

I noticed that on PVE 5.4 if you have a vm that uses cloud-init drive and the selected storage is ceph or other shared storage, live migration/HA will fail due to the destination already have the file exist

did you select mdev?