Search results

Performance is always dependent on hardware so if your nics are not quality hardware you will never be able to saturate a Gb uplink. Quality hardware does not include Realtek.

OCFS2 is abandon ware so hardly a candidate for customers running in production!

An even number of nodes is always a bad thing for deciding quorum.

This is strange since the drivers for Mellanox 10Gbit in FreeNAS/FreeBSD is the ported drivers from OmniosCE. I have no problems with either Mellanox 10 or 20 Gbit in both Omnios and FreeNAS.

I would suggest Omnios which is now OmniosCE.

If you do not require your storage to be HA as well ZFS over iSCSI is a very stable solution.

Most enterprise SAN's has the ability to make snapshots of LUN's so instead of making snapshots in proxmox you make the snapshots on the SAN.

And if you shut down two monitor servers?

In generel clusters with an even node numbers is a disaster waiting to happen. Always use odd numbers for nodes in a cluster or in Ceph terms odd numbers for monitors.

nodes for quorum must be physical. Using a VM for quorum is dangerous and plain stupid.

Hi all, Maybe Stratis could be a future storage solution for Proxmox. The solution combines the best parts from ZFS and BTRFS but doing it using existing kernel tools. DM and XFS. Article series: 1) overview: https://opensource.com/article/18/4/stratis-easy-use-local-storage-management-linux...

Ask the devs. I have mentioned it a couple of times.

In the file /usr/share/perl5/PVE/QemuServer.pm find this: if($path =~ m/^iscsi\:\/\//){ $devicetype = 'generic'; } and change it to this if($path =~ m/^iscsi\:\/\//){ $devicetype = 'block'; } After this change...

This is not an option since almost all proxmox operations requires root privileges. You can, however, disable login using password to the proxmox nodes since proxmox's access is certificate based.

If you don't need live migration between nodes and HA you could create a local ZFS pool on LUN's presented to the node from the SAN. An experimental live migration method including local disks exists but is only available from command line.

I was referring to the drivers pciid database;-)

My guess is a missing update of the pciid database.

Have you tried consulting your hardware supplier?