Search results

Hello, i asked about this from a linux openzfs developer, and he says it will probably not work, and then said that he is planning to start using virtiofs himself at some point, and when he does he will take a look at DAX, and the possibility of adding openzfs support for it.

Hello, i'm about to try this, and i will read up on https://virtio-fs.gitlab.io/howto-qemu.html but just wanted to ask in advance, if you actually tried enabling NUMA, and what happened? I want to use virtiofs in a VM that has NUMA enabled, and has cpu pinning because it has a GPU with pcie...

I know i'm repeating myself, but imo i think adding numactl options to simply bind a VM to a NUMA node, and to the RAM attached to that node would be an alternative more simple solution, as from what i understand the linux kernel is very good at choosing which cores to utilize automatically. I...

Well as @Dot 's patches now have been accepted, you can now try pinning your GPU passthrough VM to certain cores. The main idea is to use the cores of the CPU that the PCIE lanes are connected to. I don't know how the the AMD cpu's work, but you can install hwloc, which includes lstopo, which...

Just updated PVE and now i see CPU AFFINITY setting. Great work @Dot, congratulations.. Now about those numactl settings.. :D

So did the micro stuttering stop with numactl pinning as well? My aim is to find out if pinning cores is really necessary, versus pinning numa nodes, in order to simplify setups, and not needing to "count and manage cores" Also once again i'm an amateur in this, what i'd like to know is how to...

Hi, again i must state i'm no expert.. but a few things keep bothering me about this pinning stuff. Most of the people here that want to pin processes to certain cores need it for gaming setups, or to get maximum performance of other pcie devices such as storage controllers or nvme disks. But...

Edit2: to add to my questions below, i'm not sure that taskset is the correct tool to use for this issue at all. It seems to me that in the beginning of this thread the taskset tool was chosen without much consideration at all: from taskset manpage: And from numactl manpages: So i ask, are...

Hello, and thankyou @Sandbo for giving me a starting point. I have expanded the service a bit to use with my Solarflare SFN7022 NIC's Blacklisting the sfc.ko driver was not an option for me, since host uses the physical NIC ports as well, plus that would prevent me from creating more VF's on the...

Thankyou for this. Based on your response i also find it not worth implementing on my setup. I try to implement all reasonable optimisations, but with a focus on usability and stability, 10% extra performance for the price of a highly 'undynamic' system is not worth it. My case is a 2x Xeon...

Hello, i wanted to quickly ask, should i still enable hugepages by adding hugepagesz=1G default_hugepagesz=1G to KCL in latest PVE? I can't find alot of information about anon hugepages, and if it's still recommended to add the above kernel command line.. Thx.

It would be nice to be able to set this in the CPU settings for VM, but in the meantime i found this https://github.com/ayufan/pve-helpers It allows you to assign cpu cores to a VM, as well as set irq affinity using the 'Notes' area for a VM. Edit: based on preliminary testing, with a USB...

Hello I'm running a VM with pcie passthrough to a nvidia quadro GPU, and trying to set cpu and ram affinity to the same NUMA node that hosts the GPU (numa 0) It's my understanding both the RAM and the CPU's should be run only on the same NUMA node the GPU is attached to.. As i understand it...

Just stumbled upon this thread and have to comment, because i don't see the issue here.. if cloning from a snapshot is not implemented because of problems dealing with clones that are dependent on a parent, then you can just zfs send | receive from the snapshot to a new zvol on the same storage...

hi. As a quick reply, since your zpool was not exported cleanly, you will need to import it once using the -f (force) switch, you could do it for example by using a zfs version of the sysresccd found here http://xyinn.org/gentoo/livecd/ You can also ask for advice in #zfsonlinux on freenode irc...

Hi, i'm running drbd9 in testing as well, it looks very promising.. don't know the solution to your issue, but generally, try #drbd irc channel on freenode, and the documentation at http://drbd.linbit.com/users-guide-9.0/ in your case first, do: # drbdadm adjust all on all nodes, and read...

A short note, after enabling cluster firewall if using drbd9, all connection will be lost. Suggestion is to add in Datacenter -> firewall -> ipset, create ipset drbd9, add ip of all drbd9 nodes. Then in rules, create IN, ACCEPT, make sure the interface is correct, destination +drbd9, leave rest...

Thankyou. after dist-upgrade this seems to be the case.

hi i'm also getting these in my logs: systemd[1]: Cannot add dependency job for unit watchdog-mux.socket, ignoring: Unit watchdog-mux.socket failed to load: No such file or directory. systemd[1]: Failed to reset devices.list on /system.slice: Invalid argument # systemctl status...

regarding i/o error handling of drbd, it can be configured to pass on i/o errors to the upper layer or to automatically detach. see http://drbd.linbit.com/users-guide-9.0/ch-troubleshooting.html#s-hard-drive-failure