Recent content by Humbug

Hallo, nachdem ich auf Kernel 5.11 aktualisiert habe mag icinga keine diskstats via procfs mehr auslesen: perl -MData::Printer -MSys::Statistics::Linux::DiskStats -E'my $lxs = Sys::Statistics::Linux::DiskStats->new; $lxs->init; p $lxs->get;' Sys::Statistics::Linux::DiskStats: no diskstats...

Hi MikeQ, thanks a lot for sharing this workaround!

I was able to fix the problem. - changed clocksource for VMs (clocksource=acpi_pm, was kvm_clock previously) - on the virtualization host /etc/sysctl.conf # 0 avoids swapping processes out of physical memory for as long as possible. vm.swappiness = 0 # lessen the IO blocking effect that...

Hallo nochmal, ich habe das Problem inzwischen lösen können. Die letzten beiden Änderungen, die vielleicht den Ausschlag gegeben haben könnte waren: - Dateisystemüberwachung auf dem Virtualisierungsserver erhöht (fs.inotify.max_user_instances=256) - Zeitgeber auf den VMs geändert...

Lucio, i added your request for a qcow2 cluster size parameter to the ticket https://bugzilla.proxmox.com/show_bug.cgi?id=1989 What application is your picture taken from?

I filed a feature request for qcow2 l2-cache-size parameter here: https://bugzilla.proxmox.com/show_bug.cgi?id=1989

Thanks! I filed a feature request for qcow2 l2-cache-size parameter here: https://bugzilla.proxmox.com/show_bug.cgi?id=1989

I filed a feature request for qcow2 l2-cache-size parameter here: https://bugzilla.proxmox.com/show_bug.cgi?id=1989

There is a qemu parameter to set the size of qcow2 L2 cache properly (default is 8 GB disk size max) which has a dramatic impact on performance, but it is currently unusable with Proxmox PVE...

Is there any workaround? E.g. can i create the qcow2 image manually, so i don't need to add the parameter later? And is there any plan to integrate this parameter as a Proxmox PVE feature in future?

Looks like you're currently running deadline (the one in the brackets is active). Yes, give it a try.

I just read that ionice generally does not work on NFS: https://forum.proxmox.com/threads/about-of-ionice-in-vzdump.16485/#post-84955 Maybe you can use another location instead? As far as i know all other availble io schedulers do not honour ionice parameters, so you would have to use CFQ...

Were you able to fix the problem? Did changing /proc/sys/fs/inotify/max_user_watches help?

I'm not an expert but i have similar problems (not involving NFS though). What you could try as a workaround is a) install the CFQ i/o scheduler on the host and then b) use vzdump in combination with ionice parameter in the shell. See man vzdump.

I installed latest BIOS and intel-microcode package. Can i still use "noibrs noibpb" kernel parameters to disable these functions? Or do i have to downgrade BIOS again?