User Tools

Site Tools


proxmox

BIOS Tweaking

IBM lab tests show that enabling the x2APIC support for Red Hat Enterprise Linux 6 guests can result in 2% to 5% throughput improvement for many I/O workloads.

NUMA Info

numactl --hardware

When possible, create smaller VMs, instead of “Monster” VMs, that fit into a single NUMA node.

Disk performance (not for SSD)

echo 100000 > /sys/block/sdX/queue/nr_requests

List CEPH Snapshot

rbd --pool ceph-pool snap ls vm-100-disk-1
rbd snap rm ceph-pool/vm-100-disk-1@vzdump

Windows Preparation

sysprep.exe [/oobe | /audit] [/generalize] [/reboot | /shutdown | /quit] [/quiet] [/unattend:answerfile]

Import HyperV

qemu-img convert -O qcow2 disk.vhdx /var/lib/vz/images/112/vm-112-disk-1.qcow2

BBR

/etc/sysctl.conf

net.ipv4.tcp_congestion_control=bbr

Jumbo Frames

/etc/network/interface

pre-up ip link set <interface name> mtu 9000
post-up ip link set eth2 mtu 9000 && ip link set eth3 mtu 9000 && ip link set bond1 mtu 9000

Instantly ip link set <interface name> mtu 9000

HA Logic

shutdown: stops VMs, then move them to other nodes
reboot: stops VMs, put the into freeze state

Migrate Hyper-V to KVM

kvm-img convert -f vpc -O raw disk.vhd disk.img

OpenVZ / PROXMOX Tips

Proxmox “no subscription” patch
/usr/share/pve-manager/ext6/pvemanagerlib.js Modify

-if (data.status !== 'Active') {
+if (false) {

System limits Hints for Redhat Linux

Set ULIMIT

When you are using large JAVA applications, Databases
in Proxmox Hypervisor edit /etc/security/limits.conf

*    soft nofile 65535
*    hard nofile 65535
root soft nofile 65535
root hard nofile 65535
*    soft stack unlimited
*    hard stack unlimited
root soft stack unlimited
root hard stack unlimited
*    soft memlock unlimited
*    hard memlock unlimited
root soft memlock unlimited
root hard memlock unlimited

Cluster

pvecm create YOUR-CLUSTER-NAME
pvecm status
pvecm add IP-ADDRESS-CLUSTER (from new node)
pvecm nodes
pvecm delnode NAME

ext3 vs ext4

ext3 have better performance, ext4 much faster fsck.

How to edit Proxmox config without quorum

/etc/default/pve-cluster so that DEAMON_OPTS="-l"

(set local mode) and reboot
or try.

systemctl stop pve-cluster
pmxcfs -l 

Change node IP address

/etc/network/interfaces
/etc/hosts
/etc/pve/cluster.conf

Update

edit nano /etc/apt/sources.list and add public repo

# PVE pve-no-subscription repository provided by proxmox.com, NOT recommended for production use
deb http://download.proxmox.com/debian stretch pve-no-subscription

edit nano /etc/apt/sources.list.d/pve-enterprise.list
disable enterprise by hash #

apt-get update && apt-get dist-upgrade

Graphs Problems

Clean directory /var/lib/rrdcached/* then restart.

Network Bond

1) Remove vmbr0
2) Create bond0
3) Assign eth0 eth1 eth2 etc + Balance-RR (Switch support 2Gbit / up-down) or Balance-TLB (any switch 2Gb up / 1Gb down)
4 Create vmbr0 set IP and Bridge ports bond0


cat /proc/net/bonding/bond0

Root minimum size

Approx. 8GB

Speedup migration

datacenter.cfg

migration: type=insecure,network=172.24.16.0/24

KSM Sharing default at 50% used

/etc/ksmtuned.conf

KVM IO

Cache -> Writeback
Discard on -> LVM Thin

Disable the USB tablet device

Do not use balloon driver

IO Scheduler in KVM

Set noop or deadline

Mount ext4 in KVM

/dev/X /c ext4    defaults,noatime,nodiratime,nobh,commit=40,barrier=0,data=writeback 0 2

SSD / LVM-Thin

Edit harddisk setting and check “discard”

Check IOMMU Support

#!/bin/sh
if [ $(dmesg | grep ecap | wc -l) -eq 0 ]; then
  echo "No interrupt remapping support found"
  exit 1
fi

for i in $(dmesg | grep ecap | awk '{print $NF}'); do
  if [ $(( (0x$i & 0xf) >> 3 )) -ne 1 ]; then
    echo "Interrupt remapping not supported"
    exit 1
  fi
done

Single node

You can disable

systemctl disable zed
systemctl disable zfs
systemctl disable zfs-import-scan
systemctl disable zfs-mount
systemctl disable zfs-share
systemctl disable zfs-zed

pve-ha-crm
pve-la-hrm
pve-firewall
pvefw-logger

Convert QCOW2 to RAW

qemu-img convert -f qcow2 -O raw image.qcow2 image.raw
proxmox.txt · Last modified: 2018/07/30 10:49 by Jan Forman