Proxmox

Tonton Jo

Join the community:

Support my work, give a thanks and help the youtube channel:

Informations:

This repository contains somes scripts and tips for Proxmox hypervisor and backup server

Find here more of my videos about proxmox

Proxmox Toolbox - all-in-one setup tool!

If you want to setup and update your Proxmox VE / BS server, please have a look at my Proxmox Toolbox

• Set no subscription sources
• New update command (proxmox-update) wich automatically remove message if needed
• Security with Fail2ban
• snmp
• mail notifications
• Backup and restoration of Proxmox VE / BS configurations

CPU_SCALE

A home server may be power consuming and i wanted something to have a compromise between performance and energy saving without having to power-off the server Here it is: CPU_SCALE.SH

• The script, when configured to start at boot or manually executed for testing purpose, will check for the cpu load and ramp-up to a better CPU Gouvernor when calculation power is needed.
• You can configure the ramp-up and ramp-down time
• You can choose what CPU gouvernor you want to use (i recommand switching between schedutil and powersave)

Here's a power consumption report for my server:

Usefull commands:

You'll find there some usefull commands used for proxmox and more generally debian

Menu:
1 - Debian General
1.1 - System
1.2 - Drives Management
1.3 - Zpool Management
1.4 - Monitoring
1.5 - PCI express

2 - Proxmox Virtual Environement
2.1 - Services management
2.2 - Proxmox commands
2.3 - VM Management

3 - Proxmox Backup Server
3.1 - Recover / add an existing datastore

4 - Usefull tools

1 - Debian General

1.1 - System

1.1.1 - Date and time:

dpkg-reconfigure tzdata

timedatectl set-ntp true

1.2 - Drives Management

1.2.1 - Find a disc with ID:

ls /dev/disk/by-id/ -la

ls /dev/disk/by-id/ -la | grep "serial"

1.2.2 - list disk informations: Replace X

lsblk -o name,model,serial,uuid /dev/sdX

1.2.3 - find disk UUID or partition UUID

ls -l /dev/disk/by-uuid

ls -l /dev/disk/by-partuuid

1.2.4 - Wipe Disk

wipefs -af /dev/sdX

1.2.5 - Read actual partition status after change

hdparm -z /dev/sdX

echo 1 > /sys/block/sdX/device/rescan

1.2.6 - Find the actual blocksize of all disks - Usually 4k

lsblk -o NAME,PHY-SeC

1.3 - Zpool Management

1.3.1 - Remove import of removed pools at startup:

• Identify your pools:

systemctl | grep zfs

• Remove import of a pool at boot (remove old pools from load)

systemctl disable zfs-import@zpoolname.service

1.3.2 - Find ARC RAM usage for Zpool:

awk '/^size/ { print $1 " " $3 / 1048576 }' < /proc/spl/kstat/zfs/arcstats

1.3.3 - Find Compression ratio and used space:

zfs list -o name,avail,used,refer,lused,lrefer,mountpoint,compress,compressratio

1.3.4 - Replace Zpool Drive:

zpool replace pool /old/drive /new/drive

1.3.5 - Mark a pool as OK - Clear errors on pool and drives

zpool clear "poolname"

1.3.6 - Get Zpool version:

zpool --version

1.3.7 - Ugrade a zpool:

zpool upgrade "$poolname"

1.3.8 - Zpool options settings

• Get your $poolname

zfs list

• find actual options - filter if necessary

zfs get all 

zfs get all | grep atime

• Set new option value

zfs set atime=off $poolname

zfs set compression=off $poolname

1.3.9 - If pool is created with /dev/sdX instead of ID (wich may lead to mount fail)

zpool export $poolname

zpool import -d /dev/disk/by-id -aN

1.4 - Monitoring

1.4.1 - ZFS live disk IO

watch -n 1 "zpool iostat -v"

1.5 - PCI express

1.5.1 - Determine bus speed of a PCI-E Device ( and others infos if you remove the grep part )

• Identify your device:

lspci

• Get infos

lspci -vv -s 2a:00.0

1.5.2 - Find IOMMU groups

An IOMMU group is the smallest set of physical devices that can be passed to a virtual machine

#!/bin/bash
shopt -s nullglob
for g in $(find /sys/kernel/iommu_groups/* -maxdepth 0 -type d | sort -V); do
    echo "IOMMU Group ${g##*/}:"
    for d in $g/devices/*; do
        echo -e "\t$(lspci -nns ${d##*/})"
    done;
done;

1.6 - CPU

1.6.1 - Change CPU Gouvernor

Informations:

https://www.kernel.org/doc/Documentation/cpu-freq/governors.txt

Get actual CPU gouvernor

cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor

List availables CPU Gouvernors

cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_governors

Apply CPU Gouvernor

echo "ondemand" | tee /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor

Make CPU Governor reboot resilient

crontab -e

@reboot echo "ondemand" | tee /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor >/dev/null 2>&1

1.7 - Benchmark and performance test

1.7.1 - Test with DD - oflag=dsync -> ignore cache for accurate result

dd if=/dev/zero of=/$pathtostorage/test1.img bs=1G count=1 oflag=dsync

1.7.2 - Test speed of drive (hdparm needed)

hdparm -t /dev/$sdX

2 - Proxmox Virtual Environement

2.1 - Services management

2.1.1 - Stop all services:

for i in pve-cluster pvedaemon vz qemu-server pveproxy pve-cluster; do systemctl stop $i ; done

2.2 - Proxmox commands

2.2.1 - Regenerate console login screen (usefull after ip changes)

pvebanner

2.3 - VM Management

2.3.1 - Disk passtrough

qm set $vmid -scsi0 /dev/sdX

2.3.2 - Appliance Import

If OVA: extract content

tar -xvf "path_to_appliance.ova"

qm importdisk $vmid path_to_appliance_disk.vmdk local-lvm

2.3.3 - Appliance Export

• Identifiy the disk of a vm

qm config $vmid

• Check absolute path ($absolutepath) for drive as seen by the OS

absolutepath=$(pvesm path local-lvm:vm-100-disk-0)

echo $absolutepath

• Export the drive wanted, watch for the format to be correct :-)

qemu-img convert -O qcow2 -f raw $absolutepath OUTPUT.qcow2

3 - Proxmox Backup Server

3.1 - Recover / add an existing datastore

• edit "/etc/proxmox-backup/datastore.cfg"

datastore: backup
	gc-schedule daily
	keep-daily 1
	keep-last 1
	keep-monthly 5
	keep-weekly 5
	path /mnt/datastore/backup
	prune-schedule 2,22:30

Ensure the rights are corrects:

chown backup:backup -R "/path/to/datastore"

4 - Usefull tools

Ioping - usefull to simulate drive activity and therefore locating it.

s-tui - Graphical interface to monitor system performances

stress - install it with s-tui to be able to stress your system - dont use in prod