Proxmox

Links

• Proxmox VE Administration Guide
• https://pve.proxmox.com/wiki Wiki
• Monitoring Proxmox with Zabbix
• Proxmox Backup Server
• Backup and Restore
• Proxmox Cluster with Raspberry Pi as QDevice (outdate)
• External vote ( like a raspberry pi)
• Useful proxmox commands
• Proxmox VE helper scripts
• Proxmox bug tracker

Proxmox and shared storage

• blockbridge
• Proxmox vs ESXi Storage Performance - Tuning iSCSI?

Commands

Get PVE version

pveversion -v  | head -n 2

pvesm Storage manager

Check

pvesm scan 

pvecm

About pvecm output

A = Alive, NA = Not Alive
V = Vote, NV = Not Vote
MW = Master Wins, NMW = Not Master Wins [0]
NR = Not Registered

pvesh

pvesh get /cluster/resources

pvesh get /cluster/resources --output-format json-pretty

Get backup jobs

pvesh get /cluster/backup

then

pvesh get /cluster/backup/{backupid}

To change the vms included in the job:

pvesh set /cluster/backup/{backupid} -vmid 100,101,102

Get backup errors

pvesh get /nodes/pve01/tasks --typefilter vzdump --errors

Documentation

Proxmox API

• Proxmox VE API
• Proxmox API viewer

View api: https://your.server:8006/api2/json/ ?

Invalid token name

pve uses separator '=', but pbs wants ':'

Ballooning

• Dynamic Memory Management

Ballooning memory limit 80%:

autoballooning is done by one of our daemons (pvestatd) and this limit is hardcoded at the moment

Directory structure

/etc/pve/qemu-server

The VM configs

vmstate? sems related to snapshots

/var/lib/vz

/var/lib/vz/template/iso

 

 

Proxmox cluster

https://pve.proxmox.com/wiki/Cluster_Manager

Cluster manager

pvecm status
pvecm nodes

 

HA status

ha-manager status

VM configuration

Hard disk caching

"If you're using ZFS as your backing store, you should leave the vdisk caching set to 'No cache' (default)."

Monitoring proxmox cluster with zabbix

https://github.com/takala-jp/zabbix-proxmox

Terms

vram_allocated (maxmem)

Maximum amount of ram a VM is allowed to use

vram_used (mem)

Memory being used by VM

Terms

vram

maximum amount of memory a vm may use

lrm

Local Resource Manager

HOWTO

Maintenance

Rebooting a node

If HA enabled check https://pve.proxmox.com/wiki/High_Availability#ha_manager_node_maintenance If you don't want it to start migrating, 'Freeze' might be the right option for HA Settings.

otherwise just do it :)

Disk cache for guest

• https://pve.proxmox.com/wiki/Performance_Tweaks#Disk_Cache
• https://forum.proxmox.com/threads/disk-cache-wiki-documentation.125775/

Show vm configuration

qm config  101

Get VM name by ID

grep '^name:' /etc/pve/nodes/*/qemu-server/$ID.conf | awk '{print $2}'

or

pvesh get /cluster/resources -type vm --output-format yaml | egrep -i 'vmid|name' | sed 's@.*:@@'

or

grep "name:" /etc/pve/nodes/*/*/<vmid>.conf | awk '{ print $2 }'

Clustering

Show cluster status

pvecm status

It seems relatively safe to restart corosync

View cluster logs

pvesh get /cluster/tasks --output-format=json-pretty

Sysctl settings for kvm guests

Still investigating, going for /etc/sysctl.d/50-kvmguest.conf

vm.vfs_cache_pressure=30
vm.swappiness=5

Installing proxmox via PXE

https://github.com/morph027/pve-iso-2-pxe

Storage

Adding another thin pool

lvcreate -L 500G --thinpool newpool vg1

after creating lvm thin pool (TODO: link to that) add to /etc/pve/storage.cfg

lvmthin: lvm-raid10
       thinpool raid10pool
       vgname raid10
       content images

Get OS information of guest

qm guest cmd 105 get-osinfo

Disks

Identify disks in linux guest

lsblk -o +SERIAL

Run fstrim from host

Assuming agent is running:

qm agent 102 fstrim

Suspend or hibernate

Suspend

Suspend does not turn off your computer. It puts the computer and all peripherals on a low power consumption mode. If the battery runs out or the computer turns off for some reason, the current session and unsaved changes will be lost.

qm suspend

in GUI: sleep

qm status: paused

Hibernate

Hibernate saves the state of your computer to the hard disk and completely powers off. When resuming, the saved state is restored to RAM.

qm suspend to disk

or in GUI: Hibe

Backups

proxmox-backup-client

export PBS_REPOSITORY="backup@pbs@pbs-server:backuprepo"

proxmox-backup-client snapshot list

proxmox-backup-client prune vm/101 --dry-run --keep-daily 7 --keep-weekly 3

proxmox-backup-client garbage-collect

vzdump limit bandwidth

--bwlimit 50000

it looks like that limits read speed, i also noticed that bad write/speed to PBS has bad effects on guests or nowadays in /etc/vzdump.conf:

bwlimit

Get total memory allocated to vms

grep memory: /etc/pve/nodes/*/qemu-server/*conf|awk '{sum+=$2} END {print sum}'

nvidia on proxmox

• Upgrade problem to 8.2 (6.8 kernel - nvidia drivers)

FAQ

Web interface stuck on "loading"

When clicking on guest on a particular node

Works on webui of that node

Different versions of PVE?

Console: unable to find serial interface

Maybe you're trying to get console on guest of another node in your cluster. To investigate why this goes wrong.

Cores or threads?

What's called "core" in the Web UI is a core from guest point of view, it would probably be a thread on the host.

Cloud-init

No CloudInit Drive found

See https://gist.github.com/aw/ce460c2100163c38734a83e09ac0439a

==The following packages have been kept back:

 proxmox-kernel-6.8==

Don't use apt-get upgrade but

apt-get dist-upgrade

Error messages

create storage failed: storage 'XX' is not online (500)

When trying to create a storage on NFS

memory: hotplug problem

a used vhost backend has no free memory slots left

echo "options vhost max_mem_regions=509" >> /etc/modprobe.d/vhost.conf

and reboot

Proxmox API call failed: Couldn't authenticate user: zabbix@pve

Funky characters in password string?

memory: hotplug problem - 400 Parameter verification failed. dimm17: error unplug memory module

bad!

Failed to establish a new connection: [Errno -2] Name or service not known

Just that, check your DNS

ConditionPathExists=/etc/corosync/corosync.conf was not met

Problably set up node with bad /etc/hosts, or forgot to join cluster

https://forum.proxmox.com/threads/cluster.103370/

https://blog.jenningsga.com/proxmox-keeping-quorum-with-qdevices/

https://pve.proxmox.com/pve-docs/chapter-pvecm.html#_corosync_external_vote_support

corosync-qdevice[11695]: Can't read quorum.device.model cmap key

On the qdevice node

Check corosync-cmapctl ?

also see https://pve.proxmox.com/pve-docs/chapter-pvecm.html#_corosync_external_vote_support

"Quorum: 2 Activity blocked"

In my case this meant boot up second real node first

 

 

On working node:

corosync-cmapctl | grep quorum.device
quorum.device.model (str) = net
quorum.device.net.algorithm (str) = ffsplit
quorum.device.net.host (str) = 192.168.178.2
quorum.device.net.tls (str) = on
quorum.device.votes (u32) = 1

https://bugs.launchpad.net/ubuntu/+source/corosync-qdevice/+bug/1733889

x86/split lock detection: #AC: kvm/1161956 took a split_lock trap at address: 0x7ffebcb378ab

The number after kvm/ is process it, this will help you find the culprit.

See:

• In-depth analysis of split locks
• Detecting and handling split locks
• The search for the correct amount of split-lock misery

Could be large Windows guests on NUMA. Can probably be ignored

Shutting down a node

Should just work. Takes guests down with it when they're not in HA

API calls

List all vms in cluster

pvesh get /cluster/resources --type vm --output-format yaml | egrep -i 'vmid|name'

or json:

pvesh get /cluster/resources --type vm --output-format json | jq '.[] | {id,name}'

Cores, sockets and vCPUs

vCPUs is what the vm uses, maximum is sockets*cores but you could set it lower to allow adding cores/vcpus dynamically.

Migrating

VM is locked (create) (500)

Not always clear why, but try

qm unlock 111

CT is locked (snapshot-delete) =

pct unlock 115

Replication

missing replicate feature on volume 'local-lvm

looks like replication of lvm isn't supported

Check if qemu agent is running

See if IP is shown under Summary, also

qm agent 105 ping

or

qm guest cmd 111 ping

qm agent ping return values

0: OK

2: VM not running

255: No QEMU guest agent configured (just disabled in vm config?) (QEMU guest agent is not running would only show when enabled in in config?)

There is no way to tell if agent is running when it's not enabled in VM config.

When VM is not running, GUI claims agent not running

Move to unused disk

If you moved disk, and decided to move back to the old one:

• detach current disk
• select the unused disk
• click Add

qemu-guest-agent.service: Job qemu-guest-agent.service/start failed with result 'dependency'.

Could mean QEMU guest agent is not enabled in vm config

Stop all proxmox services

systemctl stop pve-cluster systemctl stop pvedaemon systemctl stop pveproxy systemctl stop pvestatd

Storage (xx) not available on selected target

probably some storage mounted only on one node, so not clustered

switch to community repository

cat /etc/apt/sources.list.d/pve-enterprise.list 
#deb https://enterprise.proxmox.com/debian/pve buster pve-enterprise

echo "deb http://download.proxmox.com/debian/pve buster pve-no-subscription" > /etc/apt/sources.list.d/pve-no-subscription.list 

apt update

W: (pve-apt-hook) You are attempting to remove the meta-package 'proxmox-ve'!

cehck sources.list :)

 

Storage

Could not determine current size of volume

When trying to grow a disk another secret!

Add local disk or LV to vm

That would be passtrough

qm set 101 -scsi1 /dev/mapper/somevolume

Make sure node node can't migrate: ?? PVE won't try that anyway, but still

 

Storage migration failed: block job (mirror) error: drive-scsi0: 'mirror' has been cancelled

Maybe moving disk to LVM, check for 4MiB alignment. qemu-img resize to 4MiB aligned size.

fstrim guests

qm guest <ID> fstrim

qmp command 'guest-fstrim' failed - got timeout

seems to be a windows thing

No disk unused

when trying to create thin volume, use command line?

qcow image bigger than assigned disk

Probably snapshots

Backups

backup write data failed: command error: protocol canceled

Temporary network failure?

storing login ticket failed: $XDG_RUNTIME_DIR must be set

Temporary bug, ignore it

PBS GC & Prune scheduling

https://pbs.proxmox.com/docs/prune-simulator/

qmp command 'backup' failed - got timeout

https://github.com/proxmox/qemu/blob/master/qmp-commands.hx

 

dirty-bitmap status: existing bitmap was invalid and has been cleared

• https://qemu-project.gitlab.io/qemu/interop/bitmaps.html

unexpected property 'prune-backups' (500)

When for example Add: iSCSI Uncheck "Keep all backups" in "Backup retention"

 

FAILED 00:00:02 unable to activate storage

TODO

 

VM 101 Backup failed: VM is locked (snapshot)

Check if there's no snapshot running (how?)

qm unlock 101

qmp command 'blockdev-snapshot-delete-internal-sync' failed - got timeout

Another job for

qm unlock 101

qmp command 'blockdev-snapshot-delete-internal-sync' failed - Snapshot with id 'null' and name 'mysnapshot' does not exist on device 'drive-scsi1'

Verify there is no such snapshot at all:

qemu-img snapshot -l vm-114-disk-1.qcow2

and then delete the entire system from [mysnapshot] in the vm config file

lvremove snapshot 'xx' error: Failed to find logical volume "pve/snap_vm-103-disk-0_xx"

Most likely the logical volume doesn't exist anymore. no idea how this can happen, but:

qm listsnapshot 103
qm delsnapshow 103 xx --force

will most likely cry about failed to find again, but with some luck:

qm listsnapshot 103

and it should be gone.

If you get

VM is locked (snapshot-delete)

just

qm unlock 103

(probably after checking there's not something else locking it)

can't acquire lock '/var/run/vzdump.lock' - got timeout

Check if vzdump is running, otherwise kill it (cluster?)

You could change lockwait in vzdump.conf, or as --lockwait parameter. Default is 180 minutes

 

VM 101 Backup failed::= VM is locked (snapshot-delete)

Check /etc/pve/qemu-server/101.conf for 'snapstate'

If that says 'delete' for a snapshot try deleting the snapshot:

qm delsnapshot 101 snapname

If that throws like Failed to find logical volume 'pve/snap_vm-101-disk-0_saving

 qm delsnapshot 101 snapname --force

If it says VM is locked (snapshot-delete) us

qm unlock XXX

When you get does not exist on device 'drive-scsi0 you might also need to delete the line "lock: snapshot-delete" from the 101.conf file

Restoring single file from (PBS) backup

Check Mounting of archives with fuse

Requires package proxmox-backup-file-restore:

proxmox-file-restore

proxmox-file-restore failed: Error: mounting 'drive-scsi0.img.fidx/part/["2"]' failed: all mounts failed or no supported file system (500)

Maybe because of lvm?

Backup log

Upload size

Seems to be in kilobytes

Error: VM quit/powerdown failed - got timeout

qm stop VMID

if that complains about lock, remove the lock and try again

 

You have not turned on protection against thin pools running out of space.

Seems noboby knows how, just monitor it?

serial console from command line

qm terminal <id}

enable serial console in guest

• https://pve.proxmox.com/wiki/Serial_Terminal

looks like this is not needed:

systemctl enable serial-getty@ttyS0.service

in /etc/default/grub

GRUB_CMDLINE_LINUX_DEFAULT="console=ttyS0 console=tty0"

ttyS0 is for qm terminal, tty0 is for the "console" buttion in UI

1. debian based

update-grub

1. redhat based

grub2-mkconfig --output=/boot/grub2/grub.cfg

 

add

serial0: socket

to /etc/pve/qemu-server/[vmid].conf and restart

 

agetty: /dev/ttyS0: not a device

systemctl status useless again, means the serial bit is missing from <vmid>.conf

TASK ERROR: command 'apt-get update' failed: exit code 100

subtle way of telling you to get subscription of at least change the sources list

Import vmdk to lvm

https://pve.proxmox.com/wiki/Qemu/KVM_Virtual_Machines#_importing_virtual_machines_and_disk_images

Can't apply changes to memory allocation

Maybe try enabling NUMA in CPU settings

 

 

Adding hardware shows orange

The keyword here is "PENDING", possibly ion /etc/pve/qemu-server/<id>.conf

Maybe something is not supported (Options->Hotplug), options: reboot or click "revert"

"Connection error 401: no ticket"

Login session expired?

can't lock file '/var/lock/qemu-server/lock-102.conf' - got timeout (500)

Maybe someone else has/had webui open, otherwise just remove it

TASK ERROR: Can't use string ("keep-all=0,keep-last=3") as a HASH ref while "strict refs" in use at /usr/share/perl5/PVE/VZDump.pm line 502.

Classic, means incorrect syntax in your /etc/pve/storage.cfg

 

The current guest configuration does not support taking new snapshots

• You're using raw instead of qcow2. Convert: Hardware->Hard disk "Move Disk"
• you might be using lvm thin over iscsi, then you can't have snapshots

WARNING: Device /dev/dm-21 not initialized in udev database even after waiting 10000000 microseconds.

Until someone fixes it:

udevadm trigger
      

Also look for link to dm-21 in /dev/disk/by-id

"connection error - server offline?"

Try reconnection browser

 

 

Find vm name by id

qm config 100 | grep '^name:' | awk '{print $2}'

or a bit cruder"

grep name: /etc/pve/nodes/*/qemu-server/101.conf |head -n 1 | cut -d ' ' -f 2

Started Proxmox VE replication runner.

??

Find ID by name

grep -l "name: <NAME>"  /etc/pve/nodes/*/qemu-server/*conf| sed 's/^.*\/\([0-9]*\)\.conf/\1/g'
      

 

Can't migrate VM with local CD/DVD

Remove the CD :)

 

 

Memory allocated to VMs

qm list|egrep -v "VM|stopped" | awk '{ sum+=$4 } END { print sum }'

Ceph

Got timeout(500)

Check

pveceph status

Possibly problem with ceph mgr

 

vzdump: # cluster wide vzdump cron schedule

1. Automatically generated file - do not edit

edit it anyway?

iSCSI

iscsid: conn 0 login rejected: initiator error - target not found

pvesm scan iscsi <targetip>

and

iscsiadm -m session -P 3

udev high load

Check

udevadmin monitor

KERNEL[426405.347906] change   /devices/virtual/block/dm-8 (block)
UDEV  [426405.359582] change   /devices/virtual/block/dm-8 (block)

ls -al /dev/mapper/

pve-vm--113--disk--0 -> ../dm-8

So vm/lx '113' is the one.

In general see https://forum.proxmox.com/threads/udev-malfunction-udisksd-high-cpu-load.99169/ since it could be usdisks2

start failed: org.freedesktop.DBus.Error.Disconnected: Connection is closed

Most likely that VM isn't running.