PVE Cluster Node Maintenance Mode?

[cloudguy]

HI Everyone,

Is there a way to invoke maintenance mode on a Proxmox cluster node? I'm looking for the semi-automated functionality of Proxmox being able to:

1. Live Migrate all running machines on a node;
2. Place this node into maintenance mode such that no new VMs can be started on that node until the status is cleared/lifted.

Same as vSphere.

Is this possible today?

 

[spirit]

Not possible for now.
Maybe can you do a feature request in https://bugzilla.proxmox.com ?

I would like to work on it in the coming weeks.

 

[hotwired007]

+1 for live migrate all running machines to another node

 

[tom]

already noted in bugzilla,

see https://bugzilla.proxmox.com/show_bug.cgi?id=173

 

[jleg]

+1 for live migrate all running machines to another node

...our "workaround" for now ;-)

#!/bin/bash
if [ -z $1 ] ; then
 echo "usage: $0 <target-node> [offline]"
 echo "       Migrate all CTs and VMs on local node to target node."
 echo "       Online- (Live-)migration is default."
 echo "       In case live migration fails, please try offline mode."
 exit
fi

if [ ! -d /etc/pve/nodes/$1 ] ; then
 echo "Target node $1 does not exist!"
 exit 1
fi

ping -c 2 -W 2 $1 >/dev/null 2>&1
if [ $? != 0 ] ; then
 echo "node $1 is not reachable."
 exit 1
fi

if [ "x$2" == "xoffline" ] ; then
 OPT=""
else
 OPT="-online"
fi

CFG="/etc/pve/local/"

FILES1=$(find $CFG/openvz -maxdepth 1 -name "*.conf")
FILES2=$(find $CFG/qemu-server -maxdepth 1 -name "*.conf")
FILES="$FILES1 $FILES2"
for VE in $FILES ; do
 V=`basename $VE`
 VEID=`echo $V|cut -f1 -d\.`
 TYPE="OVZ"
 CMD="pvectl"
 if [ `expr match $VE ".*/qemu-server/.*"` -ne 0 ] ; then
  TYPE="KVM"
  CMD="qm"
 fi
 echo "migrating $TYPE: $VEID to $1..."
 $CMD migrate $VEID $1 $OPT
 echo "$VEID done."
done

 

[cloudguy]

Pretty cool script. Thanks.

I filed a bug (enhancement request) already: https://bugzilla.proxmox.com/show_bug.cgi?id=262.

 

[christophe]

Up!

 

[mir]

May I suggest using this version of the script instead?

Changes: Only migrate running CT's and VM's

#!/bin/bash
if [ -z $1 ] ; then
 echo "usage: $0 <target-node> [offline]"
 echo "       Migrate all running CTs and VMs on local node to target node."
 echo "       Online- (Live-)migration is default."
 echo "       In case live migration fails, please try offline mode."
 exit
fi


if [ ! -d /etc/pve/nodes/$1 ] ; then
 echo "Target node $1 does not exist!"
 exit 1
fi


ping -c 2 -W 2 $1 >/dev/null 2>&1
if [ $? != 0 ] ; then
 echo "node $1 is not reachable."
 exit 1
fi


if [ "x$2" == "xoffline" ] ; then
 OPT=""
else
 OPT="-online"
fi


for OVZ in $(pvectl list|grep running|awk '{print $1}'); do
  echo "Migrating OVZ: $OVZ to $1..."
  pvectl migrate $OVZ $1 $OPT
  echo "$OVZ done."
done


for KVM in $(qm list|grep running|awk '{print $1}'); do
  echo "Migrating KVM: $KVM to $1..."
  qm migrate $KVM $1 $OPT
  echo "$KVM done."
done

 

[mo_]

doesnt HA automatically move all the VMs back onto the node youre emptying if you were to use something like that?

 

[m.ardito]

HA could restart those machine elsewhere (AFAIK random node) after they are shut down because their node is shut down. Not live migrated.

Marco

 

[christophe]

Hi,

Right, I'm not at work right now but I patched this kind of script to migrate HA managed VMs 'runnig. Stopped ones did not migrate).

Thanks,

Christophe.

 

[smbtech]

This is SWEET! THX!!!!

 

[cvk98]

https://github.com/cvk98/Proxmox-host-maintenance-mode

This script allows you to prepare the Proxmox host for maintenance. After the launch, you are prompted to select a host. Next, checks are carried out for the availability of free cluster resources, the presence of containers (lxc), as they are rebooted during migration. It also checks the availability of local resources for the VM, because they do not allow the VM to migrate. If all the checks are passed, the migration begins. The result should be a completely free host.