Skip to content
Proxmox Cluster Config

Proxmox Cluster Config

Proxmox Cluster

Edit VM settings when Cluster Quorum is not able to be reached

If one node in a two node cluster is down then cluster quorum will not be able to be reached. In this case use the following command to set the cluster to only require one vote for changes. When doing this ensure to not make any changes to cluster configuration (adding new nodes to cluster etc)

1
pvecm expected 1

Proxmox Two-Node Cluster – Running With One Node Offline

In a two-node Proxmox cluster, shutting down one node causes the remaining node to lose quorum.

When quorum is lost, /etc/pve becomes read-only and configuration changes may fail with errors such as:

1
unable to open file '/etc/pve/nodes/<node>/qemu-server/<vmid>.conf.tmp' - Permission denied (500)

Check Cluster Quorum

Run:

1
pvecm status

A two-node cluster with one node offline may show:

1
2
3
4
Expected votes: 2
Total votes:    1
Quorum:         2
Quorate:        No

Temporarily Allow Single-Node Operation

Only do this when you are certain the other Proxmox node is powered off or otherwise intentionally unavailable.

1
pvecm expected 1

Check the result:

1
pvecm status

You should now see:

1
2
3
Expected votes: 1
Total votes:    1
Quorate:        Yes

The remaining node can now modify VM configurations and /etc/pve normally.

Test /etc/pve Write Access

Optional test:

1
2
touch /etc/pve/test
rm /etc/pve/test

If this succeeds, /etc/pve is writable.

When the Second Node Returns

Normally no manual reset is required.

When the second Proxmox node rejoins the cluster, the expected vote count should automatically return to:

1
Expected votes: 2

Verify with:

1
pvecm status

Important Warning

pvecm expected 1 disables the protection normally provided by quorum.

Do not use it if there is any possibility that both nodes are running but unable to communicate with each other.

That situation could allow both nodes to independently believe they control the cluster, creating a split-brain condition.

Safe example:

1
2
3
4
host-1    OFF
host-2    ON

pvecm expected 1

Unsafe example:

1
2
3
4
5
host-1    ON
host-2    ON
Cluster network broken

DO NOT force expected votes to 1

After a Reboot

pvecm expected 1 is temporary.

If the surviving node reboots while the other cluster node is still offline, quorum will normally be lost again.

Run:

1
2
3
pvecm status
pvecm expected 1
pvecm status

Recommended Long-Term Solution

If one node will regularly be powered off, configure a QDevice on a separate always-on Linux machine.

The voting arrangement then becomes:

1
2
3
host-2    1 vote
host-1    1 vote
QDevice   1 vote

Quorum requires two votes.

This means either Proxmox node can be powered off while the remaining node and QDevice maintain quorum.

1
2
host-2 + QDevice = Quorate
host-1 + QDevice = Quorate

A QDevice is preferable to regularly using:

1
pvecm expected 1