PingWise connects to Proxmox VE with a small script on each node. It asks Proxmox itself how the node and its guests are, with pvesh, which root can use without a password, and sends that to us every two minutes. No API token is created, we hold no password of yours, and the web interface on port 8006 can stay closed to the internet.
Connecting
Integrations → Connect Proxmox VE. The next page shows a few commands with a key, once. Paste them as root on every node of the cluster, over SSH or in the node's Shell in the Proxmox web interface. One connection is one cluster, or one standalone node.
The connection's page then lists the nodes that reported, the guests found and the monitors that would be added. Nothing is created until you press Import and keep in sync. Proxmox VE 8 and 9 are supported.
What you get
For each node:
- a ping monitor on its public address;
- a status monitor, down when the node has not reported for 6 minutes;
- warnings for CPU and memory that stay high, a storage filling up, a drive failing SMART, a worn SSD or NVMe drive, a ZFS pool that is not ONLINE and a degraded software RAID;
- a page under Servers with charts, drives and guests.
For each VM and container:
- a status monitor taken from Proxmox: running is up; stopped but set to start at boot is down; stopped otherwise is paused;
- a ping monitor when it has a public address. A VM's addresses come from the QEMU guest agent when it runs inside the VM, a container's from Proxmox. A guest with only private addresses is listed as not reachable from the internet;
- warnings for CPU and memory that stay high.
Templates are skipped.
Tags
Add a tag to a guest in Proxmox to change what we do with it:
pingwise-sshadds an SSH check on port 22;pingwise-httpadds an HTTP check;pingwise-offleaves the guest out.
One alert for a node
A guest's monitors wait for their node. While the node is down, its guests' alerts are held and you get one alert for the node, not one per guest.
Keeping in step
A new guest gets its monitors at the next report. A stopped guest has its monitors paused, and resumed when it runs again. A guest that is deleted, or a node that leaves the cluster, has its monitors paused at once and removed 24 hours later. A guest that migrates to another node moves with it. If a report would pause or remove more than a fifth of everything at once, nothing is changed and we look into it.
What the script sends
You can read the whole script on the install page. It sends the node's name and IP addresses (public and private), CPU and memory use, storage use, each drive's model, SMART health and wear, ZFS and RAID state; and for each guest its id, name, type, status, tags, CPU and memory use and addresses. It never sends consoles, configuration, passwords, or anything inside your disks or guests.
Removing it
On each node, as root:
rm /etc/cron.d/pingwise-sync /usr/local/sbin/pingwise-sync /etc/pingwise-sync.conf
Then disconnect the connection, or remove a single node on its page under Servers.
Last change: 11 November 2026.