Why SSH is useful next to the web interface
Proxmox VE runs on Debian, so a host can be checked with normal Linux tools and with Proxmox commands. SSH still works when the web interface on port 8006 is not responding, and it makes it possible to see host problems that the guests only feel as slowness. The commands here run on the host and most of them need root.
Host and version
Start with the version, the load and the state of the Proxmox services. In a cluster, also check quorum, because without it guests cannot be started or changed.
pveversion
uptime
systemctl --failed
systemctl status pveproxy pvedaemon pvestatd --no-pager
pvecm statusVirtual machines and containers
qm covers virtual machines and pct covers containers. pvesh reads the same API that the web interface uses, so it also returns cluster-wide data in a form suitable for scripts.
qm list
pct list
pvesh get /cluster/resources --type vm --output-format jsonStorage: df is not enough
df shows mounted file systems only, so it does not show how full an LVM-thin pool or ZFS pool is. A full thin pool can stop or damage guests, and it is often overprovisioned, so watch its data and metadata use. For ZFS, check the health and capacity of the pool.
pvesm status
sudo lvs -a
zpool status -x
zpool listMemory on a ZFS host
The ZFS ARC cache lives outside the normal page cache, so a host with ZFS may seem almost out of memory in free even when it is healthy. Read the ARC size separately. A virtual machine appears on the host as a kvm process, whose memory may stay allocated after the guest has freed it inside, so host and guest numbers do not have to match.
awk '$1=="size"{print $3/1024/1024/1024" GiB"}' /proc/spl/kstat/zfs/arcstats
free -h
ps -eo pid,rss,cmd --sort=-rss | grep '[k]vm' | head -5Logs and backups
A failed backup or a disk error often shows up first in the journal. Check tasks and logs regularly, and do not rely only on a green status in a dashboard. SMART data needs smartmontools, which Proxmox installs by default, but disks behind a RAID controller may need extra options.
journalctl -p err -b --no-pager | tail -30
pvesh get /nodes/$(hostname)/tasks --errors 1 --limit 10
smartctl -H /dev/sdaWhat CH-J Server Manager shows on Proxmox
System Monitor has no Proxmox-specific logic. It treats the host as a Linux server and shows CPU, RAM, swap, file systems, uptime and network interfaces. It does not list guests, does not use the Proxmox API and counts ZFS ARC as used memory. For qm, pct and pvesm, use the SSH terminal. Disk Manager recognises Proxmox storage: guest volumes are labelled, and destructive changes are not offered for protected host storage such as LVM, ZFS or the system disk.