Scheduling

cv4pve-autosnap takes one snapshot per guest each time it runs and has no scheduler of its own: the schedule is the one of your system. Run it from a machine outside the cluster that is always on (a management VM, a server of the monitoring), so the snapshots do not depend on a single node.

Put the connection in a parameter file

The token should not be visible in the crontab or in the list of processes. Write the connection options in a file, one per line, readable only by the account that runs the job:

/etc/cv4pve/autosnap.conf

--host=pve1.local,pve2.local
--api-token=autosnap@pve!snap=<uuid>
--vmid=@tag-snapshot

chmod 600 /etc/cv4pve/autosnap.conf

and pass it with @, followed by the command:

cv4pve-autosnap @/etc/cv4pve/autosnap.conf snap --label=daily --keep=7

In a file the value needs no quotes. Global options such as --max-parallel can go in the file too. The rules of the file are in Options in a file.

Schedules

Linux (cron)

/etc/cron.d/cv4pve-autosnap

# every 2 hours, keep one day
0 */2 * * *  cv4pve  /usr/local/bin/cv4pve-autosnap @/etc/cv4pve/autosnap.conf snap --label=2hourly --keep=12 >> /var/log/cv4pve-autosnap.log 2>&1
# every night at 01:30, keep one week
30 1 * * *   cv4pve  /usr/local/bin/cv4pve-autosnap @/etc/cv4pve/autosnap.conf snap --label=daily --keep=7 >> /var/log/cv4pve-autosnap.log 2>&1
# Sunday at 03:15, keep one month
15 3 * * 0   cv4pve  /usr/local/bin/cv4pve-autosnap @/etc/cv4pve/autosnap.conf snap --label=weekly --keep=4 >> /var/log/cv4pve-autosnap.log 2>&1

Files in /etc/cron.d take the user as sixth field; here a dedicated cv4pve account that can read the parameter file.

Windows (Task Scheduler)

schtasks /Create /TN "cv4pve-autosnap daily" /SC DAILY /ST 01:30 /RU SYSTEM `
  /TR "C:\cv4pve\cv4pve-autosnap.exe @C:\cv4pve\autosnap.conf snap --label=daily --keep=7"

Protect C:\cv4pve\autosnap.conf so that only the account of the task can read it. To keep the output, run the command through cmd /c "… >> C:\cv4pve\autosnap.log 2>&1".

Keep the schedules apart

Proxmox VE locks a guest while it takes one of its snapshots: a second snapshot of the same guest at the same moment fails. Start the jobs of different labels at different minutes (in the example the daily job runs at 01:30 and the weekly at 03:15, never on the hour of the 2-hourly job) and leave each job the time to finish before the next one.

Before an upgrade

The same command works by hand, right before a risky change, with a label of its own:

cv4pve-autosnap @/etc/cv4pve/connection.conf --vmid=1010 snap --label=preupgrade --keep=3

An option can be given only once: if the parameter file already holds --vmid, a second one on the command line stops the tool with Option '--vmid' expects a single argument but 2 were provided. Keep a file with only the connection (--host, --api-token) for runs like this one.

Watch the result

The exit code is 1 when a guest did not get its snapshot or its old ones were not removed: let your scheduler or monitoring alert on it. For more than the exit code (a message to a chat, metrics to Prometheus) use a hook script.